Backup & Archiving Emerging Trends. Presented by Mosaic Technology

Transcription

1 Backup & Archiving Emerging Trends Presented by Mosaic Technology

2 Introduction 3 Tape Backup Not Keeping Pace 3 Disk Alternatives 4 Continuous Data Protection 5 Disk To Disk To Tape (DDT) 6 Back-end Compression 7 Filer-Based Data Protection 7 Disk-Based Data Protection 8 Total Cost Of Ownership: Disk vs. Tape 9 Essential TCO Elements 10 A Market View 11 TCO Example (2TB Data) 12 Summary 14 2

3 : Disk Alternatives Explored Introduction Over the last two decades, computers evolved from simple productivity tools into the core infrastructure of the entire global economy. That rise to prominence is reflected in the increasing focus of IT in all business processes and in evolving oversight regulatory mandates which require stringent controls around data handling and access. Business information and data repositories have become indispensable to the success of any business. Today, workers require information access in order to do their jobs effectively. As a result, the need to protect information and the ability to restore any lost or corrupted data is an essential element of corporate planning and long-term success. Organizations increasingly set goals for business continuity that approach 100% uptime, where data is constantly online and accessible. A key component of near 100% uptime is the protection of data from software malfunction, system failure, or intentional or accidental corruption of data. Managing data protection for the enterprise has become a strategic and challenging issue for companies, accounting for significant percentage of overall spending for IT organizations. Tape Backup Not Keeping Pace Despite the importance of, and the scale of investment in, data protection, companies often lack tools to fully protect and manage their information assets. Many businesses are making a commitment to improve business efficiency and reliability. They are trying to redesign data protection processes, overhaul systems and upgrade applications. Yet these same companies are using backup and restore solutions that have remained largely unchanged for over a decade. Traditional backup and restore solutions account for nearly 30% of the total cost of data ownership, a significant amount when you consider that these solutions: Fail to backup data more than 20% of the time. Fail to restore data 27% of the time from apparently successful backups. Do not protect at least 30% of a corporation s critical data. This data often resides in remote offices or with mobile workers and is costly to bring into a centralized data center. This distributed data, however, is critical information and the amount of this data is growing rapidly. While these numbers may have been historically tolerable, they are no longer good enough to satisfy emerging enterprise requirements. From a regulatory standpoint alone this level of inefficiency would expose many companies to fines and/or legal actions. Businesses need to demand more from their backup and restore solutions in order to improve: Integrity Businesses can no longer tolerate lower integrity for backups than for primary systems. The data for restores must be there when customers need it with regular checks for integrity. Speed It is no longer speeds and feeds that matter but process efficiency. Backups must be quick, non-disruptive and bandwidth efficient. 3

4 Scale Operational efficiency demands a consistent, integrated solution that grows to meet new organizational requirements in data centers, in remote facilities and for distributed clients. Data duplication is a root cause of the data protection challenge. Traditional data protection and archival processes extracts, sends and stores multiple copies of the same data over and over again. For most companies, weekly full backups store the same data ten to fifty times in a year. Daily incremental backups and copies for reliability can double or triple this amount. Creating each of these data duplicates uses valuable server, network, storage and human resources -- resulting in escalating total cost of ownership. After just sixty days of operational backups, a company with 1TB of primary data needs over 10TB of storage to maintain the necessary backups using a traditional backup strategy. Most enterprises use primary disk to store current data and interact with applications; and use tape to back on a periodic schedule. A typical backup schedule calls for daily incremental backups and weekly complete backups. The problems with this scheme are well documented but bear repeating: 1. Tape performance limitations make squeezing into backup windows and hitting recovery time objectives a near impossibility for most IT administrators. 2. Backup reliability and the data integrity characteristics of tape are limited. 3. Costs of managing tape backup infrastructure are skyrocketing as the amount of data on secondary media accumulates. Even when a restore from tape succeeds, the operation is extremely time-consuming. Despite these significant issues and limitations, tape remains the de facto backup medium. Disk Alternatives IT administrators are realizing the limitations associated with traditional data protection solutions. Than and the increasingly stringent requirements of today s business climate have pushed the tape-centric backup architecture to its breaking point. In a November 2004 end-user study of next generation backup technology and trends, Goldman Sachs Global Investment Research learned that many administrators can no longer tolerate the limitations of tape. In fact, two-thirds of those surveyed (66%) indicated that they will start to deploy disk-based backup in the coming year. These tape-based backup limitations combined with disk s decreasing cost and increased performance is changing the backup landscape. A number of data protection solutions that incorporate disk as a key component of their solution architectures are now available. These architectures fall into five categories: Continuous data protection: captures and stores on disk every change to the data volume being protected and requires an alternative backup strategy for medium to long-term data protection due to the volume of transactions being managed. Disk to disk to tape (DDT): sometimes called virtual tape, -- takes advantage of the speed of disk by providing a short-term buffering or staging area for backup data prior to the data being written to tape. 4

5 Back-end compression: focuses on compressing the data to be protected once it is at the backup server, thus reducing the overall tape storage requirements. Filer-based data protection: relies on the filer as a primary storage control point for all data being protected and then leverages snapshot and management techniques on the filer to provide short-term data protection. This approach still relies on some level of tape archive for long-term filer data protection. Disk-based data protection: manages data compression at the source, thereby significantly reducing overall storage requirements and minimizing network bandwidth use for full backups. Due to efficiencies gained by this approach, disk can be used as the medium for short and long-term data protection. Each approach has strengths, weaknesses, and a different architectural fit within the enterprise data protection scheme. Which approach is right for your organization depends on your corporation s specific financial, operational and business goals. Most of these architectures, including continuous data protection, disk-to-disk-to-tape, back-end compression, and filer-based data protection, provide incremental improvements to the backup problem. Yet they still rely on tape as the ultimate destination for protected data. Consequently, IT departments implementing these solutions still must contend with the very real issues and limitations of tape-based technology. The only solution designed from the ground up to take full advantage of the benefits inherent in disk technology to solve the backup problem is disk-based data protection. This solution provides a new way to protect data. It better accommodates the driving financial, operational and business imperatives of integrity, speed and cost-effective scale. Continuous Data Protection The advent of transaction journaling in database systems and the application of journaling concepts to modern file systems provides the genesis for continuous data protection (CDP) solutions. Description In a continuous data protection solution, every write operation to a specified volume is captured and duplicated to a transaction log. This log may be used for audit purposes or may be replayed against a baseline to recreate an image of a volume, as it existed at any point in time. Strengths The primary strength of CDP centers on the continuous nature of transaction logging. CDP does not utilize a backup window. Every write operation is logged in real-time. Transaction logging also lets you recover lost data back to any point in time. However, due to the volume of data stored during the transaction logging process, typically, only several days of data is kept online at any given time. Which means that CDP solutions are used primarily for short-term rollback of selected high-value data. Weaknesses Due to its limitations, CDP is generally used for specific applications and for short windows of protection. While every transaction may be logged under this architecture, it s impractical for operators to select last good write from the millions of stored writes. Applications must sort through stored transactions to determine consistent restore points, which can require significant indexing and post-processing of data. 5

6 This architecture can degrade performance, because it introduces latency in every write operation. Each write operation must be written to the primary data location and to the transaction log. CDP also consumes large amounts of disk storage on active systems, which must exist on the primary data path of the server being protected. Finally, this approach still relies on an alternative backup strategy for medium to longterm storage due to the amount of data generated to enable transaction level restores. Architectural Fit Continuous Data Protection is ideal for applications where users need the ability to restore data back to very specific points in time. CDP solutions are usually deployed in conjunction with another approach for backup and restoration. Disk To Disk To Tape (DDT) Tape-based architectures have been the mainstay of the data protection challenge based primarily on the initial cost difference of tape versus disk media. However, over the past decade, disk buffers were introduced to provide faster backup and restores, without dramatically changing tape backup architecture. Description DDT solutions, or virtual tape libraries, make disk arrays appear as a tape library to existing backup software. The advantage is that DDT does not require installation of new applications or major redesign of existing backup processes. These virtual tape systems store a tape image of a full or incremental backup to disk before the data is ultimately written to tape. DDT Strengths The primary strength of DDT centers on maintaining the status quo. The software environment requires no changes and minimal impact to operational activities is minimal. The increased performance of disk-based tape emulation can shorten backup windows, and the greater reliability of disk storage significantly increases the probability of successful backups. DDT also serves as an impedance match for network backups. Often networks cannot supply data at the fast, constant rate tape drives require. When this happens, tape drives go into stop/start mode, which negatively impacts effective throughput of the tape device. The performance of disk-based tape emulation tracks linearly with network throughput. DDT Weaknesses Over the past decade, backup and archiving systems have become highly attuned to tape systems constraints. These constraints define the limitations we experience with backup today, including requirements for repeated large data transfers, difficulty of maintaining multiple indexes to track data location, and relatively inefficient use of tape capacity. Today, an effective backup strategy typically requires a tape library five to ten times as large as the amount of data being protected. In order to hold repeated full copies of the data -- in addition to incremental copies an enterprise with 1 TB of data can often require a 10 TB tape solution to provide adequate data protection. DDT solutions still 6

7 rely on tape systems for medium to long-term data storage. As such they do not eliminate any tape driven constraints DDT solutions also introduce additional complexity and cost in the overall backup architecture with the deployment of another set of hardware and increased overhead of managing the DDT solution. DDT may also require additional backup software licenses since typical backup software packages are licensed on a per drive and/or tape library basis. Architectural Fit DDT solutions are an extension of the current disk to tape solution implemented over the last two decades. DDT solutions provide incremental improvement to a few of the problems that organizations face with tape-based backup solutions, including backups that take too long to complete, the low reliability of data restores, and excessive time and effort spent recovering data from tapes. Back-end Compression Back-end compression solutions are an evolutionary improvement of the disk-to-diskto-tape solutions aimed at reducing the amount of disk needed for staging and the amount of tape needed for long-term archives. Description Back-end compression solutions seek to reduce the total data storage requirements by employing various data compression techniques. While reducing hardware costs, these solutions inherit both the advantages and disadvantages of disk-to-disk-to-tape solutions. Back-end compression solutions can be implemented as tape emulation devices or as NFS mounted file systems. Back-End Compression Strengths In addition to the strengths of disk-to-disk-to-tape solutions, this approach reduces the hardware cost through more efficient storage of the backup data and provides for longer retention of data on disk compared to disk-to-disk-to-tape solutions of the same capacity. Back-End Compression Weaknesses In addition to the weaknesses of disk-to-disk-to-tape solutions, this approach is only effective within the device the compression software is loaded on. Also, given the amount of CPU resources required to perform compression and decompression of data, there is a potential performance impact when introducing back-end compression into data protection architectures. Architectural Fit Back-end compression solutions are typically deployed in place of or in addition to diskto-disk-to-tape solutions to reduce the overall hardware that must be purchased to support the DDT deployment. Filer-Based Data Protection Dedicated NAS filer appliances are a very stable and easy to administer platform for many storage applications. Given the appliance nature of these devices and the close integration and control of the underlying file system, there has been a natural evolution 7

8 to supply logical volume management services such as block level snapshots and replication. Description Modern filer appliances create block-level snapshots, point-in-time captures of the entire file server that can be used when restores are necessary. These snapshots are generally limited in number and stored on the primary storage device. In addition, some filer appliances enable block-level replication across a dedicated network or even WAN in order to copy these snapshots to another device or location. Filer-Based Data Protection Strengths The primary strength of this architecture -- similar to disk-based solutions is that block level change detection is more efficient than incremental backups. Snapshot administration is integrated with filer administration, and delivers highly reliable shortterm data protection and online restore for data contained in snapshots. Weaknesses In some cases, snapshots can overrun their storage allotment and reduce the overall storage available on the file server. In addition, snapshots work only for the file server in question; they do not work for heterogeneous systems with their own internal or external storage, which require their own data protection solution. Furthermore, most organizations still archive snapshots to tapes due to the cost of snapshot storage and the limitations on the number of snapshots supported by a given file server. Architectural Fit Filer-based data protection solutions are usually deployed in conjunction with another approach for backup and restoration. They provide many of the same benefits as DDT solutions, including reduction in backup windows and rapid restores for recent data. Many organizations use filer-based data protection to protect filer data in the short-term but still use traditional tape-based backup solutions to protect data for the long-term. Disk-Based Data Protection Historically, tape backup has been the default choice for protecting enterprise data. Most backup and recovery software packages are built around tape media and automated tape libraries. With its dramatic price declines, disk storage has become increasingly attractive as an addition or replacement for tape media in modern data protection solutions. Disk storage has a number of important advantages over tape media, including random access to stored data, higher reliability and faster access times. As a result, data protection solutions are quickly evolving to adopt hard disk storage as an integral part of backup and recovery for the enterprise. Description Unlike other architectures, disk-based data protection takes full advantage of disk storage as the final repository for backup and archival data. By re-designing the backup architecture around the inherent advantages of disk storage, disk-based data protection solutions can eliminate the need for tape storage altogether. Using content-addressable storage and other high-efficiency techniques, these solutions store full copies of backup data in a fraction of the amount of space required for traditional full and incremental tape backups. 8

9 For example, Avamar Axion s intelligent agents identify and filter redundant data stored in files within a single system, across systems, and over time at each client. Each unique sub-file data block is only backed up once within the enterprise. This reduces by over 99% overall data required for a full daily backup transmitted over the network and stored on the data protection server. Disk-based data protection solutions also offer fault tolerance capabilities through the use of conventional disk RAID or other models providing far higher availability than other solutions. Using data replication tools enhances disaster recovery efforts by mirroring data across geographically distributed sites. With its efficiency advantages, disk-based data protection provides online access to all backup and archival data and replication, within the cost-envelope of traditional tape solutions. Disk-Based Data Protection Strengths The entire purpose of backup and archive is to enable timely restoration of data when needed. Disk-based data protection provides online access to backup data that can be immediately restored to a desired location. Disk-based backup solutions also eliminate the challenges associated with long backup windows for traditional tape backups. By only sending the changed blocks of data on a daily basis, disk-based data protection solutions move quickly through the backup process and greatly reduce network bandwidth use for daily backups. Another advantage of disk-based data protection solutions is the far higher level of data integrity when compared to traditional backup solutions. Because all backup data is available on disk, these solutions perform regular consistency checks and validation, ensuring that data can be restored successfully when it is needed. By eliminating tapes, these solutions significantly reduce the manual effort and cost of managing data protection. Disk-Based Data Protection Weaknesses Disk storage is familiar. It s been used reliably in enterprises for decades as primary storage. In order to adopt a disk-based data protection solution, enterprises must be willing to make fundamental changes in their data protection architecture. Because these solutions replace existing backup solutions, companies and management may be reluctant to change current data protection architecture. And, because disk-based data protection solutions are relatively new, they may not provide the range of platform support available compared with legacy solutions. Architectural Fit Disk based systems were not deigned to fit into the existing tape-based backup paradigm. Disk-based data protection solutions were designed from the ground up to take full advantage of hard disk storage as the final repository for backup and archival data. As a result, these solutions have a tremendous advantage in efficiency over existing backup solutions. Disk-based data protection solutions provide backup, recovery, archive and replication to an offsite location for disaster recovery -- all for the same cost as a traditional tape backup solution with offsite tape vaulting. Total Cost Of Ownership: Disk vs. Tape A total cost of ownership (TCO) analysis measures the cost of acquiring and operating a new technology solution against a current installation. In the late 1990s, Gartner Group estimated the 9

10 cost of operating a unit of primary storage could be five or more times the initial cost of acquisition, which highlights the importance of an accurate TCO analysis for enterprise storage buying decisions. The cost of operating secondary storage for data protection also dwarfs the initial acquisition price by three to four times. This section identifies components required for accurate TCO analysis. Then we ll show an example that compares total cost of ownership for a tape backup solution and Avamar Axion, an efficient disk-based data protection solution. Essential TCO Elements When considering a disk-based data protection solution, it is important to include the following cost factors in a TCO analysis: Acquisition of new hardware and software Annual hardware and software maintenance Operational media Staff administration Network communications (especially wide area network usage) Where offsite data storage is a consideration, the analysis should include the following cost elements, when applicable: Offsite vaulting Disaster recovery service (second site hardware, software and staff) Additional media requirements Disaster recovery tests or reviews Network communications When you compare a disk-based data protection solution to a tape-based solution, the value of improved restore times and increased reliability can be substantial. Enterprises continue to pay hidden costs for lost productivity due to long restore times or worse, the cost of reconstructing data for restores abandoned or never initiated due to the limitations of tape backup. Annual data protection costs range from $30-$300/GB, while data reconstruction costs are typically estimated to be much higher, in the range of $900/MB. In some environments, the demand for regular restores necessitates expensive layers of redundancy that quickly inflate the total cost of data protection for an enterprise. A separate means to evaluate the value of fast restore apart from other cost considerations is outlined later in this section. According to a recent study conducted by The Taneja Group, many enterprise customers do not calculate actual data protection costs or actual success rates in performing backup and restores. Traditional data backup systems cost up to $100/GB per year, according to Avamar studies. Using a more conservative cost estimate of $60-$80/GB per year, this section s analysis compares an average market view cost of ownership for tape solutions versus Avamar Axion, a disk-based data protection solution, for three business cases: a small, medium and large enterprise. 10

11 Disk (Avamar Axion here) provides customers with the value of backup, remote replication and sufficient communications necessary to accomplish remote replication for the cost of local tape backup alone A Market View Tape backup solutions create multiple redundant copies of primary data due to inherent limitations of tape as a serial medium and the need to increase the reliability of data restores. Most enterprises back up all data in any particular server once per week, although some enterprises perform this process more frequently. This backup -- called a full backup -- contains a complete copy of all the data in the system at that point in time. On all other days of the week, incremental backups are preformed. These backups contain copies of only those files that have changed. To restore data, the last full backup is restored and then any intervening incremental backups are applied. To increase reliability and reduce restore times, most enterprises perform cumulative incremental backups. Cumulative incremental backups collect and store copies of all the files that have changed since the last full backup was taken. Some solutions also offer an incremental forever approach, where the data protection system can rebuild a virtual full from an initial full and the incremental backups. On average, enterprise systems have a daily file change rate of 3% per day. Therefore, cumulative incremental backups range from 3% on day one to almost 18% on day six, with a daily average over 10% (for our discussion we ll use an average of 10% in the TCO calculations). Over the course of a week, a set of one full and six incremental backups results in 1.6 times the amount of primary data storage per week. Enterprises also must decide how many weeks of full and incremental backups to maintain in automated libraries for near-line restore. Tapes rotated out of a library must be recovered from onsite or offsite storage, prolonging restore times. Companies, on average, leave between one and two months of data in the library, sometimes combined with a set of clones (copies) of the latest full backup in case the first set of tapes is corrupted. For medium and large enterprises, this discussion assumes six to seven weeks of nearline retention or about 10 times primary data storage. For small enterprises, who often maintain less data in near-line storage for cost reasons, we ll assumes two weeks of data in near-line storage or 3.2 times primary data storage. By maximizing its use of disk storage technology, Avamar Axion provides customers with efficient full backups that are smaller than traditional tape incremental backups. This eliminates the need for incremental backups. Avamar tested its Axion solution in over a hundred environments with remarkable consistency. Avamar found that initial levels of redundancy across files and systems are between 40% and 70%. For example, with 100TB of primary storage, the initial backup can be stored in 30-60TBs. This initial redundancy or commonality depends on how homogeneous the original systems are. For example, a set of 10,000 Windows laptops can sometimes exceed 70% initial redundancy. This white paper assumes an average initial redundancy of 55%. Commonality for daily backups is significantly higher. Some commonality ranges for systems include: Windows systems 99.6% and 99.9%; 11

12 Solaris and Linux -- between 99% and 99.5%; Highly active Oracle database servers, as low as 98%. This white paper assumes daily commonality of 99.2% (each day s full backup is.8% of the primary data being protected). Additional assumptions are as follows: 1. All hardware and software is calculated at list prices. Hardware configurations are detailed separately; tape backup software pricing is based on configurations from marketleading, enterprise backup software vendors. 2. Tape library configurations are ½ full of drives and ½ full of media at acquisition. 3. Hardware support is 15% of list price per year. 4. Software support is 20% of list price per year. 5. Tape media cost is based on a 100GB tape cartridge at $ Administrative staff efficiency is based on a 2002 Giga Group analysis that evaluated benefits of disk-based vs. tape backup systems in a number of customer environments. 7. The average, fully burdened cost of a storage administrator is $75,000 per year. 8. All hardware and software is depreciated/amortized over three years. 9. All costs are normalized $/GB per year, averaged over three years. 10. Pricing for wide area network (WAN) costs is included when backup data or replication data is transferred over a WAN (see figure 1). WAN costs used in this TCO analysis represent an average of cost quotations we have solicited for bandwidth ranging from fractional-t1 to OC3. Note: although individual enterprises will find some variability in the cost components of the following scenario, several real-world analyses yielded consistent overall results. Note: The example below compares a tape system with an Avamar Axion Disk solution. However, the same rational and methodology can be applied to other disk solutions as well. TCO Example (2TB Data) Gartner, in a 2003 study, identified mid-tier enterprises as companies with $50-$500 million revenues and employees. They found that nearly 90% of the companies had less than 1.5 TB of data under management. This data level scenario can also be applied to independent departments within larger enterprises. Based on this model we can then build a TCO. Tape hardware costs (see Figure 2) include one entry-level tape library, with 5TB of slot capacity and two LTO Generation 2 tape drives. The backup software costs include support for ten client systems. An entry-level server is included as the backup server. The disk solution is a standard Avamar Axion E configuration, an easy-to install data protection appliance with 2TB of usable storage and software licenses for ten client systems. A summary of comparative costs for this scenario is shown in Table 1. 12

13 Table 1: Cost Comparison 2TB Backup Only A disk system in this case Axion -- can provide online restore capability at less than half the total cost of a similarly sized tape backup system. In addition, while the tape library can hold only two weeks of active data, the disk system can hold a minimum of seven weeks (or significantly longer with the advanced retention policies available with Axion). All data stored in the disk system over this period is online and available for immediate restore. Even if the enterprise already has a fully depreciated, installed tape system, savings in maintenance, media and staffing alone support the cost of acquiring and operating a new disk systems such as Axion Backup and Offsite Data Storage Large enterprises routinely move data offsite for disaster recovery, but Gartner Group discovered that 49% of mid-tier enterprises do not move their data offsite at all due to cost constraints. In this scenario, two items are added to the tape costs: additional tape media and transportation/warehousing of offsite media. With its remarkable efficiency, disk provides small enterprises an affordable offsite solution over a wide area network (WAN). In the Axion case, data can be efficiently replicated on a scheduled basis to a second Axion E system at a second site, adding the following costs: a second Axion E, Axion Replicator software and the cost of WAN communications. A comparative summary of costs is shown in Table 2. Table 2: Cost Comparison 2 TB Backup and Replication Even with the additional cost of a second disk system, Replicator software and a T1 WAN line, the total cost of backup and nightly electronic replication with disk is the same as tape backup with offsite data movement by truck. With Axion, a small enterprise that pays for offsite tape 13

14 vaulting can actually afford full electronic replication, with the added benefit of immediate, online restores. Disk efficiency enables a small enterprise to have both online backup and restore AND offsite replication for the same cost as traditional tape backup and offsite archiving by truck. For even greater cost savings, a single Axion E system can be installed at an offsite location, separated from the primary servers, in order to provide offsite data protection without the cost of a second Axion E. Backups then occur across the WAN, adding only $4/GB per year for a total cost of $18/GB per year. An enterprise can begin with this approach and eventually migrate to a dual Axion E implementation as budgets allow. Summary For several decades, IT administrators have coped with the operational challenges of managing tape-based backups. Today, cost-effective data protection solutions incorporate disk as a key component of their architecture, providing significant business and operational benefits. These solutions can be categorized into five categories: continuous data protection, disk to disk to tape (DDT), back-end compression, filer-based data protection and disk-based data protection. Each of these solutions has its strengths and weaknesses. The first four solution architectures are additive in nature to the legacy backup solution, still relying on tape infrastructure to archive data over the medium to long-term. These solutions add complexity and cost for the benefits they provide. The only solution that replaces the existing tape infrastructure, significantly reduces the total cost of ownership of data protection and resolves the shortcomings of tape-based backup solutions is disk-based data protection. A Total Cost of Ownership analysis shows that in addition to performance advantages, disk clearly offers short and long terms saving over tape. 14