Transcript
Tape Technology Report For Tape It’s a New Game with New Rules
BY: Fred Moore President www.horison.com
Abstract To fully address the enormous digital data challenge, the tape industry has been rearchitecting itself and the renaissance is well underway. The tape industry continues to deliver new technologies and compelling functions including unprecedented cartridge capacity increases, vastly improved bit error rates yielding the highest reliability of any storage device, a media life of 30 years or more, and faster data transfer rates than any previous tape or disk technology. Many of these innovations have resulted from technologies borrowed from the HDD (Hard Disk Drive) industry and have been used in the development of both LTO (Linear Tape Open) and enterprise tape products. Additional functionality including LTFS, RAIT, and the Active Archive adds further value to the tape lineup. Disk technology has been advancing, but the progress for tape has been greater over the past 10 years. Disk and flash will need a huge assist from tape to cost-effectively meet the coming waves of storage demand as hyper-scale and exascale computing looms just around the corner. Fortunately, today’s tape technology is nothing like the tape of the past. It’s time to bring your understanding of tape up to date as the future for tape technology has never been brighter. The era of modern tape has arrived!
Horison Inc. www.horison.com
Page 1
Did You Know These Facts About Tape? Have you wondered what has been happening in the tape storage industry, if anything? • tape is cheaper ($/GB) to acquire than disk, • tape is less costly to own and operate (lower TCO) than disk, • tape is more reliable than disk, • the capacity of a tape cartridge is higher than the capacity of the largest disk drive, • the media life for modern tape is 30 years or more for all new media, • with LTFS tape now has media partitions for faster “disk-like” access, and • the 10-year roadmap for tape is well defined and solid. If you didn’t know - you are not alone – but it’s time to bring your understanding of tape up to date. The Era of Modern Tape Arrives - Major Tape Enhancements Since 2000 By 2000, a new era of tape technology was underway as the tape industry was busy re-engineering itself. Key tape developments yielded higher capacities, much longer media life, vastly improved drive reliability, lower acquisition and TCO (Total Cost of Ownership) and much faster data rates than any previous tape or disk technology. Troublesome tape issues of the past including edge damage, stretch, tear, loading problems, and media alignment with older tape formats such as DAT, DDS, DLT, Travan, and 8MM tape were successfully addressed. Coupled with the advent of Barium Ferrite (BaFe) media, each of these developments has helped to redefine the future of tape. With the internet, cloud, big data and IoT waves promising unprecedented data growth, the timing for advanced tape couldn't be better. Reliability Ratings Soar for Tape For years MTBF (Mean Time Between Failure) was used to measure storage device reliability but this has given way to bit error rate (BER) as the de-facto standard measure of reliability. Several factors have contributed to improve tape reliability. PRML (Partial Response Maximum Likelihood) is the most effective error detection scheme and is widely used in modern disk drives. Borrowing from the disk industry, LTO drives switched to PRML from RLL (Run Length Limited) encoding as PRML attempts to correctly interpret even the smallest changes in the analog signal. PRML can correctly decode a weaker signal enabling a much higher recording density while allowing tape to surpass disk in reliability. Today, both LTO and enterprise tape products are more reliable than any disk drive. Times have changed! Storage Technology Reliability Ratings
BER (Bit Error Rate)
Enterprise Tape (T10000x, TS11xx, LTO-7)
1 x 10E19 bits
LTO-5, 6, and 7, Flash SSD
1 x 10E17 bits
Enterprise HDD (FC/SAS)
1 x 10E16 bits
Enterprise HDD (SATA)
1 x 10E15 bits
Desktop HDD (SATA)
1 x 10E14 bits Source: Vendor’s published product specifications.
Key point: Tape has the highest reliability of any reliable storage device. Horison Inc. www.horison.com
Page 2
Comparisons between disk and tape reliability have long been the subject of lively debates and much of the debate stems from past user experiences with aging tape technologies, however the experience is completely different with modern tape. The reliability for today’s enterprise and open systems tape drives and media range from one to three orders of magnitude higher than the most reliable Fibre Channel disk drive and this trend is expected to continue to favor tape in the future. Customers have indicated for years that their most frequently perceived cause of tape failure was due to media and handling errors and this perception has lingered. Modern tape media has made significant strides to eliminate these now out of date concerns. Special prewritten data tracks on the tape called servo tracks (a track that allows the tape drive head to stay aligned with tracks on the tape) keep the tape heads on the correct track to meet the requirement for completely accurate recording and reading of stored data. With the older linear tape products, the edges of the tape media served as servo tracks. Since 2000 enterprise and LTO drives have eliminated this issue by combining the pre-recorded servo tracks on the media (away from the edge) along with developing more ruggedized cartridge shells that are relatively impervious to handling damage. Tape Media Considerations For years, Metal Particulate (MP) pigment was the primary tape media type. MP is mainly made of iron (Fe) therefore it will oxidize over time and its magnetic property will deteriorate. To slow down the oxidation process, the outer layer of MP is intentionally oxidized from the beginning. All generations of LTO cartridges prior to LTO-6 have exclusively used the MP pigment. LTO-6 uses MP and BaFe. Along with magnetic material advances, tape media have increased substrate dimensional stability with reduced thickness, much smoother surfaces, lower defect densities, and increased edge slitting precision. Tape media must be highly reliable, portable and rugged enough to be transported without impacting reliability, and must have a high capacity and very long-life for archival applications. Barium Ferrite Arrives – A Game Changer for Tape Media Barium Ferrite (BaFe) is made of an oxide and therefore it does not lose its magnetic property due to oxidation. The data capacity of magnetic tape is determined by the number of magnetic particles in its magnetic layer. The much smaller BaFe particles are one of the main advantages for using BaFe as it allows more particles-per-unit volume and therefore improves the Signal to Noise Ratio (SNR) and reliability. Therefore, magnetic tape using BaFe media has a longer archival life of more than 30 years based on accelerated life tests. In January 2010, scientists at IBM Research in Zurich demonstrated recording data onto a new, advanced prototype BaFe tape media developed by Fujifilm Corporation at a density of 29.5 Gb/in2 potentially yielding a native (uncompressed) capacity of 35 TB on a single cartridge. Because of this joint R&D effort, several new tape technologies were also developed including improved precision control of read-write head positioning, more than a 25-fold increase in the number of tracks, new detection methods to improve read accuracy, and a new low friction read-write head. In 2011, Fujifilm delivered Horison Inc. www.horison.com
Page 3
the world’s first commercial BaFe data storage cartridge for the IBM 3592JC and STK T10000C enterprise tape drives and for LTO-6 drives in 2013. On April 9, 2015 Fujifilm Recording Media USA in conjunction with IBM announced a new record in areal density of 123 Gb/in2 on linear magnetic particulate tape had been achieved. This marks the fourth time in less than 10 years that Fujifilm and IBM have combined to achieve record-breaking storage capacities on tape. This density breakthrough equates to a standard LTO cartridge capable of storing up to 220 TB of uncompressed data, more than 36 times the capacity of the current LTO -7 tape. A tape of this size can preserve the human genome (600 GB) of 360 people on a single cartridge and represents the highest capacity storage media ever demonstrated, including all HDDs, Blu-ray disc and solid state NAND flash devices. Drive Type and Media
Capacity (native)
Data Transfer Rate (native)
Channels/ head
Tracks
Areal Density
LTO-7
BaFe
6.0 TB
300 MB/sec
32
3,584
4.3 Gb/in2
LTO-6
MP, BaFe
2.5 TB
160 MB/sec
16
2,176
2.2 Gb/in2
TS1140
BaFe
4.0 TB
250 MB/sec
32
2,560
3.2 Gb/in2
T10000D BaFe
8.5 TB
252 MB/sec
32
4,608
4.93 Gb/in2
TS1150
BaFe
10 TB
360 MB/sec
32
5,120
6.52 Gb/in2
TS1155
BaFe
15 TB
360 MB/sec
32
7,680
9.78 Gb/in2
Key point: The tape industry has pushed capacity, reliability and media life to record levels. BaFe media demonstrations indicate continued advancements in tape technology for many years ahead. Future Data Recording Projections Areal density refers to how many bits of information can be stored on a given surface area of a magnetic disk drive or tape media. Tape areal densities have reached 9.78 Gb/in2 while HDD areal densities are in 1,300 Gb/in2 range, 130 times denser. Future density scenarios (see INSIC 2016 chart below) project that annual HDD areal density growth rates will not maintain their past 35-40% values and will likely slow toward 16%. Tape areal density growth rates are expected to double the HDD rate averaging about 33% annually. The surface area available to increase HDD capacity on disk platters is getting crowded while tape cartridges have a surface area roughly 200 times greater than HDD to work with. To increase capacity, many HDDs have increased the number of platters from three up to seven while using helium filled disks to reduce friction. Modern tape drives operate at areal densities that are more than two orders of magnitude smaller than the latest HDDs. It should therefore be possible to continue scaling tape technology at historical rates for at least the next decade before tape begins to face challenges related to the super-paramagnetic effect which today’s HDDs are facing. The smaller the magnetic particle, the more data there is in a single bit cell. The net result of these areal density scenarios is a sustained volumetric and total capacity storage advantage for tape technology. Horison Inc. www.horison.com
Page 4
NAND flash memory has surpassed hard disk drive (HDD) technology in areal density for the first time with announcements of 2,770 Gb/in2. The transition to 3D-flash with up to 64 layers is imminent and SSD continues to apply pressure the HDD industry.
Areal Density of Hard Disk and Tape - Laboratory Demonstrations and Products
Key point: Honestly, did you realize magnetic tape has such a recording advantage over HDD? Tape Roadmaps Show Continued Progress The tape industry has delivered numerous enhancements in the past 15 years with significant improvements in drive reliability, media life, and data rate and cartridge capacity. The LTO Consortium publishes a well-defined roadmap (see below) with each successive LTO generation arriving in approximately two-year intervals insuring LTO will continue to be the primary tape technology while steadily improving the acquisition price, capacity and performance, and will reduce cost of ownership over previous models. LTO’s stated direction is that “an LTO Ultrium drive is expected to read data from a cartridge in its own generation and at least the two prior generations. An LTO drive is expected to write data to a cartridge in its own generation and to a cartridge from the immediate prior generation in the prior generation format.” This eases customer conversion efforts by extending the life of the older media while newer LTO tape drives replace prior versions. In addition, the most recent LTO-7 format has expanded the “history buffer” in the compression engine, giving it a 2.5:1 compression ratio, up from 2:1 on previous drives, yielding an average of 15 TB of compressed capacity per cartridge. With the media life for all new LTO and enterprise class tape being 30 years or more, tape offers a highly secure, portable, long-term storage and archive medium. Horison Inc. www.horison.com
Page 5
Key Point: The progress of future LTO tape systems is well defined, highly attainable, expected to support several more years of technology advancements. Expect similar improvements and progress for enterprise tape. Security and Media Life Security features are included on LTO and enterprise tape drives to address countless compliance and legal requirements including Write-Once-Read-Many (WORM), data encryption to protect data at rest, and various write-protect capabilities. Since tape is removable media, physical cartridges can be easily transported to another location in the case of extended power outages or natural disasters which have become more common in recent years. Given the number of natural disasters potentially forcing data centers to go without electricity, media portability remains the last, but valuable line of defense for data protection. As a result, the traditional “truck access method” hasn’t lost its value. Cybercrime occurrences are all on HDDs, but are negated on tape with a line of defense called the “tape air gap” since tape data is stored offline and is not accessible to hackers. Disk Challenges are Mounting Shifting the Storage Landscape Technology roadmaps and the approaching technology limits indicate HDD storage is entering a squeeze play in the data center. HDDs are increasing in capacity -but not in performance - as the IOPS (I/Os per Second) for HDDs have basically leveled off. The potential for more concurrently active data sets increases as HDD capacity grows and the increased contention for the single actuator arm causes more response time delays. HDDs also have a much higher TCO and use considerably more energy than tape or SSD. Excessive RAID rebuild times are becoming a concern and it can now take several days to rebuild a failed HDD. This means that the disk subsystem will run in degraded mode during the lengthy rebuild period impacting performance. As HDD capacities continue to increase, total time required for the RAID rebuilding process will become prohibitive for many IT organizations and higher capacity HDDs could force a replacement for traditional RAID architecture implementations. A notable shift in the storage Horison Inc. www.horison.com
Page 6
landscape is underway as high-performance data moves from HDD onto flash SSD while lower activity and archive data migrate from HDD to modern tape. For the foreseeable future, HDDs will remain the home for primary storage, mission-critical data, OLTP, the highest availability applications and databases, but HDD shipment growth rates have declined nearly 35% since its highpoint in 2013. The “storage squeeze play” has HDD’s caught in the middle as flash and tape deliver increasing value. Key Factors That are Changing the Rules for Tape Source: Horison, Inc. Reliability T10000x, TS11xx, LTO-7 1 x 10E19 bits Tape drive BER LTO-5, LTO-6 1 x 10E17 bits____________________________ Note: BER (Bit Error Rate) Flash SSD 1 x 10E17 bits___SSD BER__________________ Hard Read Errors per Total Enterprise HDD FC/SAS 1 x 10E16 bits Disk drive BER Bits Read Enterprise HDD SATA 1 x 10E15 bits Desktop HDD SATA 1 x 10E14 bits Capacity The popular LTO tape drive family is now in its 7th generation with LTO-7 and future generations through LTO-10 are defined in the LTO roadmap. The IBM TS1155 cartridge currently has the industry’s largest media capacity at 15 TB native and 45 TB compressed @ 3-1. Data Transfer Rate 360 MB/sec for TS1155 tape, 300 MB/sec for LTO-7, ~175 MB/sec for disk. Device Performance Trends Much faster data rates and improved access time with LTFS lie ahead for tape. Few if any performance gains left for future HDDs. Capacity Growth Roadmaps heavily favor tape over disk as 220 TB capabilities have been demonstrated by Fujifilm and IBM using BaFe materials. Tape presently has roughly 200 times more recording surface area than HDDs. Scaling Capacity Tape scales capacity by adding media, disk drives scale capacity by adding more drives resulting in more energy consumption. Security LTO and enterprise tape drives offer both WORM and encryption. Long Life Media The shelf life for all new LTO and enterprise BaFe tape media is rated at (Shelf life) 30 years or more making it ideal for long-term archival storage. Disk drives typically have a 4-5-year lifespan before replacement or failure. Improve Tape File Search LTFS arrived with the LTO-5 tape cartridge implementing two distinct, and Access Time individually addressable, unequal partitions with the first quick read Performance (LTFS) partition containing descriptive metadata that enables the quick search capability (random-like) of the data contained in the second partition. Energy Efficiency Heavily This is becoming a goal for most data centers for archival data - “If data Favors Tape isn’t being used, it shouldn’t consume energy”. The 5-year tape energy (green initiatives) cost is approximately 5% of HDD for equivalent capacity. Acquisition Price Tape has a significantly lower purchase price ($/GB) than disk. TCO 5-year TCO for disk is typically 6-15x higher than tape. Management Capability Typical tape administrator can manage PBs (1x1015) of automated tape. Typical disk administrator can manage TBs (1x1012) of data. Key point: HDDs are caught in the middle as storage administrators strive to optimize their storage infrastructure to address high performance applications with SSD and archival demands with tape.
Horison Inc. www.horison.com
Page 7
Total Cost of Ownership Favors Tape over Disk Tape’s significant cost per gigabyte and Total Cost of Ownership (TCO) advantage compared with other storage mediums clearly makes it the most cost-effective technology for long-term data retention. Keep in mind that tape capacity can scale without adding more drives – this not the case with HDDs where each capacity increase requires another drive, with more energy and cooling. Tape holds a compelling value proposition over disk in environmental requirements and data longevity. Some excellent TCO studies are publicly available and show the TCO for HDDs is approximately 6 to 15 times higher than for the equivalent capacity tape systems. See the Clipper Group TCO Analysis. Emerging Applications Accelerate Future Tape Growth Tape has been expanding its historical role as a backup solution to address a much broader set of requirements including data archives and disaster recovery services. Digital archives consisting of unstructured data, digital images, multi-media, video, social networks, compliance and surveillance data are the fastest growing data category experiencing a CAGR (Compounded Annual Growth Rate) in the range of 60% annually. Just ten years ago, large businesses generated roughly 90% of the world’s digital data. Today an estimated 75-80% of all digital data is generated by individuals rather than by large businesses, however most of this data will eventually wind up back in a large business or cloud provider’s data center requiring unprecedented archival requirements and much of this will be destined for tape. The Big Data Era is here and the value of the digital archive is increasing as the benefits of analyzing and working with very large datasets enable analysts to project new business trends, prevent diseases, and improve security and national defense, and much more. Note that much of this data will never be stored, or will be retained for only a brief time while data stored for longer retention is frequently compressed resulting in lower net storage requirements. Presenting an ever-moving target, the boundaries of digital archives now reach petascale (1x1015) or exascale (1x1018) and will approach zettascale (1x1021) capacity levels in the foreseeable future. Meeting these storage requirements only with disk and/or flash will become financially prohibitive for most businesses. Key point: The TCO of tape over disk is compelling for archival storage, and with reliability having surpassed HDDs, the pendulum has shifted to tape to address much of the exploding demand for longterm storage. LTFS Enables Faster Data Access and Interchange Improvements To improve the access and interchange capabilities of tape, a new, long awaited file system specification for LTO called LTFS (Linear Tape File System) became available with LTO-5 in 2010. Originally developed by IBM, LTFS provides an easier way to access and archive data to tape without the need for another backup software product. With the new dual partitioning functionality of LTFS, one partition holds the index and the other contains the content, allowing the tape to be self-describing. The metadata of each cartridge, once mounted, is cached in server memory. Metadata operations, such as browse directory tree structures and file-name search, are performed more quickly in server memory and do not require Horison Inc. www.horison.com
Page 8
physical tape movement. The faster file access capability provided by LTFS becomes more important as tape capacities continue to increase and the number of files per tape steadily increases. LTFS makes archiving and retrieving data easier than ever before for tape drive and library applications. Key point: Expect LTFS partitioning and its future iterations to provide even greater access capabilities for tape and attract ISVs (Independent Software Vendors) to exploit its capabilities. RAIT is Arriving Providing Much Higher Transfer Rates RAIT (Redundant Arrays of Inexpensive Tape) is available with HPSS (High Performance Storage System) and aggregates bandwidth across multiple tape drives in parallel significantly increasing throughput. RAIT uses multiple tapes loaded in parallel for writing and reading data and provides parity for data reconstruction like RAID does for HDDs. Interest in RAIT is increasing as it aims to take advantage of significantly higher future tape transfer rate increases which are projected to yield tape data rates 5x faster than HDDs by 2025. Thus, the much higher transfer rates of HPSS based RAIT will position RAIT for the HPC, hyperscale, cloud and enterprise markets. See INSIC data rate projection chart for tape and HDDs (above). Energy Consumption - Tape Means Green Storage A commonly stated objective for many CIOs today is that “if data isn’t used, it shouldn’t consume energy”. In response to this directive, the movement of archival data from HDD to more reliable, much more energy efficient, and more cost-effective tape storage is actively underway. Unlike storage providers, energy providers have shown little interest in lowering their rates (price per unit) and average data-center energy costs are growing at 10-20% per year or more per unit consumed. The world’s data centers now consume almost as much energy as the country of Spain and data centers consume just over 2% of the total US electrical output. Best practices for using less energy in the data center focus on the two highest areas of energy consumption – servers and disk storage. Tape cartridges spend most of their life in a library slot or on a shelf and consume no energy when not mounted in a tape drive. As capacity demands increase, tape capacity can be added (scaled) without adding more drives and more energy consumption. This not the case with HDDs where each capacity increase requires another drive. Energy costs for tape capacity are typically less than 5% of the equivalent amount of disk capacity. The limits of power distribution in many data centers is being approached, forcing organizations to explore new cooling techniques such as water-cooled racks, outdoor and mobile Horison Inc. www.horison.com
Page 9
cooling, or in some cases, building another data center. Building another data center is normally a last resort and is extremely expensive mandating that energy consumption be properly managed. Average IT electrical consumption rates for data centers are summarized in the chart below. Average Electrical Power Usage for Data Centers Chillers, cooling, pumps, air-conditioning Uninterruptible power supply Air movement, circulation, fans etc. Misc. lighting, security, perimeter surveillance Total support infrastructure – external consumption Servers Disk drives, control units Tape drives, robotic tape libraries Network gear, SAN switches and other devices… Total IT gear – internal consumption
24% 8% 10% 3% 45% 30% 12% 3% 10% 55%
Source: Horison, Inc. and estimates/averages from various industry sources.
Heat is the enemy of all IT technology as it impacts reliability and makes adhering to recommended environmental requirements a critical factor. Various utility companies are restricting the amount of power some businesses can use at certain times of the day making data center energy management more critical. Clearly, exceeding the power limits a utility company can deliver to a given facility should be avoided as moving to a new facility can be cost prohibitive. Hyperscale data centers such as Amazon, Apple, Facebook, Google and Microsoft face enormous energy management challenges which encourage using tape for storing enormous quantities of less active data. Key point: Shifting less-active, archival and inactive data from disk to tape storage and virtualizing servers are the most significant ways of reducing energy consumption in the data center. Tape as NAS Emerges to Improve Performance. The innovative Tape as NAS solution stands in front of a tape library and presents itself as a file system (normally LTFS). This allows organizations to take advantage of low cost tape storage while maintaining the ease of use and faster access that disk-based storage offers. An LTO tape library integrates with a front-end NAS for standard NAS (CIFS/NFS) mounts and LTFS to deliver the newest active archive architecture. Data arrives at the NAS disk cache and is written to tape, files remain on disk cache until the cache is full, at which time the oldest files are reduced to metadata pointers only on disk while the files migrate to tape. File searches continue to see all archived files and only when a read request is received are files moved back from tape to disk cache for user access. A tape library as a NAS enables users to leverage familiar file system tools, and even drag and drop files directly to and from a tape cartridge, just like a disk-based NAS. For DR, second and third copies can be created to be stored in different places. Key point: Tape as NAS provides serves as an active archive providing disk-like management and much faster access times for archival data. Horison Inc. www.horison.com
Page 10
What about Tape in the Cloud? Tape’s role in the cloud provider market is quickly growing as tape is now being used for archival storage solutions to address and reduce the higher HDD cloud provider cost issues. Using HDDs for archival storage is a strategy – just not a very cost-effective one, and the cloud providers now understand this as their storage requirements soar. Since tape media is portable, using tape for cloud storage becomes highly advantageous if the cloud provider shuts down or should you want to quickly move your archive set to another cloud provider. Moving large amounts of archival data on available network bandwidth can take days or even weeks. Using a truck or airplane to move portable tape media is much faster and far less expensive than using bandwidth for large amounts of data! Key point: Cloud providers are deploying tape for their lowest-cost, most secure, long-term archival storage offerings. Storing archival data in the cloud represents a significant future growth opportunity for tape storage providers and a much lower cost solution for cloud providers. Summary Today’s reality is that the magnetic tape industry has made considerable progress in many areas over the past 10 years surpassing disk in many categories. The continued role for disk is well established but is facing growing technological challenges making capacity increases and any additional disk performance gains questionable. Tape has expanded its position as an effective complement to flash and disk for the foreseeable future due to its lower TCO, high reliability, higher capacities, faster data rates, and significantly lower energy costs. Because of this progress, the tape industry is aggressively repositioning itself to address many new high capacity, long-term and big data storage repositories which now represent more than 65% of the world’s total stored digital data. The current and future technology improvements in the tape industry suggest that tape will continue to be the most cost-effective storage solution for the enormous archival and Big Data opportunities that lie ahead, whether on-site, at a remote location, or in the cloud. Bottom-line: The tape renaissance is well underway, for the tape industry it’s clearly a new game with new rules. End of report
Horison Inc. www.horison.com
Page 11
