Intel's launch of the 6Gb/s DC S3500 bolsters their existing SSD product stack with an MLC mainstream enterprise SSD. Featuring Intel's third-generation PC29AS21CA0 (an 8-Channel ASIC) in concert with 20nm MLC NAND provides the DC S3500 with up to 75,000/11,500 random read/write IOPS and 500/450 MB/s of sustained sequential read/write speed.
The mainstream SSD market is booming and many administrators are choosing to deploy mainstream SSDs to address performance challenges in read-centric and mixed workload environments. The high performance envelope of enterprise SSDs boosts system performance up front, and the long-term advantages of radically lower power consumption, space reduction, and cooling costs drive TCO down even further.
One of the most pressing needs in today's datacenters is optimization of the compute layer. Performance is often hamstrung by I/O bound systems, leading to servers that operate with much lower CPU utilization than possible. The explosion of virtualization and the resulting 'I/O Blender' of sporadic highly random workloads require storage solutions with consistently high performance and low latency. Boosting the efficiency of the storage subsystem feeds the CPU with enough data to operate efficiently, creating more production and density per server.
The Intel DC S3500 (codename Wolfsville) is designed to provide a more cost-friendly alternative for those in need of acceleration with enterprise-class features at a lower price point. The DC S3500 (DC for Datacenter and S for SATA) provides a mainstream alternative to the Intel DC S3700, yet retains many similar characteristics.
The DC S3500 aligns itself to replace the Intel 320 customer base, offering seven times the endurance and 85% improved random read performance over the 320 Series. The 320 arrived in early 2011 and a replacement for the 3Gb/s 320 series is overdue. Intel's goal is to replace the 320 with an SSD with similar characteristics to the DC S3700.
The primary differentiator between the S3500 and the S3700 is the NAND employed. The DC S3500 utilizes 20nm MLC NAND in contrast to the 25nm HET-MLC NAND onboard the DC S3700. This provides a lower endurance threshold that is more suitable to read-centric and mixed workloads. The DC S3500 features a maximum of 450 TB Written for an 800GB capacity drive, and we expect endurance to scale with the various capacity points. The DC S3700, in contrast, offers up to 14.6PB of endurance for the 800GB model.
The DC S3500, with the same controller as the DC S3700, also focuses on performance consistency. Performance variability can rob applications of performance. Individual 'hangs' and lags from outlying I/O can significantly affect application performance simply because applications are forced into waiting for the next I/O to complete. The DC S3500 does not have quite the robust write latency specifications of the DC S3700, focusing more on read latency performance. The DC S3500 guarantees a maximum read latency of 500ms 99.9% of the time during a 4k random workload, and 50/65uS typical read/write latency.
Many administrators have experimented with utilizing standard client SSDs in enterprise environments and have suffered the consequences of inconsistent performance. The DC S3500 will rival client-side offerings with a price of roughly $1.22 per GB at launch. Delivering performance consistency and reliable performance at this price range will provide Intel a solid entrant into the booming value/mainstream SSD market.
Providing solid performance with a host of enterprise features such as end-to-end data path protection, 256-bit AES encryption, ECC protected memory, power loss protection, and a five year limited warranty rounds out the impressive feature set of the DC S3500.
Intel DC S3500 Architecture
The DC S3500 employs many of the same approaches that revolutionized performance consistency for the Intel DC S3700. The consistent performance of the DC S3700 is due to the new Intel controller, the PC29AS21CA0 (an 8-Channel ASIC), and the firmware being designed for low latency optimization.
Intel switched from utilizing a compressed binary tree system to a fully uncompressed 1:1 mapping of the NAND flash. This eliminates the need for defragmentation of the mapping table and reduces associated I/O latency concerns. In order to access such a large indirection table quickly, Intel keeps this map located in the 1GB of ECC DDR3-1600 DRAM. The large tables necessitate more cache for the SSD, with varying capacities of DRAM on each model. The memory is ECC protected to ensure the validity of the cached table data.
The 6Gb/s PC29AS21CA0 controller provides sequential read and write speeds of 500/450 MB/s, respectively, for the 800GB model. The SSD also features 75,000 random read IOPS and 11,500 random write IOPS. The write performance scales with various capacity points. Power consumption is slated at up to 5W (typ) and an idle of 650mW.
Enhanced power protection comes in the form of two radial electrolytic capacitors (rated for 105C at 3.5V/47uF) that take a unique approach of placement into a cutout section of the PCB. These capacitors flush data in-transit to the NAND in the event of a host power-loss issue. The SSD features self-diagnostics of the capacitor, and upon failure of the capacitor, will automatically switch the SSD into write-through mode. Users can also monitor the capacitor via SMART data.
Intel has taken several steps to protect user data, with CRC (Cyclic Redundancy Checks), firmware and logical block address verification built into the firmware. CRC consists of a hash tag used to validate data and identify data corruption. This protects the data from its original issuance, through the various levels of internal cache (SRAM and DRAM), and down to the NAND. AES-256 bit encryption support rounds out the feature set.
The Intel DC S3500 comes in a wide range of capacities, with the 2.5 inch SSDs coming in 80, 120, 160, 240, 300, 480, 600 and 800GB capacities. The DC S3500 is also offered in the 1.8-inch form factor, with capacities of 80, 240, 400 and 800GB.
Intel provides the Solid-State Drive Toolbox for easy management of the SSD. The DC S3500 features an UBER rating of 1 sector per 10E17 bits read and a MTBF of 2 million hours. The SSD features a five year warranty.
The DC S3500 and DCS3700 are designed for different environments, and we include the workload breakdown below the specifications.
Intel DC S3500 Internals
The DC S3500 comes in a 7.5mm z-height metal alloy casing. The 2.5-inch 7.5mm form factor is well suited for slim applications, but for those who demand high density there are 1.8-inch models as well.
The case of the SSD has plastic spacers that keep the PCB snug in the case. The controller, NAND and DRAM is BGA mounted.
The PC29AS21CA0 controller is near the 6Gb/s SATA ports, flanked by two Micron 512MB DDR3-1600 ECC DRAM packages. These DRAM packages are used for caching the large 1:1 mapping scheme employed by Intel.
The Intel 29F32B08MCMF2 20nm MLC NAND is responsible for providing the lower price structure of the DC S3500. There are eight packages per side of the PCB.
The two capacitors occupy a cutout on the side of the PCB. This is a unique capacitor deployment - most applications feature capacitors mounted directly on the PCB. The 3.5V/47uF capacitors are rated for temperatures up to 105C and will flush data to the NAND in the event of a power loss issue.
The proprietary PC29AS21CA controller is a departure from Intel's previous enterprise-class SSDs, with an 8-channel architecture as opposed to the old 10-channel design.
Test System and Methodology
We utilize a new approach to HDD and SSD storage testing for our Enterprise Test Bench, designed specifically to target long-term performance with a high level of granularity.
Many testing methods record peak and average measurements during the test period. These average values give a basic understanding of performance, but fall short in providing the clearest view possible of I/O QoS (Quality of Service).
'Average' results do little to indicate the performance variability experienced during actual deployment. The degree of variability is especially pertinent, as many applications can hang or lag as they wait for I/O requests to complete. This testing methodology illustrates performance variability, and includes average measurements, during the measurement window.
While under load, all storage solutions deliver variable levels of performance. While this fluctuation is normal, the degree of variability is what separates enterprise storage solutions from typical client-side hardware. Providing ongoing measurements from our workloads with one-second reporting intervals illustrates product differentiation in relation to I/O QOS. Scatter charts give readers a basic understanding of I/O latency distribution without directly observing numerous graphs.
Consistent latency is the goal of every storage solution, and measurements such as Maximum Latency only illuminate the single longest I/O received during testing. This can be misleading, as a single 'outlying I/O' can skew the view of an otherwise superb solution. Standard Deviation measurements consider latency distribution, but do not always effectively illustrate I/O distribution with enough granularity to provide a clear picture of system performance. We use histograms to illuminate the latency of every single I/O issued during our test runs.
Our testing regimen follows SNIA principles to ensure consistent, repeatable testing. We attain steady state through a process that brings the device within a performance level that does not range more than 20% during the measurement window. Forcing the device to perform a read-write-modify procedure for new I/O triggers all garbage collection and housekeeping algorithms, highlighting the real performance of the solution.
We measure power consumption during precondition runs. This provides measurements in time-based fashion, with results every second, to illuminate the behavior of power consumption in steady state conditions. Power consumption can cost more over the life of the device than the initial acquisition price of the hardware itself. This significantly affects the TCO of the storage solution. We also present IOPS-to-Watts measurements to highlight the efficiency of the storage solution.
Our test pool features SSDs of varying capacity and it is important to bear this in mind when viewing results. We placed the Intel DC S3500 into the value category due to its current price point of $1.22 per GB (this is subject to market fluctuation). All SSDs in this category retail from $1.00 to $1.50 per GB. The first page of results will provide the 'key' to understanding and interpreting our new test methodology.
4K Random Read/Write
We precondition Intel DC S3500 for 18,000 seconds, or five hours, receiving reports on several parameters of workload performance every second. We then plot this data to illustrate the drives' descent into steady state.
This chart consists of 36,000 data points. This is a dual-axis chart with the IOPS on the left and the latency on the right. The dark blue dots signify IOPS during the test, and the light blue dots are latency measurements during the test period. We place latency data in a logarithmic scale to bring it into comparison range. The lines through the data scatter are the average during the test. This type of testing presents standard deviation and maximum/minimum I/O in a visual manner.
Note that the IOPS and Latency figures are nearly mirror images of each other. This illustrates the point that high-granularity testing can give our readers a good feel for the latency distribution by viewing IOPS at one-second intervals. This should be in mind when viewing our test results below.
We provide histograms for further latency granularity below. This downward slope of performance happens very few times in the lifetime of the device, typically during the first few hours of use, and we present the precondition results only to confirm steady state convergence.
Each QD for every parameter tested includes 300 data points (five minutes of one second reports) to illustrate the degree of performance variability. The line for each QD represents the average speed reported during the five-minute interval.
4K random speed measurements are an important metric when comparing drive performance, as the hardest type of file access for any storage solution to master is small-file random. One of the most sought-after performance specifications, 4K random performance is a heavily marketed figure.
The Intel DC S3500 averages 62,817 IOPS with a 4K random workload at QD256. This places it between the Seagate Pro 600 and the Samsung SM843.
Garbage collection routines are more pronounced in heavy write workloads. This leads to more variability in performance. The Intel DC S3500 exhibits a very tight performance range at the bottom of the chart. The DC S3500 averages roughly the same 4K random write speed as the Samsung SM843 at 11,705 IOPS at QD256.
Both the CloudSpeed 500 and the 600 Pro exhibit significant performance variability, but in the entry-level tier of SSD's aimed at read-centric workloads higher performance is still a desired result.
Our write percentage testing illustrates the varying performance of each solution with mixed workloads. The 100% column to the right is a pure write workload of the 4K file size, and 0% represents a pure 4K read workload.
The Intel DC S3500 provides a very tightly defined range as we mix in more write activity, and while the Seagate 600 Pro and the CloudSpeed 500 exhibit more variability, they provide significantly better speed in heavier write scenarios.
The Intel DC S3500 delivers a very tight latency range in line with its consistent performance. The DC S3500 provides 2,006,218 I/O's (57%) at 20-40ms, and 1,479,527 I/O's (42.3%) in the 10-20ms range.
We record the power consumption measurements during our precondition run. We calculate the stated average results during the last five minutes of the test, after the device has settled into steady state.
The Intel DC S3500 averages 5.08 Watts during the 4K random preconditioning.
IOPS to Watts measurements are generated from data recorded during our precondition run, and the stated average is from the last five minutes of the test.
The DC S3500 offers the lowest Watt per IOP measurement with 4k write activity at 3,065 IOPS. It is important to note that the results are with heavy write activity and SSDs with faster write speeds benefit tremendously. Our mixed read/write workloads will give us a clearer view of the DC S3500 IOPS per Watt in its intended workloads.
It is important to note that the DC S3500 consumed 2.96 Watts in a 4k random read workload. This gives it an IOPS per Watt average of 21,293 IOPS per Watt in a 4K read workload, but among the other competition, this is average.
8K Random Read/Write
8K random read and write speed is a metric that is not tested for consumer use, but for enterprise environments this is an important aspect of performance. With several different workloads relying heavily upon 8K performance, we include this as a standard with each evaluation. Many of our Server Emulations below will also test 8K performance with various mixed read/write workloads.
The average 8K random read speed of the Intel DC S3500 comes in at 39,413 IOPS at QD256, placing it between the Seagate 600 Pro and the Samsung SM843.
The average 8K random write speed of the DC S3500 is 1,626 IOPS at QD256. The Intel again matches the SM843 on random write performance, but also provides a very tight performance range.
The Intel DC S3500 falls quickly into the lower performance range during the test period.
The Intel DC S3500 again exhibits a very tight latency range with 50.1% of I/O's falling into the 40-60ms range, and 48.5% falling into the 20-40ms range.
Power consumption for the Intel DC S3500 averages 3.87 Watts during heavy write activity.
The DC S3700 averages 1,626 IOPS per Watt during the preconditioning period. It is important to note that the results are with heavy write activity and SSDs with faster write speeds benefit tremendously. Our mixed read/write workloads will give us a clearer view of the IOPS per Watt performance in the DC S3500's intended workloads.
128K Sequential Read/Write
The 128K sequential speeds reflect the maximum sequential throughput of the SSD using a realistic file size encountered in an enterprise scenario.
The Intel DC S3500 comes in at an average of 439 MB/s in sequential read speed, falling below the rest of the entrants.
The DC S3500 features the lowest performance in our 128K sequential write testing with an average of 419 MB/s. We also note more variability with this workload than the other SSDs in the test pool.
The Intel DC S3500 continues its pattern of variability in our write percentage testing, but manages to keep up with the closely matched SM843.
The Intel DC S3500 displays a bit of the write variability in our latency results with 71.4% of I/O's at 60-80ms, and 28.5% of I/O's in the 80-100ms range.
The DC S3500 averages 4.72 Watts during sequential write testing.
The Intel DC S3500 provides 88 MB/s per Watt, the lowest average of our test pool.
Database/OLTP and Webserver
This test emulates Database and On-Line Transaction Processing (OLTP) workloads. OLTP is in essence the processing of transactions such as credit cards and high frequency trading in the financial sector. Enterprise SSDs are uniquely well suited for the financial sector with their low latency and high random workload performance. Databases are the bread and butter of many enterprise deployments. These are demanding 8K random workloads with a 66% read and 33% write distribution that can bring even the highest performing solutions down to earth.
The Intel DC S3500 averages 16,931 IOPS during the measurement window.
The DC S3500 averages 92.9% of I/O's at 10-20ms, and 5.7% in the 20-40ms range.
The Intel DC S3500 averages 3.98 Watts during the precondition run.
The Intel DC S3500 averages 4,248 IOPS per Watt during this test. The OLTP/Database mixed workload is a close representation of the intended environment for the DC S3500.
The Webserver profile is a read-only test with a wide range of file sizes. Web servers are responsible for generating content for users to view over the internet, much like the very page you are reading. The speed of the underlying storage system has a massive impact on the speed and responsiveness of the server that is hosting the website, and thus the end-user experience.
The Intel DC S3500 averages 23,765 IOPS at QD256 in this read-centric workload.
The DC S3500 provides 61.9% of I/O's at 10-20ms, 31.7% at 8-10ms, and 6.2% at 6-8ms.
The DC S3500 averages 2.5 Watts during the measurement window, much lower than the power draw during the other workloads with write activity.
The DC S3500 averages 9,464 IOPS per Watt.
The File Server profile represents typical file server workloads. This profile tests a wide variety of different file sizes simultaneously, with an 80% read and 20% write distribution.
The Intel DC S3500 averages 31,452 IOPS at QD256, falling into second place.
The DC S3700 provides 55% of I/O at 8-10ms, and 43.5% in the 6-8ms range.
The DC S3500 averages 2.32 Watts in steady state.
The DC S3500 averages 13,529 IOPS per Watt, placing third in the test.
The Intel DC S3500 brings enterprise-class features for the price conscious, rounding out Intel's product stack with a current-generation value SSD contender. One of the immediate effects of the DC S3500 will be increased competition in the mainstream and value SSD market. Much like the introduction of the Intel 910 helped to lower prices of all PCIe Workload Accelerators, the DC S3500 will pressure other contenders in the market with its low price point.
The Intel DC S3500 will retail for roughly $1.22 per GB initially, and price will be subject to fluctuation. This places the SSD into our value category of SSDs that range from $1.00 to $1.50 per GB. Exact pricing for enterprise SSDs is hard to come by due to fluctuations and volume pricing. However, this range provides us a good comparison pool. It is important to remember that a difference of even 25 cents per GB can add up quickly when purchasing a large number of high capacity SSDs.
We noted a definite separation in our test pool with the slightly higher-cost Seagate 600 Pro and SMART CloudSpeed 500. These SSDs feature much more robust write performance, lending them an advantage in many of our tests. The Samsung SM843 and the Intel DC S3500 were closely matched, though the SM843 tends to sport higher read speeds.
One weakness of the DC S3500 is the sequential read/write performance. The sequential read performance of the DC S3500 was lower than other entrants. The sequential write performance experienced significant variability and an average speed lower than the other SSDs in the test pool. In random read testing the DC S3500 fared well, but once we mix in write workloads, performance fell quickly. The random write speeds were within expectations for the SSD, reaching 11,000 IOPS in a 4k random write workload.
We experienced slightly better performance from the DC S3500 in our mixed workloads, though it still did not reach chart-topping status. The power consumption for heavy write workloads is higher than other SSDs, and in mixed workloads this only improved slightly. From an IOPS to Watts standpoint, there are more efficient SSDs on the market, but not many with the same blend of pricing and features.
As the chart above illustrates, endurance should always be a driving factor behind any SSD purchasing decision. Small differences in price can equate to large differences in performance and endurance, and for those that wade into the value market, it is important to take a number of factors into consideration. Features such as end-to-end data protection, host power loss protection, and a five year warranty separate this class of SSD from client-side hardware. One area that the DC S3500 does excel in is predictable performance consistency. This is in stark contrast to client-side hardware that experiences tremendous variability and poor latency performance.
The Intel DC S3500 performed within expectations during our testing with the exception of higher power consumption than other competing SSDs. In comparison to any HDD, the DC S3500 will offer a tremendous power saving advantage.
The five year warranty provides customers with peace of mind that the DC S3500 will live to its endurance specification. The low price point, weighing in at 50% cheaper than the DC S3700 and 20% cheaper than the Intel 320, will also draw in many customers. For those looking to replace HDDs with a value alternative, the Intel DC S3500 fits the bill.