NOTE: My original article contained embedded calculation errors that significantly distorted the end results. These problems have since been corrected. I apologize to anyone who was accidentally misled by this information, and sincerely thank those diligent readers who brought the issues to my attention.
Some issues seem so obvious they’re hardly worth considering. Everyone knows that Solid State Drives (SSD) are more energy-efficient than spinning disk. They don’t employ rotating platters, electro-mechanical motors and mechanical head movement for data storage, so they must consume less power – right? However, everyone also knows the cost of SSD is so outrageous that they can only be deployed for super-critical high performance applications. But does the reputation of having exorbitant prices still apply?
While these considerations may seem intuitive, they are not entirely accurate. Comparing the Total Cost of Ownership (TCO) for traditional electro-mechanical disks vs. Solid State Disks provides a clearer picture of the comparative costs of each technology.
- For accuracy, this analysis compares the purchase price (CAPEX) and power consumption (OPEX) of only the disk drives, and does not include the expense of entire storage arrays, rack space, cooling equipment, etc.
- It uses the drive’s current “street price” for comparison. Individual vendor pricing may be significantly different, but the ratio between disk and SSD cost should remain fairly constant.
- The dollar amounts shown on the graph represent a 5-year operational lifecycle, which is fairly typical for production storage equipment.
- Energy consumption for cooling has also been included in the cost estimate, since it requires roughly the amount of energy as the drives consume to maintain them in an operational state.
- 100 TB of storage capacity was arbitrarily selected to illustrate the effect of cost on a typical mid-sized SAN storage array.
The following graph illustrates the combined purchase price, plus energy consumption costs for several popular electro-mechanical and Solid State Devices.
From the above comparison, several conclusions can be drawn:
SSDs are Still Expensive – Solid State Drives remain an expensive alternative storage medium, but the price differential between SSD and electro-mechanical drives is coming down. As of this writing there is only an x5 price difference between the 800GB SSD and the 600GB, 15K RPM drive. While this is still a significant gap, it is far less that the staggering x10 to x20 price differential seen 3-4 years ago.
SSDs are very “Green” – A comparison of the Watts consumed during a drive’s “typical operation” indicate that SSD consumes about 25% less energy than 10K RPM, 2.5-inch drives, and about 75% less power than 15K RPM, 3.5-inch disks. Given that a) each Watt used by the disk requires roughly 1 Watt of power for cooling to remove the heat that is produced, and b) the cost per Kwh continues to rise every year, this significant difference become a factor over a storage array-s 5-year lifecycle.
Extreme IOPS is a Bonus – Although more expensive, SSDs are capable of delivering from 10- to 20-times more I/O’s-per-second, potentially providing a dramatic increase in storage performance.
Electro-Mechanical Disks Cost Differential – There is a surprisingly small cost differential between 3.5 inch, 15K RPM drives and 2.5 inch 10K RPM drives. This may justify eliminating 10K disks altogether and deploying a less complex 2-tiered array using only 15K RPM disks and 7.2K disks.
Legacy 3.5 Inch Disks – Low capacity legacy storage devices (<146GB) in a 3-5-inch drive form-factor consume too much energy to be practical in a modern, energy-efficient data center (this includes server internal disks). Any legacy disk drive smaller than 300 GB should be retired.
SATA/NL-SAS Disks are Inexpensive – This simply re-affirms what’s already known about SATA/NL-SAS disks. They are specifically designed to be inexpensive, modest performance devices capable of storing vast amounts of low-demand content on-line.
The incursion of Solid State Disks into the industry’s storage mainstream will have interesting ramifications not only for the current SAN/NAS arrays, but also may impact a diverse set of technologies that have been designed to tolerate the limitations of an electro-mechanical storage world. As they say, “It’s a brave new world”.
Widespread deployment of SSD will have a dramatic impact on the storage technology itself. If SSDs can be implemented in a cost-effective fashion, why would anyone need an expensive and complex automated tiering system to decrement data across multiple layers of disk? Because of its speed, will our current efforts to reduce RAID rebuild times still be necessary? If I/O bottlenecks are eliminated at the disk drive, what impact will it have on array controllers, data fabric, and HBAs/NICs residing upstream of the arrays?
While it is disappointing to find SSD technology still commands a healthy premium over electro-mechanical drives, don’t expect that to remain the case forever. As the technology matures prices will decline when user acceptance grows and production volumes increase. Don’t be surprised to see SSD technology eventually eliminate the mechanical disk’s 40-year dominance over the computer industry.
For those of you interested in examining the comparison calculations, I’ve included the following spreadsheet excerpts contain detailed information used to create the graph.
With predictable regularity someone surfaces on the Web, claiming they have discovered a way to turn slow SATA arrays into high performance storage. Their method usually involves adding complex and sophisticated software to reallocate and optimize system resources. While there may a few circumstances where this might work, in reality it is usually just the opposite.
The problem with this concept is similar to the kit car world several decades ago. At the time, kit-build sports cars were all the rage. Automobile enthusiasts were intrigued by the idea of building a phenomenal sports car by mounting a sleek fiberglass body on the chassis of a humble Volkswagen Beetle. Done properly, the results were amazing! As long as their workmanship was good, the end results would rival the appearance of a Ferrari, Ford GT-40, or Lamborghini!
However, this grand illusion disappeared the minute its proud owner started the engine. Despite its stunning appearance, the kit car was still built on top of an anemic VW bug chassis, power train, and suspension!
Today we see a similar illusion being promoted by vendors claiming to offer “commodity storage” capable of delivering the same high performance as complex SAN and NAS systems. Overly enthusiastic suppliers push the virtues of cheap “commodity” storage arrays with amazing capabilities as a differentiator in this highly competitive market. The myth is perpetuated within the industry by a general lack of understanding of the underlying disk technology characteristics, and a desperate need to manage shrinking IT budgets, coupled with a growing demand for storage capacity.
According to this technical fantasy, underlying hardware limitations don’t count. In theory, if you simply run a bunch of complex software functions on the storage array controllers, you somehow repeal the laws of physics and get “something for nothing”.
That sounds appealing, but it unfortunately just doesn’t work that way. Like the kit car’s Achilles heel, hardware limitations of underlying disk technology govern the array’s capabilities, throughput, reliability, scalability, and price.
• Drive Latencies – the inherent latency incurred to move read/write heads and rotate disks until the appropriate sector address is available can vary significantly.
For example, comparing performance of a 300GB, 15K RPM SAS disk to a 3TB 7200 RPM SATA disk produces the following results:
• Controller Overhead – Masking SATA performance by adding processor capabilities may not be the answer either. Call it what you will – Controller, SP, NAS head, or something else. A storage controller is simply a dedicated server performing specialized storage operations. This means controllers can become overburdened by loading multiple sophisticated applications on them. More complex processes also means the controller consumes additional internal resources (memory, bandwidth, cache, I/O queues, etc.). As real-time capabilities like thin provisioning, automated tiering, deduplication and data compression applications are added, the array’s throughput will diminish.
• “Magic” Cache – This is another area where lots of smoke-and-mirrors can be found. Regardless of the marketing hype, cache is still governed by the laws of physics and has predictable characteristics. If you put a large amounts of cache in front of slow SATA disk, your systems will run really fast – as long as requested data is already located in cache. When it isn’t you must go out to slow SATA disk and utilize the same data retrieval process as every disk access. The same is true when cache is periodically flushed to disk to protect data integrity. Cache is a great tool that can significantly enhance the performance of a storage array. However, it is expensive, and will never act as a “black box” that somehow makes slow SATA disk perform like 15K RPM SAS disks.
• Other Differences – Additional differentiators between “commodity storage” and high performance storage include available I/Os per second, disk latency, RAID level selected, IOPS per GB capability, MTBF reliability, and the Bit Error Rate.
When citing the benefits of “tricked out” commodity storage, champions of this approach usually point to obscure white papers written by social media providers, universities, and research labs. These may serve as interesting reading, but seldom have much in common with production IT operations and “the real world”. Most Universities and research labs struggle with restricted funding, and must turn to highly creative (and sometimes unusual) methods to achieve specific functions from a less-than-optimal equipment. Large social media providers seldom suffer from budget constraints, but create non-standard solutions to meet highly specialized, stable, and predictable user scenarios. This may illustrate an interesting use of technology, but have little value for mainstream IT operations.
As with most things in life, “you can’t get something for nothing”, and the idea of somehow enhancing commodity storage to meet all enterprise data requirements is no exception.
If you believed all the media hype and vendor pontifcations three years ago, you would have thought for sure that Fibre Channel was teetering on the edge of oblivion. According to industry hype, 10Gbps Ethernet and the FCoE protocol were certain to be the demise of Fibre Channel. One Analyst even went so far as to state, “IP based storage networking technologies represent the future of storage”. Well as they say, “Don’t believe everything you read”.
In spite of a media blitz designed to convince everyone that Fibre Channel was going extinct, industry shipments and FC implementation by IT storage professionals continued to blossom. As 16Gbps Fibre Channel rapidly grew in acceptance, the excitement around 10GbE diminished. In a Dell’Oro Group report for 4Q12, fibre channel Director, switch, and adapter revenues surpassed $650 million, while FCoE champion Cisco suffered through soft quarterly results.
So what makes Fibre Channel network technology so resilient?
• Simplicity – FCP was designed with a singular purpose in mind, and does not have to contend with a complex protocol stack.
• Performance – a native 16Gbps FC port is 40% faster than a 10GbE network, and it too can be trunked to provide aggregate ISL bandwidth up to 128 Gbps.
• Low Latency – FC fabric is not penalized by the additional 2-hop latency imposed by routing data packets through a NAS server before it’s written to disk.
• Parity of Cost – The dramatic reduction in expense promised by FCoE has failed to materialize. The complexity and cost of pushing data at NN_Ghz is fairly consistent, regardless of what protocol it used.
• Efficiency – Having a Fibre Channel back-end network supports such capabilities as LAN-less backup technology, high speed data migration, block-level storage virtualization, and in-fabric encryption.
An excellent indicator that Fibre Channel is not falling from favor is Cisco’s recent announcement of their new 16Gbps MDS 9710 Multilayer Director and MultiService Fabric Switch. Cisco was a major proponent of 10GbE and the FCoE protocol, and failed to update their aging MDS 9500 family of Fibre Channel Directors and FC switches. (http://searchstorage.techtarget.com/news/2240182444/Cisco-FC-director-and-switch-moves-to-16-Gbps-new-chassis) This left Brocade with a lion’s share of a rapidly growing 16 Gbps Fibre Channel market. For Brocade, it produced a record quarter for FC switch revenues, while Cisco struggled with sagging sales.
Another influencing factor in FC longevity of is the average IT department’s need for extremely high-bandwidth storage network capabilities. Prior to 10GbE technology, Ethernet LANs performed quite well at 1GbE (or some trunked variation of 1GbE). The majority of the fibre channel world still depends upon 4Gbps FC, with 8Gbps technology recently starting to make significant inroads in the data center. Given the fairly leisurely pace of migration to higher performance for the SAN and NAS fabric technology. Except for a fairly small percent of IT departments that actually require high performance / high throughput, the lure of a faster interface alone has a limited amount of allure.
So which network technology will win? Who knows (or even cares)? There are usually bigger issues to overcome than what the back-end “plumbing” is made of. It’s far more important to implement the most appropriate technology for the task at hand. That could be Ethernet, Fibre Channel, Infiniband, or some other future network scheme. The key is to select your approach based on functionality and efficiency, not what is being hyped as “the next great thing” in the industry. In spite of all the hyperbole, Fibre Channel isn’t going away any time soon.
As Samuel Clemens (aka Mark Twain) said after hearing that his obituary had been published in the New York Journal, “The reports of my death are greatly exaggerated”.
Like most other things, technology suffers from advancing age. That leading-edge wonder of just a few years ago is today’s mainstream system. This aging process creates great headaches for IT departments, who constantly see “the bar” being moved upward. Just when it seems like the computing environment is under control, equipment needs to be updated.
Unless a company is well disciplined in enforcing their technical refresh cycle, the aging process can also lure some organizations into a trap. The thinking goes something like this – “Why not put off a technology update by a year or two? Budgets are tight, the IT staff is overworked, and things seem to be going along just fine.” It makes sense, doesn’t it?
Well, not exactly. If you look beyond the purchase and migration expenses, there are other major cost factors to consider.
Power Reduction: There have been major changes in storage device energy efficiency over the past decade. Five years ago the 300GB, 15K RPM 3.5-inch drive was leading-edge technology. Today, that has disk been superseded by 2.5-inch disks of the same speed and capacity. Other than its physical size, other major changes are the disk’s interface (33% faster than Fibre Channel) and its power consumption (about 70% less than a 3.5-inch drive). For 100TB of raw storage, $3577 per year could be saved by reduced power consumption alone.
Cooling Cost Reduction: A by-product of converting energy to power is heat, and systems used to eliminate heat consume power too. The following chart compares the cost for cooling 100TB of 3.5-inch disks with the same capacity provided by 2.5-disks. Using 2.5-inch disks, cooling costs could be reduced by $3548 per year, per 100TB of storage.
Floor Space Reduction: Another significant data center cost is for floor space. This expense can vary widely, depending on the type resources provided and level of high availability guaranteed by the Service Level Agreement. For the purpose of cost comparison, we’ll take a fairly conservative $9600 per equipment rack per year. We will also assume fractional amounts are available, although in the real world full rack pricing might be required. Given the higher density provided by 2.5-inch disks, a cost savings of $9,371 would be achieved.
In the example above, simply replacing aging 300GB, 15K RPM 3.5-inch FC disk drives with the latest 300GB, 15K RPM 2.5-inch FC disk drives will yield the following operational costs (OPEX) savings:
Reduced power $ 3,577
Reduced cooling $ 3,548
Less floor space $ 9,371
Total Savings $ 16,496 per 100TB of storage
Over a storage array’s standard 5-year service cycle, OPEX savings could result in as much as $82K dollars or more.
Addition benefits from a storage refresh might also include tiering storage (typically yielding around a 30% savings over non-tiered storage), reduced support contract costs, and less time spent managing older, more labor-intensive storage subsystems. There is also an opportunity for capital expense (CAPEX) savings by cleverly designing cost-optimized equipment, but that’s a story for a future article.
Don’t be misled into thinking that a delay of your storage technical refresh cycle will save money. In the end it could be a very costly decision.
It’s hard to retire a perfectly good storage array. Budgets are tight, there’s a backlog of new projects in the queue, people are on vacation, and migration planning can be difficult. As long as there is not a compelling reason to take it out of service, it’is far easier to simply leave it alone and focus on more pressing issues.
While this may be the path of least resistance, it can come at a high price. There are a number of good reasons why upgrading storage arrays to modern technology may yield superior results and possibly save money too!
Capacity – When your aging disk array was installed several years ago, 300 GB, 10K RPM, FC disk drives were mainstream technology. It was amazing to realize you could squeeze up to 45 TB in a single 42U equipment rack! Times have changed. The same 10K RPM DISK drive has tripled in capacity, providing 900 GB in the same 3.5 inch disk drive “footprint”. It’s now possible to get 135 TB (a 300% capacity increase) into the same equipment rack configuration. Since data center rack space currently costs around $3000 per month, that upgrade alone will dramatically increase capacity without incurring any increase in floor-space cost.
Density – Previous generation arrays packaged from (12) to (15) 3.5 inch FC or SATA disk drives into a single rack-mountable 4U array. Modern disk arrays support from (16) 3.5 inch disks per 3U tray, to (25) 2.5 inch disks in a 2U tray. Special ultra-high density configurations may house up to (60) FC, SAS, or SATA DISK drives in a 4U enclosure. As above, increasing storage density within an equipment rack significantly increases capacity while requiring no additional data center floor-space.
Energy Efficiency – Since the EPA’s IT energy efficiency study in 2007 (Report to Congress on Server and Data Center Energy Efficiency, Public Law 109-431), IT manufacturers have increased efforts to improve the energy efficiency of their products. This has resulted in disk drives that consume from 25% to 33% less energy, and storage array controllers lowering power consumption by up to 30%. That has had a significant impact on energy costs, including not only the power to run the equipment, but also power to operate the cooling systems needed to purge residual heat from the environment.
Controller Performance – Storage array controllers are little more than specialized servers designed specifically to manage such functions as I/O ports, disk mapping, RAID and cache operations, and execution of array-centric internal applications (such as thin provisioning and snapshots). Like any other server, storage controllers have benefited from advances in technology over the past few years. The current generation of disk arrays contain storage controllers with from 3 to 5 times the processing power of their predecessors.
Driver Compatibility – As newer technologies emerge, they tend to focus on developing software compatibility with the most recently released products and systems on the market. With the passage of time, it becomes less likely for storage arrays to be supported by the latest and greatest technology on the market. This may not impact daily operations, but it creates challenges when a need arises to integrate aging arrays with state-of-the-art systems.
Reliability – Common wisdom used to be that disk failure characteristics could be accurately represented by a ”bathtub graph”. The theory was the potential for failure was high when a disk was new. It then flattened out at a low probability throughout the disk’s useful life, then took a sharp turn upswing as it approached end-of-life. This model implied that extending disk service life had no detrimental effects until it approached end-of-life for the disks.
However over the past decade, detailed studies by Google and other large organizations with massive disk farms have proven the “bathtub graph” model incorrect. Actual failure rates in the field indicate the probability of a disk failure increases by 10% – 20% for every year the disk is in service. It clearly shows the probability of failure increases in a linear fashion over the disk’s service life. Extending disk service-life greatly increases the risk for disk failure.
Service Contracts –Many popular storage arrays are covered by standard three-year warranties. This creates a dilemma, since the useful service life of most storage equipment is considered to be either four or five years. When the original warranty expires, companies must decide whether to extend the existing support contract (at a significantly higher cost), or transitioning to a time & materials basis for support (which can result in some very costly repairs).
Budgetary Impact – For equipment like disk arrays, it is far too easy to fixate on replacement costs (CAPEX), and ignore the ongoing cost of operational expenses (OPEX). This may avoid large upfront expenditures, but it slowly bleeds the IT budget to death by having to maintain increasingly inefficient, fault-prone, and power hungry equipment.
The solution is to establish a program of rolling equipment replenishment on a four- or five-year cycle. By regularly upgrading 20% to 25% of all systems each year, the IT budget is more manageable, equipment failures are controlled, and technical obsolescence remains in check.
Getting rid of familiar things can be difficult. But unlike your favorite slippers, the LazyBoy recliner, or your special coffee cup, keeping outdated storage arrays in service well beyond their prime can cost your organization plenty.
It’s easy to hold onto the concept that IT is all about systems, networks, and software. This has been accepted wisdom for the past 50-years. It’s a comfortable concept, but one that is increasing inaccurate and downright dangerous as we move into an era of “big data”! In today’s world not about systems, networks, applications, or the datacenter – it’s all about the data!
For decades accumulated data was treated as a simply bi-product of information processing activities. However, there is growing awareness that stored information is not just digital “raw material”, but a corporate asset containing vast amounts of innate value. Like any other high-value asset, it can be bought or sold, traded, stolen, enhanced, or destroyed.
A good analogy for today’s large-scale storage array is to that of a gold mine. Data is the nuggets of gold embedded in the mine. The storage arrays containing data are the “mine” that houses and protects resident data. Complex and sophisticated hardware, software, tools, and skill-sets are simply tools used to locate, manipulate, and extract the “gold” (data assets) from its surrounding environment. The presence of high value “nuggets” is the sole reason the mining operation exists. If there was no “gold”, the equipment used to extract and/or manipulate it would be of little value.
This presents a new paradigm. For years storage was considered some secondary peripheral that was considered only when new systems or applications were being deployed. Today storage has an identity of its own that is independent from the other systems and software in the environment.
Data is no longer just a commodity or some type of operational residue left over from the computing process. “Big Data” forces a shift in focus from IT assets deployment and administration to the management of high-value data assets. It dictates that data assets sit at the center of concentric rings, ensuring security, recoverability, accessibility, performance, data manipulation, and other aspects of data retention are addressed as abstract requirements with unique requirements. Now information must be captured, identified, valued, classified, assigned to resources, protected, managed according to policy, and ultimately purged from the system after its value to the organization has been expended.
This requires a fundamental change in corporate culture. As we move into an era of “big data” the entire organization must be aware of information’s value as an asset, and the shift from technology-centric approaches for IT management. Just like gold in the above analogy, users must recognize that all data is not “created equal” and delivers different levels of value to an organization for specific periods of time. For example, financial records typically have a high level of inherent value, and retain a level of value for some defined period of time. (The Sarbanes-Oxley act requires publicly-traded companies to maintain related audit documents for no less than seven years after the completion of an audit. Companies in violation of this can face fines of up to $10 million and prison sentences of 20 years for Executives.)
However, differences in value must be recognized and managed accordingly. Last week’s memo about the cafeteria’s luncheon specials must not be retained and managed in the same fashion as an employee’s personnel record. When entered into the system, information should be classified according to a well-defined set of guidelines. With that information it can be assigned to an appropriate storage tier, backed up on a regular schedule, kept available on active storage as necessary, later written to low-cost archiving media to meet regulatory and litigation compliance needs. Once data no longer delivers value to an organization, it can be expired by policy, freeing up expensive resources for re-use.
This approach moves IT emphasis away from building systems tactically by simply adding more-of-the-same, and replacing it with a focus on sophisticated management tools and utilities that automate the process. Clearly articulated processes and procedures must replace “tribal lore” and anecdotal knowledge for managing the data repositories of tomorrow.
“Big Data” ushers in an entirely new way of thinking about information as stored, high-value assets. It forces IT Departments to re-evaluate their approach for management of data resources on a massive scale. At a data growth rate of 35% to 50% per year, business-as-usual is no longer an option. As aptly noted in a Bob Dylan song, “the times they are a-changin”. We must adapt accordingly, or suffer the consequences.