quantum networker cloning pdf
Quantum DX30 and ATL P-series Enhances LEGATO NetWorker™’s Protection Capabilities A NetWorker Cloning Scenario w/ Quantum ATL P-series Libraries July 7, 2003
Contents EXECUTIVE SUMMARY ...............................................................................................................................................3 INTRODUCTION............................................................................................................................................................4
The Current Problem .................................................................................................................................................4 The Solution ..............................................................................................................................................................4 Step 1: Nearline Storage: Backup to DX30 ....................................................................................................................5 Step 2: Clone to Offsite Removable Media ....................................................................................................................6 Concepts and Guidelines ...............................................................................................................................................7 LEGATO NetWorker Cloning ......................................................................................................................................9 The Quantum DX30 .....................................................................................................................................................14
Quantum DX30 Setup and Configuration ...............................................................................................................15 Archival and Disaster Recovery ..............................................................................................................................16 CONCLUSION .............................................................................................................................................................17 REFERENCES.............................................................................................................................................................17
EXECUTIVE SUMMARY Exponential data growth, 24x7 information accessibility, rising debilitating downtime costs, and restricted IT budgets place data centers under increasing pressure to protect and recover business-critical information in less time and at lower cost. In addition, recovery service levels and the need to provide a comprehensive onsite/offsite data protection plan increase the requirements of IT staff to deliver a multi-faceted backup solution with a limited budget. LEGATO Systems, Inc. and Quantum have come together to deliver a full solution to meet the backup and cloning from onsite to offsite disaster recovery requirements. Simply put, inserting disk or other faster-access/higher-reliability technologies into the existing storage architecture adds speed, reliability, and flexibility to a firm’s data protection capabilities. One very efficient way to accomplish these results is to emulate tape library functionality on a RAID-protected disk-based system. This approach optimizes a customer’s existing IT investment in both backup hardware and software, and minimizes the impact on existing backup operational procedures. The Quantum DX30 is a strong example of this new breed of Enhanced Backup systems. DX30-based solutions are designed to minimize required changes to pre-existing end user architecture and processes. Combined with the Quantum ATL P-series libraries, Enhanced Backup offers all the advantages of tape, with valued improvements in performance and capacity use efficiency. LEGATO NetWorker protects the business critical information of customers worldwide by simplifying and centralizing backup and recovery operations across UNIX, Windows, Linux, Macintosh OS X, OpenVMS and NetWare platforms. Built upon an open, highly scalable architecture, NetWorker reduces management overhead by automating the protection of storage assets in data centers and branch offices. This paper describes solutions and configurations developed by LEGATO Systems and Quantum to help enable IT managers to build backup and recovery solutions that meet ever increasing service level requirements. Jointly, we present a multi-faceted data protection solution that combines LEGATO NetWorker and its high-performance data protection capabilities with the Quantum product line that support today’s data protection and availability requirements.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 3
INTRODUCTION The Current Problem One of a corporation's greatest assets is the information, or the data, that it owns. This information is stored in the form of documents, databases, spreadsheets, etc., on hard disks at the desktop and in the data center. It is well known that backing up this data to an alternate medium is essential to protecting it from accidental erasure or corruption by a user or program and also from hardware failure. In some cases, having a backup of the data on a single tape may not be enough to meet today’s restore and disaster protection requirements. Today’s IT environments, with strong disaster recovery requirements, demand multiple copies of data with a lifecycle that takes storage media from onsite to offsite locations. Data protection today requires a multi-step process. The first step starts with near-line storage for fast, dependable recovery during a short period of time after the data is initially backed up. Then the backup image is migrated to removable, long-term media (tape media) for purposes of cost effective extended storage. The last stage is to clone or stage the onsite tape media to off site tape media for disaster recovery purposes. The Solution LEGATO NetWorker provides the ability to make duplicate copies of the data through a process called "cloning". Cloning ensures that the copies of the data are accurate, and in addition it keeps track of the information necessary to browse and recover the data. This paper will discuss the details of this cloning process, how it differs from simple tape duplication, what type of performance should be expected, and the future directions for NetWorker Cloning. The Quantum DX30, a disk-based emulation of a tape library, separates the role of the backup target from the backup archive. The Quantum DX30 utilizes the inherent strengths of disk (quick access, high-speed data transfer and RAID protection) to improve backup target functions. In turn, through LEGATO NetWorker cloning, the backup target data can easily copy to the Quantum ATL P-series library which uses its strengths (low cost, ability to write large block sizes, removable media and archival stability) to concentrate on what it does best — archival data storage. Furthermore, since a copy of the most recent backup is still available on the disk-based, enhanced backup system, recovery to the most recent point in time can occur in a minimal amount of time.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 4
Step 1: Nearline Storage: Backup to DX30 The first step to building a comprehensive onsite/offsite backup solution is to implement the initial backup. The Quantum DX30 provides optimal disk storage with a virtual tape library and tape drive interface to NetWorker. The DX30 can present itself to NetWorker as a P1000 with DLT7000 tape drives, and NetWorker will treat it as such. However, the elimination of handling tape cartridges makes administration simpler. The DX30’s compact, 4U, rackmount frame allows the location of multiple units either decentralized, in various departments, and/or centralized, in data centers. The DX30 reduces backup and restore time windows since it eliminates tape load an unload time, and increases seek speeds astronomically compared to tape. Media failure is also lower than tape-based systems, since DX30 uses RAID configuration with redundant hardware architecture. Administrative management effort is lower since the initial backup image is safely stored to the DX30 without the hassle of tape media management. Service levels for data restore increase as management requirements decrease, since error due to operator mistakes are less likely with less handling. The NetWorker administrator can implement Step 2 for offsite/removable media needs.
P-Series Library NetWorker Server
DX30
Original Backup Path
Restore Path
Available Image to Restore
Figure 1. Backup to DX30
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 5
Step 2: Clone to Offsite Removable Media Now that the data is safely stored on the DX30, it is time to move the backup image to removable media for offsite purposes. In Step 2a, the data is cloned to a Quantum, ATL P-Series tape library. This offsite copy gives the option of restoring from either the primary or the offsite location. The advantage of the tape-based off-site store is that the customer can keep a backup image on the local, DX30-based, nearline storage, but clone the backup image to removable media that can be moved offsite or exist in a remote location for disaster recovery or long term storage, as necessary. It this step, the NetWorker administrator can clone automatically at the end of the initial backup, and then immediately remove the clone or clone manually through the use of a script at a particular time of day, week, or month when tape services physically migrate tapes offsite. With the results of this step, customers can retain high, local, restore service levels and be protected in case of disaster with the off-site copies resident in the ATL P-series library.
ATL P-Series Library
NetWorker Server
DX30
Original Backup Path
Cloning Path
Restore Path
Available Image to Restore
Figure 2. Backup to DX30 then Clone to Tape
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 6
Concepts and Guidelines Developing a backup/restore strategy requires understanding the parts of the solution and cautions. Planning should include the: • Amount of data to be backed up • Priority of data • Data retention period • Restore requirements • Offsite requirements Amount of Data: The amount of data to be backed up drives any backup strategy. This information is necessary to determine amount of storage and removable media you will need. In this solution, the amount of data directly affects how cloning is performed. The amount of capacity on a DX30 is fixed and is not required to be removable since it provides large quantities of disk space in a small footprint. The only way to regain space on the DX30 is to expire previously-written backup images. The data capacity requirement helps determine the number of DX30s required, as well as the size of the tape library unit needed for cloning. Due to characteristics of RAID, the capacity required to protect the disks in the DX30 consumes various amounts of space. Useable space versus total space can also vary due to hardware redundancy requirements. Data Priority and Retention: Prioritization of data is important when determining the data protection lifecycle for a given set of data, since the priority determines restore service levels, length of backup image retention, and onsite / offsite location of backup image. High priority data is that which supports the key strategic processes within the business. Data might have high priority due to industry regulations or company policy on retention requirements. As an example, some organizations are required to retain email under regulatory or statutory requirements, such as financial services organizations or government agencies. New regulations, such as the Sarbanes-Oxley Act – passed in response to various corporate accounting scandals – and the Healthcare Insurance Portability and Accountability Act (HIPAA), raise the bar on record-keeping for a variety of entities, including public accounting firms, healthcare providers and insurance companies. Data retention requirements must be identified for each data set backed up. Similar retention and priority data sets should be kept in the same pools because data retention determines when media will be free for recycling. Data retention will also affect how much storage that is required to meet the retention time of a data. Higher priority data requires a higher disaster recovery level that often implies immediate long-term, off-site storage and longer nearline storage for quick restore. For example, high priority data, in the scenario described in this document would be cloned immediately from the DX30 to the P-series Library after the primary backup completes. NetWorker cloning is used to create multiple copies and meet different onsite/offsite retention requirements. Using nearline storage to hold data with low importance is inefficient. Data with low priority rarely has offsite requirements and should be archived to the P-series library and expired from the DX30 when no longer needed. Keeping the low priority data separate from all other data will allow for clean expiration and recycling of storage resources used by this data. The medium priority data, that falling in between the low and high priority will be moved to removable media quicker and possibly held in the library longer than data needing immediate offsite removal. The medium-priority data can often be cloned to removable media from the DX30 and then removed to offsite storage as time allows.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 7
Restore Service Levels: Restore service level is the window of time allowed between the end user request for recovery and the restoration of the data to its original location. The strategic importance of each data set determines the proper restore service level, which can be described as high, medium, or low. A high restore service level usually applies to databases, applications, and immediate need data. Without this data, a company’s key strategic processes cannot proceed, in part or in whole. Restore of this data must be performed immediately after corruption, failure, or data loss takes place. A backup image of this kind of data would be kept on the DX30 to facilitate quick recovery that bypasses the media handling and tape positioning time investments associated with removable tape media. The medium restore service level is that data that must be restored with less urgency than the high service level, but restore should not be delayed by having to call back removable media from offsite or require media insertion into a tape library from onsite storage facilities. These delays would be too long and add operator time and resource expenditure with tape handling. This data could be comprised of project spreadsheets, customer information, web site data, etc. The low restore service level is data that doesn’t have immediate recover needs and offsite recall of the backup medium is acceptable. This data could be user home directories, long-term email data, long term transaction data, etc. This data is to be kept for record purposes in case data must be searched or identified even after it has become useless to daily operations. Offsite Requirements: Disaster recovery and long-term storage needs dictate the need for off-site storage requirements. For over 30 years, companies have relied on off-site copies of backup data on tape to reduce risk of data loss due to inaccessibility at the local site. Offsite practices can exploit cloning to get the backup image from nearline storage to removable media, while also keeping it on nearline storage for quick recovery. Immediate deployment of off-site copies to offsite storage is necessary for purposes of recovering high priority data from a site disaster. This offsite operation can be accomplished using a WAN or MAN connection between the primary location and the offsite location. For example, LEGATO NetWorker can clone the backup data over a WAN to a P-series library in the off-site location. The Quantum ATL P-series libraries are particularly well suited to these scenarios since they can be natively deployed to a SCSI, Fibre Channel or IP Storage environment
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 8
LEGATO NetWorker Cloning In NetWorker, cloning is defined as a duplicate of the data (saveset), not of the physical backup medium. Cloning does not duplicate entire volumes, but rather specific data on volumes. NetWorker sends data from the client filesystem and writes it to tape in the form of a "saveset". A saveset is defined as a group of files or a filesystem from a single client computer backed up onto storage media. In the design phase of the cloning process, Legato chose a design whereby the data would be cloned at the saveset level, rather than at the volume level. Some of the rational would include the following: 1) Volume duplication is difficult to track. With all volumes having the same identifying labels, they cannot be distinguished from one another. 2) Savesets can span backup volumes, so duplicating a volume containing the beginning of that saveset without the volume containing the end of that saveset can result in loss of data. 3) The backup volume or physical tape cartridge is simply a medium for the data. Because a backup volume can contain data from many different sources (clients), from different periods of time, there may be situations where backup volumes contain data not suitable for cloning - i.e. aborted savesets, less important clients, etc. By cloning at the saveset level, the data for particular clients and filesystems can be singled out for cloning, while other less important or invalid data can be ignored, avoiding unnecessary duplications. 4) Physical tapes do not hold the same exact amount of data. If the destination tape holds less data than the source tape, then multiple tapes will be required. The second tape will probably not be filled to capacity. The remaining space would be due to the tapes association with a single source volume. Statistically, if no two tapes are the same length, then 50% of the time, the destination will be shorter than the source. By cloning at the saveset level, NetWorker ensures that the data is cloned reliably and efficiently, and accurately tracked and managed in the media database. Tracking cloned data During a normal backup session, NetWorker tracks each backup volume, saveset and file, recording it in the Media Database and Client File Index (CFI). These two files form a relational database, using the Saveset ID (SSID) as the unique key between them. During cloning, a new entry for the clone volume will be recorded in the Media DB, but NetWorker will use the original saveset entries in the media DB, and the original file entries in the CFI for both the original and the clone volume(s). It is possible to clone a clone saveset, or to clone the original saveset multiple times, so NetWorker keeps track of the number of clones that a saveset has, and which volumes they are contained on. Since the purpose of cloning is to create separate duplicate copies of data, NetWorker ensures that each copy of a saveset will be written to a unique backup volume, so each copy can be stored in separate physical locations. The volume that will be cloned to must be assigned to a special pool, called a clone pool, at labeling time to differentiate it from the original. Since the saveset and file index entries are shared between volumes, the effect of the browse and retention policy, as well as volume deleting, purging, or recycling must be understood to avoid inadvertently losing index entries. Taking action on the Media DB or CFI entries of a volume will also affect any clones or originals of that volume. Purging the index entries for the original volume also purges those for the clone. When the original volume becomes recyclable, by definition the clone(s) are recyclable as well. LEGATO NetWorker cloning streamlines the process of creating copies of NetWorker backup images. Two types of cloning are available, scheduled and manual. Scheduled cloning is configured through the NetWorker Administration GUI. Manual cloning can be initiated via the same GUI, or through the command line interface (CLI). Scheduled clones are started automatically at the end of a savegroup. This is configured in the group resource by changing the "clones" attribute to yes, and selecting a destination clone pool. Once the savegroup has completed the saves of its clients, NetWorker will begin cloning all completed savesets contained in that group. Scheduled cloning may lead to device contention in environments where multiple savegroups overlap. When a cloning process begins while other saves are running, the saves and clones may compete for device or volume resources. This potential contention needs to be taken into consideration when implementing scheduled cloning. Managing device contention will be discussed later in this document. LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 9
Manual cloning through the GUI is provided by the "clone volume" or "clone savesets" options. As stated above, the cloning process is performed at the saveset level. The "clone volume" option also clones at the saveset level, but it allows the user to clone all the savesets belonging to a particular volume. Previous to NetWorker 7.0, the "clone volume" feature will clone any saveset that has a fragment, or partial saveset, contained on the source volume. This included savesets that spanned from other volumes, or to other volumes. This behavior changes in NetWorker 7 and later, to clone only savesets that begin on the specified volume. Savesets spanning from a previous volume will not be cloned. With the various options detailed above, LEGATO NetWorker cloning can be tailored to answer the needs of any environment. Coupled with the Quantum DX30 and ATL P-series libraries, LEGATO NetWorker cloning answers the challenge of having a reliable solution for all aspects of data protection, from near-line storage for fast recovery to off-site storage for disaster recovery and regulatory compliance.
Figure 3. Clone Volume Window in the NetWorker GUI The "clone savesets" feature lets you specify individual or multiple savesets for cloning. This option allows you to specify these savesets based on any combination of the following criteria: • client • saveset name • volume name • pool • date • status • level LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 10
Figure 4. Clone Save Set Window in the NetWorker GUI The command line interface to cloning, (nsrclone), provides options for cloning by SSID's or by Volume. In addition, there is an option to use a text file as the input file to nsrclone. This option is useful in scripting where the savesets are selected, using custom specifications, from the output of the mminfo command. Regardless of the method used to start the cloning process, a list of savesets is ultimately passed to the nsrclone process. The nsrclone process uses this list to determine which volumes are necessary to complete the task. Nsrclone determines which host it will use to mount the source volume. The clone process can clone multiplexed savestreams to the new volume. The clone process reads and writes each saveset as encountered on the source volume during a single pass through the volume. For data verification purposes, scanner or recover are the best tools available to verify that all the data on the volume is recoverable. Volume spanning is handled just as during a routine save. If the destination volume fills before the cloning process is completed, then a new volume will be requested to and the cloning will continue to that volume when it is available. If the volume to be cloned is already mounted in a device, then NetWorker will leave that volume mounted and clone it from where it is mounted.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 11
If the source volume is not in a shared library, then NetWorker will mount the source volume on the NetWorker server or storage node that owns the library containing that volume. Where the destination clone volume is mounted will be determined by querying the "Clone Storage Node" affinity list of the client resource for the source host. If the source volume is in a shared library, then all the hosts/storage nodes sharing the library have access to the source volume. Because the "clone storage node" attribute is an attribute of the storage nodes' client resource, each storage node having access to the shared library will have its own "clone storage node" value(s). This renders the "clone storage node" attribute an invalid attribute to base the selection of the source host on. With the introduction of library sharing functionality in NetWorker 5.5.3, the jukebox resource has a new attribute, the 'read hostname' attribute, which is used to determine which host will mount the source volume. The "clone storage node" attribute is then used to determine where to mount the destination volume. In the current implementation of NetWorker, the 'read hostname' can define only one host. The result of this is that all cloning (in a shared library environment), will occur on that host. The cloning process adds very little overhead to the device reads and writes. However, moving data from disk to tape is generally faster than from tape to tape, due to the differences in the read speeds of a disk versus a tape device. LEGATO's testing indicates that disk to tape cloning is approx 30% faster than the tape to tape cloning. Tests included a single clone process running local backups and clones, (immediate save/cloning technology enabled), with non-spanning savesets (eliminates volume mounting/unmounting and device contention), of up 8 streams multiplexed, with no other activity being performed on the machine doing the cloning.
1 Stream 4 Streams (GUI) 8 Streams (GUI)
Backup time in sec's
Clone time in sec's
319 1260 2516
416 1673 3407
Time gained by Cloning from Disk versus tape 30 % faster 33% faster 33% faster
Table 1: Backup time vs. Clone time However, the performance in the real world environment can be affected by many variables. A few of the more commonly seen scenarios are discussed below. Nsrclone will clone multiplexed savesets in a single pass through the volume only if all of the SSID's are passed to a single nsrclone process. Starting four nsrclone processes for four savesets that are multiplexed will result in the backup volume being read four separate times. Scheduled and volume cloning will facilitate this automatically, however cloning from the command line requires them to be manually entered on the same command line. If multiple savesets were multiplexed to a tape volume, and only a subset of those savesets are cloned, then the cloning process will be required to read all the savesets on the volume, while writing only the subset passed to the nsrclone process. This will result in slower write speeds to the destination device, as more data is being read than is being written. Device contention can also be caused by multiplexed savesets spanning tapes. For instance, consider an environment with two backup sessions, each writing four savesets per backup volume (target sessions set to four). Two savesets on each of these volumes (tape A and tape B) complete while the other four savesets (two per volume) continue until tapes A & B are full. NetWorker will now see that there are only four savesets running, and will put all four to the same tape (tape C).
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 12
During the attempt to clone these savesets, two nsrclone processes will be started, one for tape A, and one for tape B. When the first nsrclone process reaches the end of its first backup volume, it will automatically mount tape C, to clone the remainder of its savesets. When the second nsrclone reaches the end of its first backup volume, it will also need data from tape C, and will be forced to wait until the first nsrclone process is done reading from that tape. In a large environment, this can lead to many nsrclone processes sitting idle for long periods of time waiting for access to volumes that are in use. This can lead to significant increases in the total time necessary to complete the entire cloning process. Another example of device contention would be if saveset one spans volume A & B, while saveset two begins on volume B. If two nsrclone processes (one for each saveset) are invoked, then when clone process for volume A completes cloning the saveset fragment on volume A, it will not be able to access volume B until the clone process for volume B is complete. Scheduled cloning can also encounter device contention between nsrclone processes and save processes. By definition, scheduled cloning is invoked after a savegroup completes its backup tasks. If the environment consists of savegroups that overlap, both save and nsrclone processes could be competing for the same devices. This drive contention can stall processes for long periods of time, impacting the perceived performance of the backup or cloning processes.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 13
The Quantum DX30 Backup jobs are done directly to the Quantum DX30 (the backup target), which can reach backup speeds in excess of 288 GB/hour. The software platform designed into the DX30 emulates a tape library (with two to six tape drives), allowing your backup software to function without modification. The DX30 uses RAID-protected disks to increase your confidence (beyond 99%) that the backup has been completed correctly and can be restored when needed. After the backup has completed, the IT administrator can export a copy of the backup data, using the cloning or copy capabilities of LEGATO NetWorker, from the DX30 to the ATL P4000 or ATL P7000 library for archival storage. With the DX30, the customer now has the opportunity to conceptually divide the backup system into two distinct functions—the backup target (high performance and reliability are key characteristics) and the backup archive (requires removable and transportable media). The DX30 addresses the three major concerns that have been expressed by customers: •
The DX30 reduces the time required for backup. A single DX30 system will transfer data at a rate of at least 288GB/hr, which is equivalent to 80MB/second.
•
The DX30 significantly increases the confidence of completing the backup within the allocated backup window. Today, based on the typical number of reported failed backups per year, customers are experiencing 90% to 95% confidence that they can complete their backups on any given night. Typically, as long as the backup software and the applications are appropriately configured, the most common reason why backups fail to complete is because of anomalies with tape drives and tape media. With the DX30 Enhanced Backup System, since it is based on an array of parity-protected disk drives, Quantum can improve the confidence to well over 99% that the backup target will not cause the backup to fail to complete within the backup window.
•
The DX30 reduces the latency of restoring data. With a tape library, it takes minutes before you can begin to restore a file. The fastest possible restore times with a tape library are directly determined by how long it takes to mount a required cartridge in a tape drive and seek to the proper position on the tape to begin restore. It can be several minutes before a restore is started and even then the data transfer is limited by the speed of tape. Exponentially compounding the problem is the case where the required cartridge has been moved off-site for disaster recovery purposes. Furthermore, the time it takes to restore a large amount of data from tape can far exceed service level agreements due to its sequential nature and a potentially high number of cartridge swaps required to complete the restore. With the DX30, the latency to restore is less than 10 seconds in the worst case. The restore latency is on the order of seconds—not milliseconds, like it is for disk arrays that are designed for primary storage. In addition, the enhanced backup system allows the customer to preserve the investments they have already made in backup hardware, software, operational procedures and operator training.
The best way to think of the backup target is to imagine an ATL P1000 tape library that is filled with cartridges and has its front doors and load port permanently locked shut. In this case, the backup software package can back data up to the ATL P1000. When the capacity starts filling up with backup images it can expire some of the backup images and re-use the virtual cartridges for new backups. In this case, the ATL P1000 is used as a closed repository for backup images.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 14
Quantum DX30 Setup and Configuration The DX30 is a native fibre channel device that can either be directly connected to the backup host, or into the fabric, allowing access from multiple storage nodes. Since the DX30 has a compact frame (4U), it can be put into an existing rack and doesn’t require its own floor space. The solution has been tested with multiple fibre channel switches and host bus adapters (HBA) running on a variety of operating systems. When the DX30 is installed the configuration also includes the number of virtual tape drives and cartridges. Setting up a DX30 with multiple tape drives allows for better optimization by taking advantage of NetWorker’s adjustable parameters such as parallelism and target sessions. The virtual tape drive and cartridge count settings can be changed from the web GUI. NOTE: Changing the number and/or size of the cartridges after the initial installation will cause loss of data. It is recommended to clone the data to another source that can be another DX30 or to removable media. A web interface for management is also available. The web interface can be used to configure email alert and network settings. From the web interface the user can monitor multiple aspects of the device including online status, virtual tape usage, and hardware components. Newer versions of DX30 software can also be loaded through this interface.
Figure 5. DX30 Home Page Interface
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 15
The Quantum ATL P-series Libraries for Archiving and Disaster Recovery Tape libraries are the ideal targets for archival and off-site data. Cloning from the DX30 to the ATL P4000 or ATL P7000 for on-site archival provides the most efficient way to protect medium- and low-priority data. Cloning from the DX30 to a remote ATL P-series library enables an efficient and effective disaster recovery infrastructure. Using the strategy outlined in this paper, high priority data exists in both the Quantum DX30 (available for quick restore) and in local and remote ATL P-series libraries (archived or off-site for full disaster recovery protection). The administrator has the flexibility to choose where and how long to keep the data — based on the data restoration needs. Tape libraries such as the Quantum ATL P4000 or P7000 are the ideal systems for fulfilling the backup archive function for three primary reasons: • The ability to remove and transport media. • Can be directly attached to a server via SCSI or Fibre Channel, or deployed into a Fibre Channel SAN, or work in an IP Storage environment. • Reliability of a platform that exploits common, well-known and time-tested practices, that has proven over 3 million swaps, that is a proven architecture and that has a large install base. The combination of DX30, which handles 3-6 Tb of backup data, and ATL P-series libraries, which handles over 500 Tb of backup data, allows customers to focus their purchase. They can purchase enhanced backup to handle their strategic data, while backing up low priority and some medium priority data directly to the P-series library. This schema produces very cost effective solutions from a cost/benefit standpoint. Investment is applied to reduce risk in the areas of the most potential impact.
Figure 6. DX30 Data Cartridge Status LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 16
CONCLUSION Leveraging the advanced management capability of LEGATO NetWorker and Quantum’s full line of storage devices, this cloning-based, enhanced backup solution delivers solid data protection. This data protection strategy supports the full onsite/offsite data retention lifecycle required to meet industry regulations, disaster recovery protection, and restore service level agreements. LEGATO and Quantum have teamed up to deliver this robust solution with an emphasis on manageability, reliability, and keeping IT administration and resources to a minimum. Using LEGATO NetWorker to backup to the Quantum DX30 disked based virtual tape library, companies can disperse backup targets throughout departments or data centers that defer the investment of time needed for tape handling in this initial, time-constrained step. NetWorker can then clone the backup images to removable media in Quantum’s ATL P-series libraries for local archiving and off-site storage purposes. An IT department can implement LEGATO NetWorker and Quantum’s disk and tape hardware to meet its company’s need for data protection within a reasonable cost/benefit framework.
REFERENCES LEGATO NetWorker Administrator's Guide, Release 7.0, Microsoft Windows Version LEGATO NetWorker Installation Guide, Release 7.0, Microsoft Windows Version LEGATO NetWorker Release Supplement, Release 7.0, Windows Version Quantum DX30 Users Guide, Release 6494002v2r1
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 17
Contents EXECUTIVE SUMMARY ...............................................................................................................................................3 INTRODUCTION............................................................................................................................................................4
The Current Problem .................................................................................................................................................4 The Solution ..............................................................................................................................................................4 Step 1: Nearline Storage: Backup to DX30 ....................................................................................................................5 Step 2: Clone to Offsite Removable Media ....................................................................................................................6 Concepts and Guidelines ...............................................................................................................................................7 LEGATO NetWorker Cloning ......................................................................................................................................9 The Quantum DX30 .....................................................................................................................................................14
Quantum DX30 Setup and Configuration ...............................................................................................................15 Archival and Disaster Recovery ..............................................................................................................................16 CONCLUSION .............................................................................................................................................................17 REFERENCES.............................................................................................................................................................17
EXECUTIVE SUMMARY Exponential data growth, 24x7 information accessibility, rising debilitating downtime costs, and restricted IT budgets place data centers under increasing pressure to protect and recover business-critical information in less time and at lower cost. In addition, recovery service levels and the need to provide a comprehensive onsite/offsite data protection plan increase the requirements of IT staff to deliver a multi-faceted backup solution with a limited budget. LEGATO Systems, Inc. and Quantum have come together to deliver a full solution to meet the backup and cloning from onsite to offsite disaster recovery requirements. Simply put, inserting disk or other faster-access/higher-reliability technologies into the existing storage architecture adds speed, reliability, and flexibility to a firm’s data protection capabilities. One very efficient way to accomplish these results is to emulate tape library functionality on a RAID-protected disk-based system. This approach optimizes a customer’s existing IT investment in both backup hardware and software, and minimizes the impact on existing backup operational procedures. The Quantum DX30 is a strong example of this new breed of Enhanced Backup systems. DX30-based solutions are designed to minimize required changes to pre-existing end user architecture and processes. Combined with the Quantum ATL P-series libraries, Enhanced Backup offers all the advantages of tape, with valued improvements in performance and capacity use efficiency. LEGATO NetWorker protects the business critical information of customers worldwide by simplifying and centralizing backup and recovery operations across UNIX, Windows, Linux, Macintosh OS X, OpenVMS and NetWare platforms. Built upon an open, highly scalable architecture, NetWorker reduces management overhead by automating the protection of storage assets in data centers and branch offices. This paper describes solutions and configurations developed by LEGATO Systems and Quantum to help enable IT managers to build backup and recovery solutions that meet ever increasing service level requirements. Jointly, we present a multi-faceted data protection solution that combines LEGATO NetWorker and its high-performance data protection capabilities with the Quantum product line that support today’s data protection and availability requirements.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 3
INTRODUCTION The Current Problem One of a corporation's greatest assets is the information, or the data, that it owns. This information is stored in the form of documents, databases, spreadsheets, etc., on hard disks at the desktop and in the data center. It is well known that backing up this data to an alternate medium is essential to protecting it from accidental erasure or corruption by a user or program and also from hardware failure. In some cases, having a backup of the data on a single tape may not be enough to meet today’s restore and disaster protection requirements. Today’s IT environments, with strong disaster recovery requirements, demand multiple copies of data with a lifecycle that takes storage media from onsite to offsite locations. Data protection today requires a multi-step process. The first step starts with near-line storage for fast, dependable recovery during a short period of time after the data is initially backed up. Then the backup image is migrated to removable, long-term media (tape media) for purposes of cost effective extended storage. The last stage is to clone or stage the onsite tape media to off site tape media for disaster recovery purposes. The Solution LEGATO NetWorker provides the ability to make duplicate copies of the data through a process called "cloning". Cloning ensures that the copies of the data are accurate, and in addition it keeps track of the information necessary to browse and recover the data. This paper will discuss the details of this cloning process, how it differs from simple tape duplication, what type of performance should be expected, and the future directions for NetWorker Cloning. The Quantum DX30, a disk-based emulation of a tape library, separates the role of the backup target from the backup archive. The Quantum DX30 utilizes the inherent strengths of disk (quick access, high-speed data transfer and RAID protection) to improve backup target functions. In turn, through LEGATO NetWorker cloning, the backup target data can easily copy to the Quantum ATL P-series library which uses its strengths (low cost, ability to write large block sizes, removable media and archival stability) to concentrate on what it does best — archival data storage. Furthermore, since a copy of the most recent backup is still available on the disk-based, enhanced backup system, recovery to the most recent point in time can occur in a minimal amount of time.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 4
Step 1: Nearline Storage: Backup to DX30 The first step to building a comprehensive onsite/offsite backup solution is to implement the initial backup. The Quantum DX30 provides optimal disk storage with a virtual tape library and tape drive interface to NetWorker. The DX30 can present itself to NetWorker as a P1000 with DLT7000 tape drives, and NetWorker will treat it as such. However, the elimination of handling tape cartridges makes administration simpler. The DX30’s compact, 4U, rackmount frame allows the location of multiple units either decentralized, in various departments, and/or centralized, in data centers. The DX30 reduces backup and restore time windows since it eliminates tape load an unload time, and increases seek speeds astronomically compared to tape. Media failure is also lower than tape-based systems, since DX30 uses RAID configuration with redundant hardware architecture. Administrative management effort is lower since the initial backup image is safely stored to the DX30 without the hassle of tape media management. Service levels for data restore increase as management requirements decrease, since error due to operator mistakes are less likely with less handling. The NetWorker administrator can implement Step 2 for offsite/removable media needs.
P-Series Library NetWorker Server
DX30
Original Backup Path
Restore Path
Available Image to Restore
Figure 1. Backup to DX30
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 5
Step 2: Clone to Offsite Removable Media Now that the data is safely stored on the DX30, it is time to move the backup image to removable media for offsite purposes. In Step 2a, the data is cloned to a Quantum, ATL P-Series tape library. This offsite copy gives the option of restoring from either the primary or the offsite location. The advantage of the tape-based off-site store is that the customer can keep a backup image on the local, DX30-based, nearline storage, but clone the backup image to removable media that can be moved offsite or exist in a remote location for disaster recovery or long term storage, as necessary. It this step, the NetWorker administrator can clone automatically at the end of the initial backup, and then immediately remove the clone or clone manually through the use of a script at a particular time of day, week, or month when tape services physically migrate tapes offsite. With the results of this step, customers can retain high, local, restore service levels and be protected in case of disaster with the off-site copies resident in the ATL P-series library.
ATL P-Series Library
NetWorker Server
DX30
Original Backup Path
Cloning Path
Restore Path
Available Image to Restore
Figure 2. Backup to DX30 then Clone to Tape
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 6
Concepts and Guidelines Developing a backup/restore strategy requires understanding the parts of the solution and cautions. Planning should include the: • Amount of data to be backed up • Priority of data • Data retention period • Restore requirements • Offsite requirements Amount of Data: The amount of data to be backed up drives any backup strategy. This information is necessary to determine amount of storage and removable media you will need. In this solution, the amount of data directly affects how cloning is performed. The amount of capacity on a DX30 is fixed and is not required to be removable since it provides large quantities of disk space in a small footprint. The only way to regain space on the DX30 is to expire previously-written backup images. The data capacity requirement helps determine the number of DX30s required, as well as the size of the tape library unit needed for cloning. Due to characteristics of RAID, the capacity required to protect the disks in the DX30 consumes various amounts of space. Useable space versus total space can also vary due to hardware redundancy requirements. Data Priority and Retention: Prioritization of data is important when determining the data protection lifecycle for a given set of data, since the priority determines restore service levels, length of backup image retention, and onsite / offsite location of backup image. High priority data is that which supports the key strategic processes within the business. Data might have high priority due to industry regulations or company policy on retention requirements. As an example, some organizations are required to retain email under regulatory or statutory requirements, such as financial services organizations or government agencies. New regulations, such as the Sarbanes-Oxley Act – passed in response to various corporate accounting scandals – and the Healthcare Insurance Portability and Accountability Act (HIPAA), raise the bar on record-keeping for a variety of entities, including public accounting firms, healthcare providers and insurance companies. Data retention requirements must be identified for each data set backed up. Similar retention and priority data sets should be kept in the same pools because data retention determines when media will be free for recycling. Data retention will also affect how much storage that is required to meet the retention time of a data. Higher priority data requires a higher disaster recovery level that often implies immediate long-term, off-site storage and longer nearline storage for quick restore. For example, high priority data, in the scenario described in this document would be cloned immediately from the DX30 to the P-series Library after the primary backup completes. NetWorker cloning is used to create multiple copies and meet different onsite/offsite retention requirements. Using nearline storage to hold data with low importance is inefficient. Data with low priority rarely has offsite requirements and should be archived to the P-series library and expired from the DX30 when no longer needed. Keeping the low priority data separate from all other data will allow for clean expiration and recycling of storage resources used by this data. The medium priority data, that falling in between the low and high priority will be moved to removable media quicker and possibly held in the library longer than data needing immediate offsite removal. The medium-priority data can often be cloned to removable media from the DX30 and then removed to offsite storage as time allows.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 7
Restore Service Levels: Restore service level is the window of time allowed between the end user request for recovery and the restoration of the data to its original location. The strategic importance of each data set determines the proper restore service level, which can be described as high, medium, or low. A high restore service level usually applies to databases, applications, and immediate need data. Without this data, a company’s key strategic processes cannot proceed, in part or in whole. Restore of this data must be performed immediately after corruption, failure, or data loss takes place. A backup image of this kind of data would be kept on the DX30 to facilitate quick recovery that bypasses the media handling and tape positioning time investments associated with removable tape media. The medium restore service level is that data that must be restored with less urgency than the high service level, but restore should not be delayed by having to call back removable media from offsite or require media insertion into a tape library from onsite storage facilities. These delays would be too long and add operator time and resource expenditure with tape handling. This data could be comprised of project spreadsheets, customer information, web site data, etc. The low restore service level is data that doesn’t have immediate recover needs and offsite recall of the backup medium is acceptable. This data could be user home directories, long-term email data, long term transaction data, etc. This data is to be kept for record purposes in case data must be searched or identified even after it has become useless to daily operations. Offsite Requirements: Disaster recovery and long-term storage needs dictate the need for off-site storage requirements. For over 30 years, companies have relied on off-site copies of backup data on tape to reduce risk of data loss due to inaccessibility at the local site. Offsite practices can exploit cloning to get the backup image from nearline storage to removable media, while also keeping it on nearline storage for quick recovery. Immediate deployment of off-site copies to offsite storage is necessary for purposes of recovering high priority data from a site disaster. This offsite operation can be accomplished using a WAN or MAN connection between the primary location and the offsite location. For example, LEGATO NetWorker can clone the backup data over a WAN to a P-series library in the off-site location. The Quantum ATL P-series libraries are particularly well suited to these scenarios since they can be natively deployed to a SCSI, Fibre Channel or IP Storage environment
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 8
LEGATO NetWorker Cloning In NetWorker, cloning is defined as a duplicate of the data (saveset), not of the physical backup medium. Cloning does not duplicate entire volumes, but rather specific data on volumes. NetWorker sends data from the client filesystem and writes it to tape in the form of a "saveset". A saveset is defined as a group of files or a filesystem from a single client computer backed up onto storage media. In the design phase of the cloning process, Legato chose a design whereby the data would be cloned at the saveset level, rather than at the volume level. Some of the rational would include the following: 1) Volume duplication is difficult to track. With all volumes having the same identifying labels, they cannot be distinguished from one another. 2) Savesets can span backup volumes, so duplicating a volume containing the beginning of that saveset without the volume containing the end of that saveset can result in loss of data. 3) The backup volume or physical tape cartridge is simply a medium for the data. Because a backup volume can contain data from many different sources (clients), from different periods of time, there may be situations where backup volumes contain data not suitable for cloning - i.e. aborted savesets, less important clients, etc. By cloning at the saveset level, the data for particular clients and filesystems can be singled out for cloning, while other less important or invalid data can be ignored, avoiding unnecessary duplications. 4) Physical tapes do not hold the same exact amount of data. If the destination tape holds less data than the source tape, then multiple tapes will be required. The second tape will probably not be filled to capacity. The remaining space would be due to the tapes association with a single source volume. Statistically, if no two tapes are the same length, then 50% of the time, the destination will be shorter than the source. By cloning at the saveset level, NetWorker ensures that the data is cloned reliably and efficiently, and accurately tracked and managed in the media database. Tracking cloned data During a normal backup session, NetWorker tracks each backup volume, saveset and file, recording it in the Media Database and Client File Index (CFI). These two files form a relational database, using the Saveset ID (SSID) as the unique key between them. During cloning, a new entry for the clone volume will be recorded in the Media DB, but NetWorker will use the original saveset entries in the media DB, and the original file entries in the CFI for both the original and the clone volume(s). It is possible to clone a clone saveset, or to clone the original saveset multiple times, so NetWorker keeps track of the number of clones that a saveset has, and which volumes they are contained on. Since the purpose of cloning is to create separate duplicate copies of data, NetWorker ensures that each copy of a saveset will be written to a unique backup volume, so each copy can be stored in separate physical locations. The volume that will be cloned to must be assigned to a special pool, called a clone pool, at labeling time to differentiate it from the original. Since the saveset and file index entries are shared between volumes, the effect of the browse and retention policy, as well as volume deleting, purging, or recycling must be understood to avoid inadvertently losing index entries. Taking action on the Media DB or CFI entries of a volume will also affect any clones or originals of that volume. Purging the index entries for the original volume also purges those for the clone. When the original volume becomes recyclable, by definition the clone(s) are recyclable as well. LEGATO NetWorker cloning streamlines the process of creating copies of NetWorker backup images. Two types of cloning are available, scheduled and manual. Scheduled cloning is configured through the NetWorker Administration GUI. Manual cloning can be initiated via the same GUI, or through the command line interface (CLI). Scheduled clones are started automatically at the end of a savegroup. This is configured in the group resource by changing the "clones" attribute to yes, and selecting a destination clone pool. Once the savegroup has completed the saves of its clients, NetWorker will begin cloning all completed savesets contained in that group. Scheduled cloning may lead to device contention in environments where multiple savegroups overlap. When a cloning process begins while other saves are running, the saves and clones may compete for device or volume resources. This potential contention needs to be taken into consideration when implementing scheduled cloning. Managing device contention will be discussed later in this document. LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 9
Manual cloning through the GUI is provided by the "clone volume" or "clone savesets" options. As stated above, the cloning process is performed at the saveset level. The "clone volume" option also clones at the saveset level, but it allows the user to clone all the savesets belonging to a particular volume. Previous to NetWorker 7.0, the "clone volume" feature will clone any saveset that has a fragment, or partial saveset, contained on the source volume. This included savesets that spanned from other volumes, or to other volumes. This behavior changes in NetWorker 7 and later, to clone only savesets that begin on the specified volume. Savesets spanning from a previous volume will not be cloned. With the various options detailed above, LEGATO NetWorker cloning can be tailored to answer the needs of any environment. Coupled with the Quantum DX30 and ATL P-series libraries, LEGATO NetWorker cloning answers the challenge of having a reliable solution for all aspects of data protection, from near-line storage for fast recovery to off-site storage for disaster recovery and regulatory compliance.
Figure 3. Clone Volume Window in the NetWorker GUI The "clone savesets" feature lets you specify individual or multiple savesets for cloning. This option allows you to specify these savesets based on any combination of the following criteria: • client • saveset name • volume name • pool • date • status • level LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 10
Figure 4. Clone Save Set Window in the NetWorker GUI The command line interface to cloning, (nsrclone), provides options for cloning by SSID's or by Volume. In addition, there is an option to use a text file as the input file to nsrclone. This option is useful in scripting where the savesets are selected, using custom specifications, from the output of the mminfo command. Regardless of the method used to start the cloning process, a list of savesets is ultimately passed to the nsrclone process. The nsrclone process uses this list to determine which volumes are necessary to complete the task. Nsrclone determines which host it will use to mount the source volume. The clone process can clone multiplexed savestreams to the new volume. The clone process reads and writes each saveset as encountered on the source volume during a single pass through the volume. For data verification purposes, scanner or recover are the best tools available to verify that all the data on the volume is recoverable. Volume spanning is handled just as during a routine save. If the destination volume fills before the cloning process is completed, then a new volume will be requested to and the cloning will continue to that volume when it is available. If the volume to be cloned is already mounted in a device, then NetWorker will leave that volume mounted and clone it from where it is mounted.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 11
If the source volume is not in a shared library, then NetWorker will mount the source volume on the NetWorker server or storage node that owns the library containing that volume. Where the destination clone volume is mounted will be determined by querying the "Clone Storage Node" affinity list of the client resource for the source host. If the source volume is in a shared library, then all the hosts/storage nodes sharing the library have access to the source volume. Because the "clone storage node" attribute is an attribute of the storage nodes' client resource, each storage node having access to the shared library will have its own "clone storage node" value(s). This renders the "clone storage node" attribute an invalid attribute to base the selection of the source host on. With the introduction of library sharing functionality in NetWorker 5.5.3, the jukebox resource has a new attribute, the 'read hostname' attribute, which is used to determine which host will mount the source volume. The "clone storage node" attribute is then used to determine where to mount the destination volume. In the current implementation of NetWorker, the 'read hostname' can define only one host. The result of this is that all cloning (in a shared library environment), will occur on that host. The cloning process adds very little overhead to the device reads and writes. However, moving data from disk to tape is generally faster than from tape to tape, due to the differences in the read speeds of a disk versus a tape device. LEGATO's testing indicates that disk to tape cloning is approx 30% faster than the tape to tape cloning. Tests included a single clone process running local backups and clones, (immediate save/cloning technology enabled), with non-spanning savesets (eliminates volume mounting/unmounting and device contention), of up 8 streams multiplexed, with no other activity being performed on the machine doing the cloning.
1 Stream 4 Streams (GUI) 8 Streams (GUI)
Backup time in sec's
Clone time in sec's
319 1260 2516
416 1673 3407
Time gained by Cloning from Disk versus tape 30 % faster 33% faster 33% faster
Table 1: Backup time vs. Clone time However, the performance in the real world environment can be affected by many variables. A few of the more commonly seen scenarios are discussed below. Nsrclone will clone multiplexed savesets in a single pass through the volume only if all of the SSID's are passed to a single nsrclone process. Starting four nsrclone processes for four savesets that are multiplexed will result in the backup volume being read four separate times. Scheduled and volume cloning will facilitate this automatically, however cloning from the command line requires them to be manually entered on the same command line. If multiple savesets were multiplexed to a tape volume, and only a subset of those savesets are cloned, then the cloning process will be required to read all the savesets on the volume, while writing only the subset passed to the nsrclone process. This will result in slower write speeds to the destination device, as more data is being read than is being written. Device contention can also be caused by multiplexed savesets spanning tapes. For instance, consider an environment with two backup sessions, each writing four savesets per backup volume (target sessions set to four). Two savesets on each of these volumes (tape A and tape B) complete while the other four savesets (two per volume) continue until tapes A & B are full. NetWorker will now see that there are only four savesets running, and will put all four to the same tape (tape C).
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 12
During the attempt to clone these savesets, two nsrclone processes will be started, one for tape A, and one for tape B. When the first nsrclone process reaches the end of its first backup volume, it will automatically mount tape C, to clone the remainder of its savesets. When the second nsrclone reaches the end of its first backup volume, it will also need data from tape C, and will be forced to wait until the first nsrclone process is done reading from that tape. In a large environment, this can lead to many nsrclone processes sitting idle for long periods of time waiting for access to volumes that are in use. This can lead to significant increases in the total time necessary to complete the entire cloning process. Another example of device contention would be if saveset one spans volume A & B, while saveset two begins on volume B. If two nsrclone processes (one for each saveset) are invoked, then when clone process for volume A completes cloning the saveset fragment on volume A, it will not be able to access volume B until the clone process for volume B is complete. Scheduled cloning can also encounter device contention between nsrclone processes and save processes. By definition, scheduled cloning is invoked after a savegroup completes its backup tasks. If the environment consists of savegroups that overlap, both save and nsrclone processes could be competing for the same devices. This drive contention can stall processes for long periods of time, impacting the perceived performance of the backup or cloning processes.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 13
The Quantum DX30 Backup jobs are done directly to the Quantum DX30 (the backup target), which can reach backup speeds in excess of 288 GB/hour. The software platform designed into the DX30 emulates a tape library (with two to six tape drives), allowing your backup software to function without modification. The DX30 uses RAID-protected disks to increase your confidence (beyond 99%) that the backup has been completed correctly and can be restored when needed. After the backup has completed, the IT administrator can export a copy of the backup data, using the cloning or copy capabilities of LEGATO NetWorker, from the DX30 to the ATL P4000 or ATL P7000 library for archival storage. With the DX30, the customer now has the opportunity to conceptually divide the backup system into two distinct functions—the backup target (high performance and reliability are key characteristics) and the backup archive (requires removable and transportable media). The DX30 addresses the three major concerns that have been expressed by customers: •
The DX30 reduces the time required for backup. A single DX30 system will transfer data at a rate of at least 288GB/hr, which is equivalent to 80MB/second.
•
The DX30 significantly increases the confidence of completing the backup within the allocated backup window. Today, based on the typical number of reported failed backups per year, customers are experiencing 90% to 95% confidence that they can complete their backups on any given night. Typically, as long as the backup software and the applications are appropriately configured, the most common reason why backups fail to complete is because of anomalies with tape drives and tape media. With the DX30 Enhanced Backup System, since it is based on an array of parity-protected disk drives, Quantum can improve the confidence to well over 99% that the backup target will not cause the backup to fail to complete within the backup window.
•
The DX30 reduces the latency of restoring data. With a tape library, it takes minutes before you can begin to restore a file. The fastest possible restore times with a tape library are directly determined by how long it takes to mount a required cartridge in a tape drive and seek to the proper position on the tape to begin restore. It can be several minutes before a restore is started and even then the data transfer is limited by the speed of tape. Exponentially compounding the problem is the case where the required cartridge has been moved off-site for disaster recovery purposes. Furthermore, the time it takes to restore a large amount of data from tape can far exceed service level agreements due to its sequential nature and a potentially high number of cartridge swaps required to complete the restore. With the DX30, the latency to restore is less than 10 seconds in the worst case. The restore latency is on the order of seconds—not milliseconds, like it is for disk arrays that are designed for primary storage. In addition, the enhanced backup system allows the customer to preserve the investments they have already made in backup hardware, software, operational procedures and operator training.
The best way to think of the backup target is to imagine an ATL P1000 tape library that is filled with cartridges and has its front doors and load port permanently locked shut. In this case, the backup software package can back data up to the ATL P1000. When the capacity starts filling up with backup images it can expire some of the backup images and re-use the virtual cartridges for new backups. In this case, the ATL P1000 is used as a closed repository for backup images.
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 14
Quantum DX30 Setup and Configuration The DX30 is a native fibre channel device that can either be directly connected to the backup host, or into the fabric, allowing access from multiple storage nodes. Since the DX30 has a compact frame (4U), it can be put into an existing rack and doesn’t require its own floor space. The solution has been tested with multiple fibre channel switches and host bus adapters (HBA) running on a variety of operating systems. When the DX30 is installed the configuration also includes the number of virtual tape drives and cartridges. Setting up a DX30 with multiple tape drives allows for better optimization by taking advantage of NetWorker’s adjustable parameters such as parallelism and target sessions. The virtual tape drive and cartridge count settings can be changed from the web GUI. NOTE: Changing the number and/or size of the cartridges after the initial installation will cause loss of data. It is recommended to clone the data to another source that can be another DX30 or to removable media. A web interface for management is also available. The web interface can be used to configure email alert and network settings. From the web interface the user can monitor multiple aspects of the device including online status, virtual tape usage, and hardware components. Newer versions of DX30 software can also be loaded through this interface.
Figure 5. DX30 Home Page Interface
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 15
The Quantum ATL P-series Libraries for Archiving and Disaster Recovery Tape libraries are the ideal targets for archival and off-site data. Cloning from the DX30 to the ATL P4000 or ATL P7000 for on-site archival provides the most efficient way to protect medium- and low-priority data. Cloning from the DX30 to a remote ATL P-series library enables an efficient and effective disaster recovery infrastructure. Using the strategy outlined in this paper, high priority data exists in both the Quantum DX30 (available for quick restore) and in local and remote ATL P-series libraries (archived or off-site for full disaster recovery protection). The administrator has the flexibility to choose where and how long to keep the data — based on the data restoration needs. Tape libraries such as the Quantum ATL P4000 or P7000 are the ideal systems for fulfilling the backup archive function for three primary reasons: • The ability to remove and transport media. • Can be directly attached to a server via SCSI or Fibre Channel, or deployed into a Fibre Channel SAN, or work in an IP Storage environment. • Reliability of a platform that exploits common, well-known and time-tested practices, that has proven over 3 million swaps, that is a proven architecture and that has a large install base. The combination of DX30, which handles 3-6 Tb of backup data, and ATL P-series libraries, which handles over 500 Tb of backup data, allows customers to focus their purchase. They can purchase enhanced backup to handle their strategic data, while backing up low priority and some medium priority data directly to the P-series library. This schema produces very cost effective solutions from a cost/benefit standpoint. Investment is applied to reduce risk in the areas of the most potential impact.
Figure 6. DX30 Data Cartridge Status LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 16
CONCLUSION Leveraging the advanced management capability of LEGATO NetWorker and Quantum’s full line of storage devices, this cloning-based, enhanced backup solution delivers solid data protection. This data protection strategy supports the full onsite/offsite data retention lifecycle required to meet industry regulations, disaster recovery protection, and restore service level agreements. LEGATO and Quantum have teamed up to deliver this robust solution with an emphasis on manageability, reliability, and keeping IT administration and resources to a minimum. Using LEGATO NetWorker to backup to the Quantum DX30 disked based virtual tape library, companies can disperse backup targets throughout departments or data centers that defer the investment of time needed for tape handling in this initial, time-constrained step. NetWorker can then clone the backup images to removable media in Quantum’s ATL P-series libraries for local archiving and off-site storage purposes. An IT department can implement LEGATO NetWorker and Quantum’s disk and tape hardware to meet its company’s need for data protection within a reasonable cost/benefit framework.
REFERENCES LEGATO NetWorker Administrator's Guide, Release 7.0, Microsoft Windows Version LEGATO NetWorker Installation Guide, Release 7.0, Microsoft Windows Version LEGATO NetWorker Release Supplement, Release 7.0, Windows Version Quantum DX30 Users Guide, Release 6494002v2r1
LEGATO NetWorker with Quantum Enterprise Products Disk to Disk to Tape Backup
Page 17
