September 12, 2008 | Comments: (0)
Affordable SSDs in the offing?
September is usually too early to draw year-end conclusions, but I'll go out on a limb to say that 2008 will be remembered as a key year for SSDs (solid-state drives) -- not so much for sales figures, which won't likely reflect the hype surrounding flash SSDs this year, but for the technology itself, which remains one of the most controversial to hit storage in recent times.
Flash SSDs are based on technology rich with technical details that are generally misunderstood. To clear up some of the uncertainty around flash SSDs, I spoke with Wallace C Kou. 
Kou is president and CEO of Silicon Motion, a company based in Taiwan and California, and one of the major players in SSD technology. Due to our busy schedules and distant time zones, Kou and I conducted our interview over e-mail.
* * * * * *
InfoWorld: Please introduce Silicon Motion to our readers.
Kou: Silicon Motion is an innovative technology company that designs, develops, and markets high-performance, low-power semiconductor solutions for the mobile multimedia consumer electronics market. We have three major product lines: mobile storage, mobile communications, and multimedia SoC (system-on-a-chip). Our mobile storage business is composed of microcontrollers used in NAND flash memory storage products such as flash memory cards, USB flash drives, SSDs, embedded flash applications, and card readers. Our mobile communications business is comprised of mobile TV tuners, CDMA RF ICs, and electronic toll collection RF ICs. Our multimedia SoC business is comprised of products that support MP3 players, PC cameras, and embedded graphics applications.
InfoWorld: How do you see the flash memory market today?
Kou: We believe that NAND supply is still bigger than demand and that the industry's excess supply condition will continue through to the end of 2008. Since overall demand is still soft, the movement of channel inventory is slow, which is affecting companies involved in the NAND food chain.
InfoWorld: What market segments are you targeting?
Kou: Our three product lines target three markets: solid-state storage devices, principally flash memory cards, USB flash drives, SSDs, and embedded flash devices, mobile TV involving DVB-H, T-DMB, S-DMB, ISDB-T, CMMB, and T-MMB standards, and embedded graphics applications. The end application for approximately two-thirds of our revenue, whether from sales of controllers for flash memory cards or mobile TV ICs, is the large and rapidly growing mobile phone industry.
InfoWorld: What is an SSD controller, and how does it work?
Kou: A controller plays a very important function in a SSD. It manages the data input/output between a host device and the NAND storage media. It also handles various host protocol, special custom command set, data error detection and correction, as well as NAND flash management, such as bad block management, error handling, wear-leveling, data protection during power off, performance, endurance, and reliability, etc.
InfoWorld: Please explain what a hybrid SSD is.
Kou: Hybrid SSD utilizes both SLC and MLC NAND to achieve low cost, high performance, and high endurance. Our hybrid SSD controller can, for example, manage a combination of, say 8GB of SLC and 56GB of MLC in a 64GB device. The controller utilizes our proprietary algorithms to identify frequently accessed and infrequently accessed files and puts the frequently accessed files into the SLC NAND and the infrequently accessed files into the MLC NAND. By using our hybrid SSD controllers, SSD OEMs can achieve endurance and performance similar to an SLC-only SSD, but at the cost close to an MLC-only SSD.
InfoWorld: Your company recently announced new SSDs controllers. Please describe their major features.
Kou: Our new SSD controllers support SSD OEMs who chose to use only SLC, only MLC, and a mix of SLC and MLC flash. Our controllers provide both advanced 8-bit to 24-bit ECC (error correction code) engines for enhanced reliability and "global wear-leveling," which means wear-leveling on a device basis as opposed to wear-leveling at each individual NAND flash component in the device. Global wear-leveling performs its task by taking the logical block address from the host and randomly mapping it into the physical block address of the bank of NAND flash components. In addition, our new controllers add new data protection technology to prevent data loss due to unexpected power surge or failure.
InfoWorld: Two of the new controllers have a PATA (parallel ATA) interface. Isn't that old technology?
Kou: In low-cost notebook PCs and embedded system applications, PATA is still popular because it consumes less power and has a cost advantage to the newer SATA interface. It is important for us to provide the various solutions that our customers require.
InfoWorld: Can you give an idea of how much the cost of an SSD controller impacts the final product?
Kou: The cost of the controller is a very small part of the overall cost of a SSD. The bulk of a SSD cost is from NAND flash. Despite the controller accounting for a relatively small proportion of an overall SSD cost, it plays a disproportionately large role as an enabler. Currently, most SSDs are using SLC NAND flash because of a lack of availability of suitable MLC controller solutions. When a robust MLC controller, such as ours, becomes available, SSD OEMs will be able to reduce the cost of their SSD solutions -- of the same capacity -- by one half to a third, or enable SSD OEMs to double or triple their SSD capacity for the same cost. [Emphasis mine, not Kou's (see below).]
Leading global notebook PC OEMs and consumer electronics OEMs are therefore all looking for better solutions that could enable them to use MLC NAND for SSD. SSD controllers could quickly become a mainstream storage device when robust MLC controllers are available that enable OEMs to use less expense MLC NAND flash. Our MLC controllers for SSDs have just become available for sampling.
InfoWorld: The wear-leveling logic seems to be a very critical part of the controller. Is that part stable or still work in progress?
Kou: NAND is rapidly transitioning to 40nm and 30nm process nodes. Flash makers can significantly reduce their cost by moving into these finer geometries, but achieve this benefit at a heavy cost. Endurance for these newer products is significantly reduced when compared to products based on 60nm and 70nm process technologies. The controller therefore has to provide significantly more advanced and robust wear-leveling algorithms, by randomly and equally accessing every physical block in all the individual NAND components in that SSD device, to enhance SSD endurance. We have already achieved some excellent results with our global wear-leveling approach. We will continue to develop newer and more advanced wear-leveling techniques to meet upcoming NAND flash technological challenges.
* * * * * *
I emphasize the above statement about cost reduction because the promise of getting three times the bang for the buck is not something we hear often, in storage or other areas. If Kou's prediction is correct, the price of flash SSDs will indeed soon become much more competitive with that of traditional spinning drives.
What is your view? Feel free to e-mail me, or better yet, leave a comment below.
Posted by Mario Apicella on September 12, 2008 03:00 AM
August 01, 2008 | Comments: (0)
Debating the merits of SSDs, part two
Healthy debate is often necessary to get a balanced view of an emerging technology. Somewhere between endorsements and detractions, a realistic understanding of the long-term outlook for a technology arrives. As such, I have decided to turn to a guest once again for the second installment in what I hope to be an ongoing debate over the merits of flash SSDs (solid-state drives).
For those who missed the first installment, I spoke with Joel Hagberg of Fujitsu, who questioned the long-term outlook for flash SSDs, pointing out that the much-hyped advantages of SSDs over CSDs (conventional spinning drives) -- better performance, improved reliability, and lower power consumption -- may very well be too minimal to justify the price.
This week I turn to Scott Nelson, vice president of the memory business unit at Toshiba America Electronic Components and Semiconductors, for insights on the pro-SSD side.
In our conversation, Nelson touched on several of the points Toshiba published in a recent Web document disputing what the company considers the most common misrepresentations of the technology.
First off, according to Nelson, the current 56nm flash memory format has yet to reach its full potential. By Toshiba's estimates, performance for this format should double to 200MBps for reads and 80MBps for writes. And that doesn't touch on what is in store for future generations, Nelson says.
By year's end, Nelson believes we will see the flash memory format shrink to 43nm, yielding performance that will reach 240MBps for both reads and writes. Fast-forward to 2010, and the footprint will shrink even more, well below 40nm, according to Nelson, bringing performance in excess of 400MBps, again for both reads and writes, which will exceed what SATA 2 can sustain, thus requiring the adoption of SATA 3 as a host interface for SSDs.
As the shrinking footprint might suggest, mobile devices will benefit most from these jumps in performance, Nelson says. As for Toshiba, it is betting high on MLC (multilevel cell) flash memory for laptops, which promises to be less expensive and denser, while providing more capacity than SLC (single-level cell) flash.
If those two acronyms are foreign to you, the quick take is that SLC stores a single bit per cell, whereas MLC packs two or more bits in every cell. As such, assuming similar form factors, MLC packs more data in the same space.
But if MLC is best for notebooks, where will SLC flash find its ultimate home? Nelson answers before I can voice that question: "SLC-based drives, that technology is very important for the enterprise, where they need the higher performance and the higher reliability."
Not surprisingly, Toshiba plans to enter the nascent SLC SSD space. Although Nelson did say this foray will begin in 2009, he would not reveal details regarding the capacity and performance of those upcoming enterprise drives.
Overall, Nelson believes the demand for SSDs is on the rise. But although Toshiba expects customers will buy more laptops with MLC-based SSDs going forward, Nelson admitted that that number will likely remain a small fraction of the total number of laptops sold.
Will things be different in the enterprise? Probably not. My guess is that SSDs will comprise only a fraction of the enterprise's overall drive footprint, replacing mainly small-capacity, high-RPM, expensive 3.5-inch drives when demand for performance makes SSD a viable option.
However reasonable that may sound, I can't help thinking that enterprise SSDs already have a formidable challenger in the form of fast, 2.5-inch drives, the deployment of which could significantly reduce the need for flash storage at the high end.
Of course another rival technology is the volatile, RAM-based SSD, which is understandably more expensive but bears the promise of even higher performance than flash, as the comparison between these two arrays from Texas Memory Systems suggests.
Which begs the question: Are enterprise SSDs already doomed to a marginal presence, even before reaching their prime?
Your expert opinions are welcome.
Posted by Mario Apicella on August 1, 2008 03:00 AM
July 25, 2008 | Comments: (0)
Flash vs. RAM: Why not choose both?
You won't find RAM-based SSD (solid state drive) arrays just anywhere. But if you visit an HBA vendor's lab, you surely will. After all, a single such box can deliver performance on par with a large lineup of conventional storage devices.
Just how fast are RAM-based SSD arrays, and how do they compare with flash-based storage? The RamSan-440 and the RamSan-500, both from TMS, provide an intriguing comparison.
The RamSan-440 is the latest and fastest member of the RamSan family, according to TMS. Released this week, the 4U box can serve 256GB or 512GB of storage in up to 1,024 LUNs. Two 4Gbps FC (Fibre Channel) ports are standard, but the unit can support six more. Or, if you prefer, the box can come equipped with four 10Gbps Infiniband ports.
According to TMS, the RamSan-440 can sustain random reads at 600,000 IOPS and 2KB blocks transfers at a rate of 4.5GBps. At that rate, you could copy the content of a near-full DVD in 1 second. Of course, there is no DVD device capable of keeping up with that pace, but that puts the RamSan-440's performance into perspective.
By comparison, the RamSan-500 -- a heavily cached, flash-based solution that hit the market at the end of 2007 -- would take twice as long to complete the same transfer. According to the vendor, the RamSan-500 can sustain only 100,000 IOPS –- a remarkable rate, but one that pales in comparison to what the RamSan-440 can do. Of course, that 6x performance gain comes with a similar hike in price. TMS dodged my questions on pricing, but did suggest $100 per GB for the RamSan-500, a device that you can order with as much as 2TB capacity, while the RamSan-440, at $500 per GB, carries a more challenging sticker price.
How does the RamSan-440 solve the apparent contradiction between volatile RAM and storage persistence? Well, redundant power supply, of course, plus batteries that in case of power failure will keep the array's memories alive for up to 25 minutes, according to TMS. But there's more: The RamSan-440 also includes five removable modules with enough flash-based memory to ensure a full RAM backup.
In essence, you get a very fast RAM-based storage backed up to permanent flash storage -- an interesting proposition. In basic configurations, the system will automatically initiate a backup to flash in the event of a power failure. The 512GB model can complete this backup in 6 minutes -- well before the batteries are completely discharged. To the best of my knowledge, TMS is the only vendor to offer flash backup for RAM-based storage.
It's an optional feature, but you can also set up the array to do constant background synching of RAM content to flash, which would protect data against double failures in case, for example, batteries misbehave during a power failure.
The five flash modules are protected by a RAID configuration that permits full backup of the RAM content even if one module fails, and can, obviously, be removed. Therefore, should your machine fail, you can easily transfer your backup data to a similar machine.
Did I mention that the RamSan-440's RAM is also RAID-protected?
But what about power? According to the vendor, the RamSan-440 slurps a paltry 650W at peak. For comparison, I've measured the energy absorbed by single 16-drive storage arrays mounting fast disks at around 450W.
Here's a comparison of a 512GB RamSan-440 unit and a 2TB RamSan-500 model, according to data provided by TMS:
| RamSan-440 | RamSan-500 | |
| capacity GB | 512 | 2000 |
| price/ capacity | $500 | $100 |
| estimated price | $256,000 | $200,000 |
| IOPS | 600,000 | 100,000 |
| Transfer rate GB/sec | 4.5 | 2 |
| price/ IOPS speed | $0.43 | $2 |
| price / GBsec transfer rate | $56,889 | $100,000 |
What should catch your eye is that the more expensive RamSan-440 model has much lower price-performance ratios than the RamSan-500. As such, it is a better bargain if speed is a top priority. Of course, the table offers only a rough estimate of those ratios: It will be interesting to see how the two arrays will fare in the SPC (Storage Performance Council) benchmarks.
Posted by Mario Apicella on July 25, 2008 03:00 AM
July 11, 2008 | Comments: (0)
Would you pay several times more for a technology that yields only dubious performance advantage? How about if that technology is experiencing a high rate of product returns from early adopters?
That's the central question regarding flash SSDs (solid state drives) as Joel Hagberg, vice president of business development at Fujitsu, sees it.
According to Hagberg, the much-hyped advantages of SSD over CSDs (conventional spinning drives) -- better performance, improved reliability, and lower power consumption -- are minimal and certainly do not justify adding $1,000 to the price of a notebook.
In fact, as Hagberg explains, Fujitsu customers have returned mostly negative feedback on SSD laptops, having found no significant performance advantage nor much improvement in battery life.
"The LCD panel, the CPU, and the DRAM are the three power hogs in a notebook, and those are still there with a SSD," Hagberg says. "When we did the measurements in our labs we saw a 10- to 15-minute improvement with the SSDs, not enough to warrant the price increase if you are not getting the performance increase."
Hagberg does, however, concede that SSDs provide improved reliability when compared with CSDs. And though he is not convinced of the proclaimed performance advantage of current SSDs, he also admits that there are areas -- random reads, for example -- where they are faster than CSDs.
Perhaps this means Hagberg might see a better fit for SSDs in a corporate setting, where customers are less sensitive to price differences if there is a strategic advantage to gain.
Well, no.
"Storage arrays, SAN or NAS, do more than random reads. For example, you have RAID 5 striping happening, where what you need to do to have parity is read-modify writes," Hagberg says.
As a reminder of what Hagberg is referring to, check out Figure 3 of this diagram. When a block of data in RAID 5 is updated, the storage subsystem needs to read both data and parity blocks, update them, and then write them back. In essence, two read and two write operations are required to update a single block of data.
"Then if you have to do a wear-leveling algorithm for writing, you introduce a level of complexity that reduces those I/O numbers significantly," Hagberg says.
That wear-leveling argument is the ultimate weapon that SSD detractors such as Hagberg throw at their opponents. In essence, the memory cells that flash-memory chips and SSDs are made of can sustain only a finite number of erase-write cycles. There seems to be a consensus around that number: It varies between 100,000 and 1 million cycles, depending on the "quality" of the flash memory used.
Regardless of which number we choose, to extend flash memory's ability to work without errors, manufacturers must implement a wear-leveling strategy -- in essence, distributing writes across cells so that no cell will be forced to an early failure.
If you are interested in learning more about wear-leveling algorithms, this presentation in PDF format from a university in Taiwan is useful, and in English.
But the question remains, is Hagberg right about SSDs? In a word, no, although he does raise some good points. However, I don't think that the clash between wear-leveling and RAID creates an insurmountable problem. For example, this vendor has been offering a RAID card that mounts flash SSD (PDF file) for quite some time.
Nevertheless, at the current state, many customers will probably agree that SSD can be a little pricey for their taste. But there is a range of customers that will appreciate the additional performance and will gladly pay a premium to get it. After all, many people travel first class, and they love it.
In a future column, I will host the thoughts of a pro-SSD guest. In the meantime, post your thoughts regarding SSDs in the comments below or e-mail me.
Posted by Mario Apicella on July 11, 2008 03:00 AM
June 06, 2008 | Comments: (0)
SSDs: Perfect fit for mobile computing?
There aren't many products I review that I can test for traveling. In fact, most are large enough for me to hide behind, if not inside. But the Lenovo ThinkPad T61 with a 64GB Samsung SSD (solid state drive) I am currently testing offered a rare opportunity to take my work with me on the road.
Testing SSDs has given rise to some intriguing scenarios. For example, an imaginative person at memory solutions vendor Super Talent came up with the idea of using a paint shaker to put together a torture test scenario for SSDs (movie clip with ActiveX control).
That odd combination of handyman and IT skills was meant to demonstrate how impervious SSDs are to shocks and vibrations that could, by contrast, easily damage a spinning drive. This resilience to misuse makes SSDs a good alternative in mobile devices.
Although I agree in spirit with Super Talent's test case, isn't that demonstration a bit theatrical? Of course it is: It's tantamount to shaking a book to prove that paper has better resistance to g-force than a disk drive. Then again, with SSDs commanding a cost per gigabyte still outrageously higher than that of spinning drives, it's easy to understand why SSD vendors leverage everything they can think of to promote their products.
Of course, the test does make a point. After all, laptop drives are indeed vulnerable to misuse. For example, the ThinkPad T61 that I am testing has applications and sensors capable of detecting potentially harmful shocks and stopping the disk drive before damage occurs. It's like having a seismic needle in your machine. I deactivated that protection, however, since I wanted to test the resilience to shock of the Samsung 64GB SSD.
My scenario? I hired a test assistant to ride shotgun in my pickup while downloading music or videos or playing games on that T61. Some people have a really easy life.
An important component in that test was a Pantech PX-500 mobile broadband card that kept my Sprint EvDO connection up and running throughout those long trips, but that's a topic for another column. You should be able to read more on that and other tests in an upcoming review, including how much it will cost to add that Samsung drive to the laptop, a detail that was covered by written embargo when I started my evaluation.
If that embargo surprises you, consider that prices in the SSD market are moving faster than a politician running for a photo op. Take, for example, this week's announcement by Sun about offering SSDs on its machines. I'll be the first to admit that the Sun news is only loosely related to our topic because it addresses the server space, where the requirements are considerably different from those of laptops. Think battery life, for one. Still, that announcement proves that EMC, as well as other storage vendors before it, are the vanguard of a much larger crowd marching to the SSD beat. Obviously a larger offer of SSDs should bring their price further down.
We'll keep an eye on whatever else develops in the corporate storage space, but a concurrent announcement from Super Talent, the paint-shaker test folks, is square on target for the mobile devices space. In fact, the vendor is announcing the MasterDrive KX family of 1.8-inch SSDs for laptops and ultra-mobile PCs. The drives are offered in 30GB, 60GB, and 120GB capacities.
The declared performance numbers of the MasterDrive KX seem to be on par with other devices in that class, but price is the real jaw-dropper in the Super Talent announcement: The 30GB unit should sell for about $300, a price that, although still much higher than comparable spinning drives, is a temptation, rather than a hopeless barrier for an IT manager. Sort of like contemplating that sports car instead of that cheaper sedan – it's not exactly the same cost, but it's still affordable.
Other capacities also carry a tempting sticker. The 60GB model adds only $150, and the 120GB model is a notch below $700. If durability is a concern, Super Talent suggests a life span of more than 16 years for the 30GB drive at 50GB per day of write/erase cycles. The life expectancy doubles for the 60GB unit and stretches to 66 years for the 120GB drive. Whatever model you choose, the mobile device you have will probably fail or become passé well before the drive reaches the end of its useful life.
Would you consider replacing your drive with an SSD at those prices? If not, what price point would you consider? Feel free to comment below or send me an e-mail.
Posted by Mario Apicella on June 6, 2008 03:00 AM
May 23, 2008 | Comments: (0)
Judging from the blogosphere's post-EMC World buzz, EMC's recent passion for SSDs (solid state drives) is no fly-by-night affair. In fact, at the event, which I was unable to attend, EMC suggested that by 2010, SSDs could reach price parity with, and eventually replace, FC (fibre channel) drives.
EMC is showing more than a superficial interest in SSD -- quite surprising from a company that doesn't usually jump on new technologies (remember iSCSI or, better yet, RAID 6?) and, instead, waits for other vendors to taste how sharp the cutting edge is.
In SSDs, EMC sees competitive advantage -- enough to break with its wait-and-see tradition. By offering storage solutions with drives 30 times faster than old-fashioned, still expensive FC spinners, EMC hopes to gain share against the competition. You may remember this number because EMC recently suggested in a white paper that 30-to-1 is the performance ratio between its SSDs and high-end spinning drives.
That's quite a performance jolt, but I doubt many customers are longing for such a jump. Many would rather have three times the performance using a tenth of the drives. Not only would such a storage solution be faster, it would dramatically cut down on floor space, not to mention electricity and cooling requirements -- three very expensive resources for most datacenters.
But do SSDs need to be priced on par with FC drives for that hypothetical scenario to work? No. In fact, assuming a 30-to-1 performance ratio and equal controller and enclosure costs, customers would be better off even if SSDs were 10 times more expensive than regular drives.
So why would EMC make such a bold, unrealistic prediction regarding the 2010 price of SSDs? Might it be because of signals from customers that SSDs could attract a larger number of applications than initially expected -- hence have to become much more affordable?
Pliant Technology, a startup that should begin shipping its first SSD by year end, is -- not surprisingly -- in agreement with that vision of SSD taking the center stage of application performance.
CEO Amyl Ahola doesn't mention prices, but that's what he had to say:
We see a hybrid environment where enterprise-level flash technology is used with the applications that require high-performance and disk drives are relegated to what they're best at: very low-cost storage and sequential operations as opposed to random operations. You should begin seeing this environment early next year and it should provide the performance and reliability needed for the enterprise environment.
Don't get me wrong, SSD drives will cost much less, but it will take a minor miracle to reach cost parity with spinning drives by 2010. After all, FC drives still go for a premium, and their price gap compared with other drives is minuscule versus that of SSDs.
Perhaps EMC was trying to impress Wall Street while serving a softball to SSD partners Seagate and STEC. Whatever the reason, even if a slip of the tongue, EMC's statement has brought optimism to a market segment that has been struggling for purchases. In fact, someone is already buying that Kool-Aid.
From my experience, the only thing wrong with SSDs is price. Anything that can help bring the cost down is a welcome novelty. Perhaps if we could persuade other vendors about the promise of SSDs in the enterprise, their price would indeed reduce significantly by 2010.
Which might just be the Machiavellian plan EMC had in mind.
Posted by Mario Apicella on May 23, 2008 03:00 AM
May 09, 2008 | Comments: (0)
Solid-state upgrades: Risky business
Hardware upgrades can be a blast. Slide 2GB more RAM in your machine and everything just works faster and smoother. Updating a laptop or desktop with an SSD (solid-state drive), however, can be tricky and not so rewarding, as I am finding.
I haven't tested any enterprise-class SSDs yet, but I want to share some of my findings testing flash SSD devices on laptops and desktops. In short, your machines may not be able to keep up with the drives' speed, and unfortunately there is no easy way to tell beforehand what will happen with your specific system.
What I learned during my tests is that using devices capable of sustaining very fast throughput such as an SSD is a surefire way to reveal old or slow machines: With new SSDs in place, some of the desktops in my lab suddenly showed their disk access slowness.
I was expecting the immediate top-notch performance trumpeted by the drive vendors, but in many cases, performance was on a par with that of spinning drives, even in terms of throughput, where a flash SSD is supposed to win hands-down.
Was there something wrong with my test units? I checked in with the drive vendors, and the culprit was revealed: Some motherboard chip sets that control SATA connections seem to have their performance capped and, like automobiles with the hand brake on, weren't able to produce top speeds in my benchmarks.
The good news is that in many cases a simple driver update can improve performance dramatically. On some motherboards, I saw an improvement of 10MBps to 20MBps after installing the latest drivers. Unfortunately, not every motherboard can be brought up to speed, and I don't have enough variety in my lab to compile a comprehensive list of what can be made to work faster and what can't; so before trying an SSD upgrade, browse your motherboard vendor's Web site to download, or ensure you have, the latest drivers.
And this goes for those of you who use automatic updates. I had to download a new driver manually on some of my machines, because apparently the new drivers in question are not always captured by the automated update cycles. For example, after manually updating the Nvidia nForce driver on this HP Pavilion, the burst speed -- which was well below 100MBps before the update -- cranked up to 122MBps.
If you have motherboards with Intel chip sets, don't miss this resource page, where the vendor explains which chip sets can be made to work in AHCI (Advanced Host Controller Interface) mode. Easy to do from the BIOS setup screen, this essentially releases that hand brake, cranking up performance while enabling important SATA features, as explained on the Intel resource page. Thanks to Roger Bradford, product marketing engineer for chip set and graphics marketing, for pointing me in the right direction with the Intel chip set.
I should also mention that the desktop I updated with new drivers manually is a relatively new machine. With older models, you might not be as lucky finding an updated driver. In other words, even the fastest SSD money can buy will not bring all the expected benefits if your motherboard can't keep up.
On those older machines without an upgrade option, you might be able to solve the problem with an add-on SATA card. But the few cards I tried did not give me improved SSD speed. Again, I can't categorically say that no add-on card will bring any benefit, but it will probably cost you, and unless you can verify via your vendor (or another person's experience) that your model is fast enough to keep up with an SSD, you might find yourself out of money with nothing to show for it -- obviously not where you want to be after implementing an expensive update.
How expensive? Well, the bare minimum to hold an OS and a few applications, 32GB, will cost you $900 or more per unit, retail, according to a search on Google. Brand and form factor, in addition to capacity, will also affect your price. Of course, we can't blame the SSD vendors for the update risk, but considering the money involved, that's not something that anyone can afford to jump on lightly.
Have you updated any of your machines to SSD? Care to share your experience? Feel free to comment below or send me an e-mail.
Posted by Mario Apicella on May 9, 2008 03:00 AM
April 28, 2008 | Comments: (0)
An interview with the real McCoy
There is a growing appetite for storage performance in the data center that old fashioned disk drives don't seem to be able to satisfy. Solid state drives are becoming the new stars of performance hungry applications, but with so many vendors and no unifying product standard is there a common ground for those drives? In an exclusive e-mail interview with Jim McCoy, chairman of Pliant Technology, a new company that will soon begin shipping their own SSD products, we try to find an answer to that question.
InfoWorld: What is Pliant Technology?
McCoy: Pliant Technology is developing a new class of solid state drive storage devices for the enterprise. The company is primarily focused on Enterprise Flash Drives, or EFDs, that offer very high performance, very high reliability and low power consumption compared with traditional disk drive architecture. The company came out of stealth mode in February, but we have been working on our device for about two years. Pliant was established by a team of executives and engineers from Fujitsu, Seagate and Maxtor who’ve had a track record of success in the storage industry. I expect we’ll find the same success with Pliant.
InfoWorld: Why create a new company? Aren't there already many suppliers of SSDs?
McCoy: You are correct that there are several SSD suppliers. However, we feel they do not solve the essential problem that SSD was designed to address, especially for enterprise applications. The issue is that disk drive performance is no longer adequate for high transaction rate applications. In this age, there is a huge amount of data that needs to be moved and accessed quickly, and disk drives aren’t doing it efficiently. And, so far, every supplier we’re aware of that has tried to stretch consumer type Flash SSD technology to meet the high demands of enterprise storage has not been able to do it either.
InfoWorld: How will Pliant products be different?
McCoy: Enterprise applications demand very high IOPS performance, low latency and high reliability, while operating 24/7 at or near 100 percent duty cycle levels. The Pliant EFD is different from traditional SSDs in that it has been designed specifically for this environment while being fully plug compatible with existing disk drive architectures. In addition to superior reliability, our EFD device will provide performance that is exponentially greater than other Flash-based SSDs. In fact, our device will provide greater performance than DRAM-based SSDs in the same rack space, at a lower cost and using a lot less power.
InfoWorld: What market segment are you targeting?
McCoy: In general, we are targeting the enterprise market. To go into a bit more detail, there are many applications today that have extremely high IO requirements and are in desperate need of better performance. These include any online transaction based or query based system, such as financial services, banking, manufacturing, etc. In fact, any system incorporating expensive, power-consuming supposed high-performance disk drives (15K RPM) could potentially reap huge benefits from the use of Pliant EFDs.
InfoWorld: Who will be your customers?
McCoy: Pliant Technology will sell its products on an OEM basis to storage system and server providers. We plan to bring our products to market by the end of this year.
InfoWorld: Why should end-users consider SSD?
McCoy: End users are facing many storage performance related issues today and employing multiple techniques, such as thin provisioning, virtualization, and expensive DRAM caching in addition to data center consolidation, to improve performance and efficiency. These tactics place even greater demands on already over-burdened storage systems, and too often result in painful over provisioning of disk drives. Over provisioning is the tactic of utilizing more disk drives than needed from a capacity standpoint by 3 to 1, or 4 to 1, or in some cases even 10 to 1, to achieve the required performance. Think about it: it’s not uncommon for end users to purchase and run 3 to 4 times as many disk drives as they need just to get functional performance. That IT executives have learned to accept over provisioning as “normal” is a clear statement of the need for a better solution.
This of course results in more racks, more controllers, more cache appliances, reduced reliability, and power consumption that’s going through the roof. By using EFDs together with low cost disk drives, over provisioning can not only be avoided, but higher capacity and performance levels can be achieved at lower costs and with power consumption reduced by as much as 80 percent.
InfoWorld: What capacities/performance are you targeting with your products?
McCoy: Capacity points will be similar to disk drives capacity points, and will grow with improvements anticipated by Flash chip suppliers. Specific performance specifications will be released when we formally announce our initial products later this year. While I can’t get into specifics, I can say that our performance will go far beyond the enterprise SSD vendor’s products out there today, and will look more like DRAM performance.
InfoWorld: Write performance and long time data persistence are typical weak point of SSD drives. How do you plan to overcome those limitations?
McCoy: These are only some of the issues that need to be specially addressed for enterprise applications. Traditional solutions may be adequate for low duty cycle applications, such as laptops, but absolutely are not sufficient for the enterprise. Enterprise read and write performance needs to be symmetrical and considerably higher than traditional SSDs and far better than the best of the disk drives. Data integrity and reliability must be at least as good as the best of the disk drives. In addition to providing unprecedented performance, Pliant’s new controller architecture and firmware are specifically designed to overcome Flash phenomena, such as write endurance, read disturb and program disturbs, to provide no-compromise enterprise-level reliability.
InfoWorld: Price is another aspect of SSD that customers may not like…
McCoy: I’m glad you brought this up because there is a bit of a misconception in the industry. I believe that customers will like the cost of EFD. They’re recognizing that cost per gigabyte is only one of many metrics in real-world storage applications. As I said earlier, OEM customers who pair our EFDs together with low cost higher capacity disk drives for their high transaction rate applications can provide major purchase cost savings, as much as 40 percent or more, to their customers. They’ll no longer need to over provision disk drives and can also enjoy the tremendous reliability improvement and energy efficiency of a more streamlined approach. So while initial cost savings will be significant, total cost of ownership savings will be outstanding.
InfoWorld: Can a single vendor have a significant impact on SSD prices?
McCoy: To some extent yes. As an example, the Pliant EFD does not require any on-board DRAM cache to achieve its performance levels. Other SSDs do require such cache, which is not only expensive itself but requires additional costs in terms of power backup and complexity due to its volatility.
InfoWorld: Storage vendors seem to be reluctant to move away from spinning drives: what could change that?
McCoy: How about a 40-percent cost reduction, an 80-percent power reduction, and an 80-percent lower failure rate? That’s a simple answer, but it gets right to the bottom line about what IT managers and CTOs need.
Disk disk drives are not going to evaporate in many bulk storage mid-range performance applications in the near term. But the basic physics and power requirements intrinsic to their mechanical nature makes them unsuitable compared to a good solid state solution at the high end of HDD’s historic markets. Clumsy early Flash SSD products and hybrid HDD/Flash products are confusing to customers today. Our new generation of products will change those perceptions and help customers understand the limits of mechanical HDD technology and advantages of Pliant EFD.
We think that word within the industry and recognition of the reliability and cost effectiveness of the RIGHT enterprise Flash SSD - Pliant’s EFD - product will travel fast. Adoption of Pliant EFDs will rapidly displace disk drives in high-performance applications as low-end Flash has displaced HDD storage in small and portable consumer devices. The good old HDD we’ve come to know and love will do just fine in the middle for a while.
Posted by Mario Apicella on April 28, 2008 03:00 AM
March 21, 2008 | Comments: (0)
Now that loose SSDs (solid state drives) are available, you may be wondering how best to take advantage of the technology. Here's a breakdown of where retrofitting current machines with solid state could reap worthwhile rewards.
Imation, one of the first vendors to market loose SSDs, is currently offering two SSD lines. One, a 2.5-inch drive with either 16GB or 32GB capacity; the other, a 3.5-inch device that stores up to 64GB. Both offer SATA connectivity.
Of the two offerings, the 2.5-inch format finds a natural home in laptops, allowing you to exploit one of solid state's major advantages over spinning devices: shock resistance.
Travel, as you know, takes its toll on laptops; just the right bump could easily damage a delicate spinning drive -- especially while in operation. To wit, Seagate's Momentus drives can withstand a shock of 900G for 2 milliseconds when idle, but only 300G for 1 millisecond when the drive is spinning, according to the company. So, if a beverage cart bumps you in the elbow while you are working, you are at greater risk of drive damage. Retrofitting with an SSD will mean your laptop will be less susceptible to drive damage.
As for the 3.5-inch format, which Imation sent me, laptop retrofitting is out of the question. Instead, outfitting a server or workstation with one of these loose SSDs will allow you to tap solid state's second major advantage: speed.
According to Imation, its Pro 7000 can sustain 81,000 IOPS when doing sequential reads, and it can transfer data as fast as 120MBps -- which sounds impressive but must be carefully compared with what spinning drives can do, considering the higher price of SSD drives.
Which is considerable.
A quick search on the Web shows that the 64GB Imation Pro 7000 sells for about $2,000, whereas a 16GB model can be purchased for around $900. As more vendors start offering these devices, these prices will go down significantly, but until then, careful planning is essential, as purchasing the minimum capacity necessary can save a bundle.
Without actual testing, however, it is difficult to reach a conclusion regarding the price/performance ratio of these drives. I will be running performance tests in the near future, with the intention of comparing the results with those of conventional drives. One such drive, the Western Digital Raptor, offers a similar capacity (74GB) and a SATA interface, allowing it to be connected to the same controller. From previous reviews, I know that the Raptor, which spins at 10,000 RPM, is an exceptionally fast drive, and should offer a worthwhile point of comparison for the Pro 7000.
While I'm running those benchmarks, consider this: Disk-to-disk backups can cut backup windows significantly over tape. However, disk-to-disk does not solve the problem of moving a copy to an outside vault in a satisfactory fashion. Because removable drives are susceptible to damage in transport, you still have to use tape. Drop a drive, and your backup is history. By contrast, the Pro 7000, being solid state, opens up the possibility of safely using disk drives to move backups off-site. I dropped my test drive from six feet over a concrete floor, and it's still working without problems.
Try that with a spinning drive.
Posted by Mario Apicella on March 21, 2008 03:00 AM
| STORAGE SPRAWL PODCAST |
| Listen to the latest podcast: |
| MP3 • RSS • Get Podcast • Archive • Mobilize |
TOP STORIES
ADDITIONAL RESOURCES
- Migrating to Vista
- CX4: Leading-Edge Midrange Storage for Virtualized Environments
- Turning Information Into A Competitive Advantage
- Flexible, Scalable, Enterprise Storage for Virtual Infrastructures
- Four Steps to Disaster Recovery and Business Continuity Using iSCSI
- Virtual Servers Meet Virtual Storage
