In-Depth
Tape’s Role in the Green Data Center
Long-standing technology is surprisingly efficient when it comes to “going green” with storage.
by Mark Ferelli
As data protection becomes a growing part of the IT budget, it seems like every technology that records data wants to claim primacy and indispensability. Too frequently in the frantic dash for market share, technology that has stood the test of time is falsely accused of being hopelessly out of date. So it is with magnetic tape in the data center. The veteran technology is inaccurately identified as “yesterday’s technology,” but tape technology is experiencing a Renaissance; a rebirth based on the growing recognition that tape technology is responding regularly and reliably to today’s vital business requirements.
Much as I hate to contradict the homey wisdom of Kermit the Frog, being green is easier than you think. Tape technology is already helping to keep the data center green as utility bills rise. There is a temptation in the marketplace to label tape as “yesterday’s technology,” but it is hard to make a compelling case to that effect when tape continues to solve today’s problems.
The growing awareness of the environmental impact of power consumption in the data center has led enterprises to address rising power and HVAC costs.
Greg Schultz, founder and analyst at StorageIO, reports that in the August 2007 EPA report to Congress on energy usage in U.S. data centers. During 2006, IT data centers consumed about 61 billion kilowatt hours of electricity at an approximate cost of $4.5 billion. Also reported: on average, IT data centers consume 15-20 times the energy per square foot than does a comparable office building.
Server consolidation and intelligent server virtualization have helped keep the data center green, even in intense 24/7 operation, but energy efficiency at the server level is not a complete solution. The storage infrastructure must follow suit.
Disk Dilemma
While there is no single “silver bullet” to make green the storage infrastructure, using the right technology can make an immediate impact. Deployment decisions are as important as the green labels on new product acquisitions.
Active advocacy abroad suggests that hard disk technology has a manifest destiny to store data throughout the data lifecycle, from primary storage through archival storage. Advocates point to disk’s capacities and the high I/O rates that are a part of spin physics.
However, hard disk drives are intricate electromechanical products. They incorporate diverse technologies, including precision motors, spinning platters on spindles, head-positioning electronics, advanced read/write heads, slider assemblies, and more. With so many building blocks, there are many ways a hard disk drive can fail. There are also the non-component threats of thermal buildup and rotational vibration.
Further, disk drives are enthusiastic consumers of power. StorageIO’s Schultz observes that the major power draws in the data center, according to his research for commonly deployed storage systems, are spinning hard disk drives and their enclosures, which account for on average 66-75 percent. The balance is power drawn by controllers.
Disk drive failures are a matter of “when” rather than “if.” A study from Carnegie Mellon presented at the 5th USENIX Conference on File and Storage Technologies (February 2007) points out that customers are replacing disk drives at rates far higher than estimated MTTF (mean time to failure) would justify.
The Carnegie Mellon study reported on larger high-transaction systems, including high-performance computing sites and Internet services sites running SCSI, Fibre Channel (FC), and SATA drives. The data sheets for those drives listed MTTF between 1 million to 1.5 million hours, which the study translated to annual failure rates "of at most 0.88 percent." However, the study showed typical annual replacement rates of between 2 and 4 percent, "and up to 13 percent observed on some systems."
Although data centers seek return authorizations on drives for many reasons, the study noted that a harsh environment at the customer site and intensive, random read/write operations that cause premature wear to the mechanical components in the drive were cited frequently.
The efficient use of disk can help with data center greening when a user reads and writes to the densest possible disk array to ensure capacity is maximized and more disk is not bought unnecessarily.
In archiving, on the other hand, the greenest option is tape, which uses less power and produces a lower heat output. This not only eases the bite of the utility bill but places less strain on HVAC systems. In contrast, the case can be made that using disk for archiving does more harm since disks that spin constantly use much more power and generate more heat.
Nathan Thompson, CEO of Spectra Logic and a vendor of both disk and tape solutions, notes: “Tape has a shelf life of about 30 years, if stored in a reasonable environment. Tape drives do not have the same power-up requirements as disk, which has to be powered up at least once a year. Additionally, disks continue to have problems with stiction.”
In the context of hard disk drives, stiction is the tendency of read/write heads to stick to the platters. Stiction is thought to take place due to two properties of the platters: smoothness and magnetic forces. Once the heads have stuck to the platters in a power-off mode, the disk will be prevented from spinning up and can cause physical damage to the media. Data loss is the result. Other forces that may cause stiction include electrostatics or adhesion from the inherent stickiness of silicon.
Tape Recognized as Greenest
According to Dave Vellante, co-founder and senior storage analyst at Wikibon, “Tape is the greenest of all storage technologies. By its continued existence and persistence, it appears an excellent candidate for innovation, which has been lacking in recent years. Tape remains a multi-billion dollar market, and, with most of the world’s information residing on tier 3 storage, tape, as a stand alone technology … could attract investment.”
That kind of investment bodes even better for power-conserving and thermal control.
Bob Covey, vice president of marketing at Qualstar Corporation and a veteran storage executive, notes that tape libraries are as power-efficient as individual drives. Says Covey: “Tape libraries are the ‘greenest’ storage on the planet. Our tape libraries use less power than the light bulb next to my favorite reading chair when not moving tapes. A tape drive waiting for the next backup job to start or the next restore request uses less power than a child's nightlight. Think about it: tapes in slots hold about 1 terabyte each and require no power whatsoever. Nothing else can even come close.”
More than MAID
In an effort to curb rampant power use and extend its usefulness beyond primary storage, disk drive makers are touting MAID (Massive Array of Idle Disks) architecture as a solution. In MAID, disk drives are powered down when not in use, similar to some classes of workstation. The customary advantages of MAID architecture are reduced power and cooling. The customary disadvantages are reduced application performance, the possibility of drive failure, and damage to the drive due to power surges upon spinup.
Spectra Logic’s Thompson noted that, “[MAID] is a significant technology. Copan [Copan Systems, a MAID array vendor] created ‘disk aerobics’ and offers a valid solution to stiction issues. But the fact is that you can’t shelve it as you can tape, or salt mine it.”
He also points out that power keeps going in MAID. He continues: “Copan spins down, but there is still power in use. Everyone else uses a ‘power saving mode’ where the disk spins down to half the spindle speed.”
StorageIO’s Schultz has examined the issue as well, and draws attention to the fact that even MAID does not match the energy cost savings of tape in an off-line environment.
Says Schultz: “Comparing off-line disk and tape-based secondary storage, solutions like the Sun LTO4 tape-based system has better power savings versus the best on-line or off-line disk-based solution.”
Summary
Improving power efficiency is one of the largest challenges facing today’s data center managers, particularly in large data centers or in metropolitan areas where power demand is highest and resources are being pushed to the limit.
For some, it is not just about escalating utility bills, but a more foreboding threat: running out of power capacity to support growth. New solutions and strategies to help store information more intelligently can deliver significant impact as it relates to improving power efficiency.
Start with a thoroughgoing analysis and plan projects. Evaluate workloads and configurations to determine present and future energy consumption across data center assets, ideally both servers and storage. Other assessments should include data center capacity and utilization along with facility and energy costs.
Resist the temptation to “rip and replace” existing energy-efficient tape subsystems for no other reason than the shrill advocacy that can be heard. Jon William Toigo, founder of the Green Data Project, notes: “The energy efficiency of tape versus disk has been demonstrated over and over despite marketecture to the contrary. Seventy percent of data is stored on tape, and for energy cost savings, it is the best game in town.”
Toigo resists some of the trendy logic being offered. “You hear about the statistic that 1 out of 10 tapes fails on restore. Anyone who has worked in a tape environment would fire a tape administrator with that high a failure rate.” Toigo recommends a read verify after a write operation. The extra time is worth the effort.
The assertion that “tape is dead” is one that the trade press hears constantly, but the technology persists, improves, and most assuredly doesn’t die. In the data center, tape technology continues to make it easier to be green
Mark Ferelli is a freelance technology journalist who has written about mass storage and information management technologies since 1988. You can reach the author at mcferelli@yahoo.com.