|Introduction and Specifications|
When we sit back and look at the evolving storage market and SSD technologies, there are a couple of competitive camps that are trying to gain traction, in what is without question one of the largest growth areas in storage. The major players in solid state storage solutions have emerged from old-line, incumbents in hard disk media and then also a new breed, specializing exclusively in SSD technology. Major memory manufacturers like Intel, Micron and Samsung, along with specialty shops like OCZ and Corsair, are making a push with their expertise in memory, whereas players like Western Digital and Seagate also have skin in the game, but are walking the line carefully, as the migration from spinning media to solid state slowly shapes the landscape.
LSI is the sort of company that has been a household name in storage for many years; highly respected and trusted in the HBA (Host Bus Adapter) space, especially in high availability datacenter and media server applications. If we consider companies like LSI and their competitors, like Adaptec and AMCC (3Ware), you'll note that they're in a unique position in this storage technology migration, in that they can almost take an agnostic stance, in terms of which media type and market opportunity they decide to resource development efforts. For players like LSI, with a wealth of HBA controller IP at their disposal, they can relatively easily adapt and re-tool their technologies to service NAND Flash SSD memory structures and bring compelling solutions to market as market demand develops.
The LSI WarpDrive that we have on hand for you here is representative of LSI's efforts to capitalize on the company's strong SAS RAID controller technology base, combined with the most cutting-edge SSD technologies on the market today. We'll call to your attention early here that the WarpDrive means business though. With a $11,500 MSRP and street pricing in the $7400 - $8K range (yes, we're not sure what's up with that), end-user enthusiasts are likely not going to belly-up the justification to support the cost of a storage subsystem that is much more expensive than a typical entire system build. The LSI WarpDrive is targeted to the datacenter, workstation and server markets, where gobs of bandwidth, instantaneous random access and high reliability mean money in the bank, and as a result, "total cost of ownership" becomes increasingly more practical.
Alas, we'll squelch the marketing speak, before your eyes begin to glaze over, and get down to the details. After all, we've got some of the fastest PCI Express-based SSD technology money can buy on the test bench and we intend to light it up for you to see on the pages that follow...
The first thing to notice in the WarpDrive spec sheet is that we're looking at an SLC NAND-based SSD. SLC is decidedly more expensive than MLC NAND Flash, which is what's currently powering the large majority of consumer-class SSD products on the market today. However, it also has a write endurance factor some 10X that of MLC NAND. In other words, SLC is much better suited for high availability and reliability applications like those found in a web server, data center or pro workstation environments.
Other spec notables are LSI's proprietary RAID controller, that LSI has built a custom firmware wrapper for, that affords it the ability to interface with multiple SSD volumes configured in a RAID 0 stripe. Though you have access to the RAID BIOS on the WarpDrive, there's not much to configure there, save for boot options and some reporting. The drive is pretty much hard configured to a single RAID 0 volume of 300GB for the fastest performance it's capable of delivering. Let's take a closer look at the hardware...
|The WarpDrive Up Close and Dissected|
|As we noted earlier, the LSI WarpDrive is a combination of LSI's tried and true SAS (Serial Attached SCSI) RAID controller technology, along with SandForce-based SSDs in a RAID 0 configuration. Specifically, there are six SSD cards that make up the WarpDrive in its 300GB configuration, all sporting SandForce 1500 series controllers and all taking to an LSI SAS2008 hardware RAID engine, and a very capable engine at that.
There's also a bit of non-volatile SRAM that affords the WarpDrive more resiliency and the ability to complete fast write transactions in the event of power loss or system reset. The Flash firmware block is a chunk of configuration memory that the SandForce controllers need to function. And of course LSI has their own RAID BIOS firmware for the WarpDrive, which, as we noted is basically hard configured, save for a few status and boot options; there is no option to change the RAID configuration of the WarpDrive as it stands today. It's setup for maximum performance in RAID 0.
LSI WarpDrive - Half-Height PCI Express Card - Click for high res
SSD Flash Modules - Click for high res
The design of the WarpDrive is rather elegant, in light of the various different technologies at play with the device. Similar to OCZ's RevoDrive X2 series, there are mezzanine style Flash cards plugged into a half-height PCI Express X8 HBA card design. Each module has eight Micron NAND Flash chips on board for a total of 64GB per card. Together, they offer about 384GB of total capacity, but LSI reserves excess memory space for SSD garbage collection algorithms, wear-leveling and maintenance to keep the drive performing well over time and use. Frankly these "non-native" PCI Express designs aren't quite as elegant as Fusion-io's direct attach approach that doesn't have to talk through RAID controllers and SSD controllers in tandem. However, Fusion-io's custom PCI Express SSD ASIC adds a fair bit of cost to their solution as well.
|Test System SANDRA Physical Disk Tests|
Our Test Methodologies: Under each test condition, the Solid State Drives tested here were installed as secondary volumes in our testbed, with a standard spinning hard disk for the OS and benchmark installations. The SSDs were left blank without partitions wherever possible, unless a test required them to be partitioned and formatted, as was the case with our ATTO, Vantage, and CrystalDiskMark benchmark tests. Windows firewall, automatic updates and screen savers were all disabled before testing. In all test runs, we rebooted the system and waited several minutes for drive activity to settle before invoking a test.
For our first set of tests, we used SiSoft SANDRA, the the System ANalyzer, Diagnostic and Reporting Assistant. Here, we used the Physical Disk test suite and provided the results from our comparison SSDs. The benchmarks were run without formatting and read and write performance metrics are detailed below.
SANDRA's purely sequential and rather short read/write test doesn't paint the WarpDrive in an all that impressive light. Its performance is decent, perhaps two times that of the average SSD but only marginally faster than OCZ's first gen RevoDrive. Write speeds for the WarpDrive are definitely its strong suit, according to SANDRA. However, as you'll see many of the tests moving forward here, this quick and dirty view of the WarpDrive's performance is hardly representative of what the drive is capable of.
|IOMeter 2008 Test Results|
As we noted in a previous SSD round-up, though IOMeter is clearly a well-respected industry standard drive benchmark, we're not completely comfortable with it for testing SSDs. The fact of the matter is, though our actual results with IOMeter appear to scale properly, it is debatable whether or not certain access patterns, as they are presented to and measured on an SSD, actually provide a valid example of real-world performance for the average end user. That said, we do think IOMeter is a gauge for relative available bandwidth with a given storage solution. In addition there are certain higher-end workloads you can create place on a drive with IOMeter, that you really can't with any other benchmark tool available currently.
In the following tables, we're showing two sets of access patterns; our Workstation pattern, with an 8K transfer size, 80% reads (20% writes) and 80% random (20% sequential) access and our Database access pattern of 4K transfers, 67% reads (34% writes) and 100% random access.
The first thing you'll note here is how flat the standard SATA SSD's performance was across test patterns and IO queue depth. The IO queue depth set in IOMeter essentially represents higher levels of workload requests of the same access patterns simultaneously. The Vertex LE SSD was saturated here as we turned up queue depth. However, the entire group of PCI Express-based SSDs scaled up significantly at higher request levels. OCZ's RevoDrive line-up flattens out at a queue of 144 or so, with the RevoDrive X2 offering the best performance out of the offering. The Fusion-io drives both offer significantly more IO bandwidth and also flatten out at higher queue depths.
The LSI WarpDrive offers light workload performance at a queue depth of 12 to be somewhere in the middle of the pack, faster than the more consumer/workstation ioXtreme card but about half the performance of the expensive enterprise-class Fusion-io ioDrive. Like the WarpDrive, the ioDrive is also an SLC-based SSD. Finally, however, when we scale IO requests higher, the WarpDrive really begins to kick into high gear offering performance well in excess of 100K IOPs in our less than optimized Workstation test condition. The WarpDrive actually proved to be the fastest of the group under higher queue depths and it's impressive to see it overtake Fusion-io's best offering.
Our database access pattern showed much of the same performance grouping as we saw in the Workstation setup. In fact, with its higher mix of random write requests, all PCI Express solutions here scale to even higher IO throughput levels. The LSI WarpDrive broke 140K IOPs, which is impressive and the fastest score we've recorded under these test conditions, and by a long shot if you consider anything else other than Fusion-io's expensive $8K, SLC-based, 160GB ioDrive. Again, however, to realize the full potential of the WarpDrive, you have to exercise the product with lots of concurrent requests. Average client workload performance is going to look at lot more like the numbers at an IO queue depth of 12 here. We should also note that CPU utilization across all test runs with the WarpDrive, hovered around the 8 - 10% mark, which is not as good as we expected for a hardware RAID engine, but still reasonable when you consider the throughput of the product.
|ATTO Disk Benchmark|
ATTO is another "quick and dirty" type of disk benchmark that measures transfer speeds across a specific volume length. It measures raw transfer rates for both reads and writes and graphs them out in an easily interpreted chart. We chose .5kb through 8192kb transfer sizes and a queue depth of 6 over a total max volume length of 256MB. ATTO's workloads are sequential in nature and measure raw bandwidth, rather than IO response time, access latency etc. This test was performed on blank, formatted drives with default NTFS partitions in Windows 7 x64.
In ATTO testing, the LSI WarpDrive drive showed some of the highest overall read and write bandwidth of the five SSD solutions we compared here but only at certain transfer sizes. Furthermore, as you can see, there is a definite issue with transfer sizes of 128K and 256K. This is a known bug that we confirmed with LSI and they've noted a firmware update is coming that should resolve the issue, as well as bring the large 1MB - 8MB transfer speeds seen here, more in line with the drive's performance at the 64K mark.
CrystalDiskMark is a synthetic benchmark that tests both sequential as well as random small and large file transfers. It does a nice job of providing a quick look at best and worst case scenarios with regard to SSD performance, best case being larger sequential transfers and worse case being small, random transfers.
CrystalDiskMark probably most accurately illustrates the high level performance picture of the WarpDrive. When you look at large sequential transfers, the WarpDrive shows very strong write speeds but middle of the road read throughput. However, with smaller 512K random requests, both read and write performance scale up nicely and the product easily keeps pace with Fusion-io's fastest SLC-based SSD. What's really impressive is the high queue depth throughput at the 4K QD32 mark. The WarpDrive actually smokes the Fusion-io ioDrive and offers the fastest throughput of any device we've tested here, by a significant margin. The moral of the story? The LSI WarpDrive is very well suited to data center server environments where high availability and low IO response times are critical; which is fitting of course, when you consider LSI's pedigree in the space.
|File Transfer Tests - General Performance|
Our next series of tests are what you might call more "crude measurements" in that we simply fired up our trusty stop-watch and measured the time it took to complete a copy and paste command of a single large file or a bunch of large files from one storage volume in our test system to another.
** Please note that we utilized a Fusion-io ioXtreme card as our source drive in the following tests, to read files from or copy files to, for our read and write measurements. This affords us the luxury of much higher available bandwidth from the source drive or to a target drive, such that either direction would not be as much of a limiting factor in any given test condition.
Here times are tightly grouped at the top of the stack, with the LSI WarpDrive offering competitive performance just behind that of the Fusion-io card for writes but edging it out slightly in reads.
Again we see that a single large sequential read operation is not where the WarpDrive excels, but then again the field is so tight here there really aren't any standouts. However, when it comes to large sequential writes in this test, the WarpDrive once again goes toe to toe with the fastest SLC Fusion-io card we have on hand.
|PCMark Vantage HDD|
|We really like PCMark Vantage's HDD Performance module for its real-world application measurement approach to testing. PCMark Vantage offers a trace-based measurement of system response times under various scripted workloads of tradtional client/desktop system operation. From simple Windows start-up performance to data streaming from a disk drive in a game engine and video editing with Windows Movie Maker, we feel more comfortable that these tests reasonably illustrate the performance profile of SSDs in an end-user/consumer PC usage model.
This series of Vantage tests will stress read performance in real-world usage models, with a broad mix of sequential and random read transactions of both small and large file sizes.
The LSI WarpDrive once again shows best-in-class performance in this set of Vantage HDD that are mostly read intensive workloads. The WarpDrive handily out-paces the OCZ RevoDrive X2 and even occasionally bests the Fusion-io ioDrive. The ioDrive incidentally put up the fastest Vista Startup score by a wide margin, but then again, you can't boot from the device either, nor can you from the ioXtreme. The WarpDrive put up the fastest Vista Startup score of all the bootable SSD solutions we tested. However, we'd note that when we actually installed a Windows 7 X64 image on the WarpDrive, it took about 24 seconds to get to a desktop, whereas the RevoDrive X2 was able to shave about 10 seconds off that market getting booting Win 7 in only 13.5 seconds. Clearly workstation users aren't going to be as concerned with OS boot times, as much as they are with something like video or 3D rendering, but it should be noted that LSI has a bit of work to do in this area. It seems as if the card isn't hitting its full potential in this one test simple test condition we placed on it. Finally, we should also note that there is currently a small bug, in LSI's Windows 7 drivers at least, that didn't allow the card to be detected during OS installation as a boot drive if AHCI was enabled in the motherboard BIOS. Once we disabled AHCI, we were able to load drivers and target the WarpDrive as our boot volume.
|PCMark Vantage (Continued)|
|The following PCMark Vantage HDD tests are more write intensive and in some cases stress the Achilles' Heel of the average storage subsystem, that being random write performance.
The numbers here speak pretty well for themselves so we won't wax poetic much. The WarpDrive's write prowess shows strong here and the LSI SSD puts up scores besting any of the group, with the once exception of the application loading test. Windows Movie Maker showed a sizable gain over the io-Drive and about a 14% gain over the next fastest score recorded on the OCZ RevoDrive X2.
|Performance Summary and Conclusion|
The LSI WarpDrive offered very strong performance, keeping close pace and occasionally outpacing the performance offered by Fusion-io's ioDrive line of SLC-based products. Its general weakness was in large sequential reads whereas it's strengths are more than obvious in high load random read/write requests and with respect to random writes in general. The WarpDrive offered some of the highest IOMeter response throughput scores we've measured to date in our workstation and database tests, and continued to prove-out the scores we took in synthetic testing by posting some of the best PCMark Vantage scores we've seen to date as well, again with benchmark records broken in write-intensive tests. There were some oddities observed, however, specifically in larger sequential transfers (ATTO testing).
Regardless, We've seen many PCI Express SSD products in our labs, that make use of standard hardware RAID controllers in conjunction with off-the-shelf SSD solutions based on third party controllers like SandForce. We've seen mixed results from products like this, that ranged from yawn-inspiring to jaw-dropping exciting. The LSI WarpDrive is an example of the latter and it comes tuned with a legacy of high throughput server performance backing it up with LSI's excellent SAS controller technology behind it. Make no mistake, at a current street price of $7400 - $8900 or so for 300GB of capacity, we're talking roughly $25 per Gigabyte, so there is no way the end user enthusiast crowd is going to justify the cost of this beast. However, when you consider its closest competitor performance and reliability-wise (remember, we're considering SLC-based SSD reliability here), that being the 160GB Fusion-io at around $8,000 and about a $50/GB cost model currently, the WarpDrive's price tag, relatively speaking is, dare we say, very competitively priced?
LSI actually has a TCO (Total Cost of Ownership) calculator that you can play with. Essentially it illustrates how many hard drives you would need to be able to scale the IO response rate of the LSI WarpDrive. The metric used here is page requests per second, in a web server environment. At the lowest workload of 1000 pages per second, LSI claims you'll need $17000 worth of hard drives to match the WarpDrives performance. Looking at the numbers quickly, we're assuming they're using 73GB 15K RPM drives (standard fare in web server setups) to come to this calculation. In short, if you need this kind of throughput in your datacenter or in your workstation where crunching large volumes of video media means time and money, then the WarpDrive offers some of the fastest SSD technology money can buy currently.
Short of that, the practicality of the device for most client applications if obviously not there, but we're hoping that it's obvious to you that this type of technology isn't something you're going to need for a round of Call of Duty with your buddies on a Saturday night. Clearly the product is intended for more mission critical applications; not that packing a clip and slugging it out with a virtual Fidel Castro isn't a noble cause of course. Regardless, we're hopeful that LSI will work to improve upon sequential read performance and work out some of the anomalies we saw with the WarpDrive. It's very early on in the product's introduction so there's likely a bit more performance they can wring out of this already impressive PCI Express SSD. In the mean time, dropping one of these bad boys into a high performance database server will definitely skyrocket your response times and total bandwidth, you can bet on that. And with LSI's proven track record in storage, reliability won't need to suffer as a result.