Samsung has many firsts under their belt, with the world's first 3D NAND and 3bit MLC (TLC) products leading the way. Pairing innovative hardware with a new interface technology capitalizes on the performance advantages of flash storage. Samsung also led the NVMe race with the XS1715, the industry's first certified product (5/31/2013) on the UNH-IOL NVMe Test Consortiums Integrators List (IL).
NVMe standardizes interaction with non-volatile memory on the PCIe bus through an optimized register interface, command set, and feature set. We dive in deeper into the details in our Defining NVMe article. Intel brought the first NVMe products to the retail market with the Intel DC P3700, but the XS1715 has been in production behind the OEM curtain for a longer time.
The Samsung XS1715 is a 2.5" NVMe PCIe SSD that comes in 400, 800, and 1.6TB capacities. Performance density is the most impressive aspect of the XS1715, it can deliver 750,000 random read IOPS and 115,000 write IOPS in an amazingly small 2.5" 15mm package. It also offers a beefy 3,000 MB/s in sequential read speed and 1,400 MB/s of write performance. Samsung recently announced mass production of the SM1715 series, an NVMe SSD in the traditional HHHL form factor, and capacities up to 3.2TB courtesy of 3D V-NAND. The V-NAND powered SM1715 delivers a 10DWPD threshold, which is higher than the 5.6 DWPD (1.6TB model) from the XS1715's 20nm-class eMLC Toggle NAND.
Dell currently sells the XS1715 in their popular PowerEdge R920 series as the PowerEdge Express Flash NVMe SSD. The XS1715 connects to a backplane and provides Dual-Port PCIe 3.0 x4 speed via the new SFF-8639 connector. Front bay access increases serviceability and brings hot-swapping capability to the PCIe interface. It also increases the potential number of NVMe drives deployed per server. Dell utilizes PCIe extender adaptors that allow four NVMe devices per motherboard PCIe slot. This maximizes performance and capacity density. Two of these extenders allow utilization of eight NVMe SSDs, and still leaves PCIe slots open for other hardware, such as networking and co-processors.
The extreme performance of the XS1715 is well suited for a number of high performance write-intensive applications, such as HPC, ERP, databases, transactional workloads, OLAP, and virtualized environments. The XS1715 is also bootable with supported UEFI platforms. The performance advantages are clearly evident in the four world records held by the Dell PowerEdge R920.
The 5.6 DWPD for the 800GB and 1.6TB models (7 DWPD for 400GB) is competitive with other similar-capacity PCIe offerings. The warranty period is five years and backed by the 5-year DWPD guarantee, or individual OEM warranty limitations. Dell guarantees the SSD according to the TBW (Total Bytes Written) threshold and offers software-monitoring utilities for easy tracking of remaining endurance. The XS1715 offers full data path protection, an UBER rating of 1 per 10^17, and power loss protection via tantalum capacitors. The XS1715 also features a robust two million hour MTBF.
The PMC 89HF16P04CG3 controller powers the XS1715. This 16-channel PCIe 3.0 x4 programmable flash controller provides Samsung with wide latitude in firmware customizations. The PMC controller interfaces with a wide variety of SLC and MLC NAND with the Toggle or ONFI interface, and it would not be surprising to see V-NAND compatibility with this controller in the future.
NVMe provides better latency and performance in comparison to legacy interfaces used by most SSDs, such as AHCI and SCSI. NVMe also provides a standardized approach that can leverage a common driver, instead of the varying proprietary drivers found in current-gen PCIe SSDs. NVMe's higher performance, and multiple queues, provides enhanced scaling in demanding environments. NVMe is well suited to realize the advantages of SSDs by keeping them as close to the CPU as possible.
The XS1715 provides us with a first look at the new SFF-8639 connector. Let's take a closer look at the Samsung XS1715 and its connection into the server.
Internals and Specifications
Samsung XS1715 1.6TB Internals and Specifications
The Samsung XS1715 comes in the 2.5" form factor with a 15mm z-height. The XS1715 interfaces with the PCIe bus via the SFF-8639 connector, covered in more detail on the following page.
We are greeted with a dual PCB design upon opening the case, and a thick thermal pad that helps cool the banks of capacitors.
40 capacitors and three NAND packages line the top PCB. We also note the empty pads for mounting more capacitors, which might indicate higher capacity versions with this same design in the future. The dual-PCB configuration is held tightly in place by a fastener on each corner of the PCB's. The sandwiched design and eight internal fasteners are nestled into a precision-milled case to ensure the SSD will function well in high-vibration environments.
The other side of the dual-PCB configuration also has large thermal pads to wick heat away from the PMC controller, DRAM, and 20nm-class eMLC NAND packages. There is an additional thermal pad between the two layers.
The large mating connectors on each side snap tightly into each other. We also observe the 2GB of DRAM cache next to the PMC controller.
The PMC 89HF16P04CG3 NVMEC16 controller provides Samsung with a programmable and flexible 16-channel controller. Each independent 8-bit channel can support up to eight targets. SECDED ECC protects data in the internal memory and DDR3 SDRAM, and internal end-to-end host to NAND data path protection eliminates silent data corruption. The controller offers robust ECC and flash channel RAID. Auto RAID rebuild allows transparent recovery from flash device failures, while maintaining normal operation. According to Dell's product documentation, the XS1715 will go into a recovery mode upon reboot, after an unexpected power loss. This is a common rebuild technique utilized with many PCIe SSDs.
The controller is also optimized for low-power operation through a combination of architectural and semiconductor design techniques. Advanced power-management features power-down units when not in use and processor cores go into a low power state when idle. The controller also supports the Enterprise NVMe device power-management interface, for power and performance management.
The XS1715 is available in MZWEI400HAGM-00003, MZWEI800HAGM-00003, and MZWEI1T6HAGP-00003 SKUs.
SFF-8639 and ExpressBay
Page 3[SFF-8639 and ExpressBay]
The Samsung XS1715 came to our lab with a PCIe card and custom backplane connected via three cables. This development kit is not used in actual deployments.
The PCIe card is simple, with three connectors for the wires that connect to the backplane. The cable to the left is a typical SAS cable, the middle cable is used for power delivery (the XS1715 uses 3.3v and 12v), and the cable on the right is a typical SATA cable, which might be used for refclock signals.
The backplane has mating connectors for the three wires on the back, and an SFF-8639 connector on the front.
SFF-8639 is a new connector that serves as the Swiss army knife of storage connections. SFF-8639 is a multifunction 6x unshielded connector defined by the SSD Form Factor Workgroup (SSD FFWG). SFF-8639 supports single port SATA, dual port SATA Express, dual port SAS, MultiLink SAS, and up to 4-port PCIe device configurations. Backplanes with this connector are compatible with plugs developed in accordance with SFF-8482, SFF-8630, and SFF-8680s specifications. The connector supports all of the pins for dual-port SAS, and adds an additional 39 pins to support four PCIe lanes and 10 sideband signals. 14 of the additional 39 pins are used for two additional SAS ports in MultiLink SAS connections.
In normal deployments, ExpressBay extends PCIe speed to the front panel and provides an easy connection for drives with the SFF-8639 connector. ExpressBay is a multi-protocol externally accessible storage slot that supports 2.5" or 3.5" devices compatible with the SFF-8639 connector. Standard drive bays are limited to 9 Watts of power, but ExpressBay bumps that up to 25 Watts - the same as the typical ceiling of a PCIe slot on the motherboard. Additional airflow is also provided for dissipation of higher heat levels.
SFF-8639 also brings hot-swapping capability to the PCIe interface. A closer examination of the plug interface reveals a 3-level contact engagement sequence. The mating sequence of the contacts is timed to occur in the proper order. This is accomplished through a carefully designed offset between contacts on each side of the plug and receptacle.
Dell utilizes PCIe extender adaptors to connect the PowerEdge R920's drive bays to the PCIe slots on the motherboard. The direct PCIe connection yields a maximum bandwidth of 8 GB/s.
Test System and Methodology
We designed our approach to storage testing 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).
While under load, all storage solutions deliver variable levels of performance. 'Average' results do little to indicate 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. 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. This testing methodology illustrates performance variability, and includes average measurements, during the measurement window.
IOPS data that ignores latency is useless. 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 utilize high-granularity I/O latency charts to illuminate performance during our test runs.
Our testing regimen follows SNIA principles to ensure consistent, repeatable testing, and utilizes multi-threaded workloads found in typical production environments. We tested the 1.6TB Samsung XS1715 against the 1.6TB Intel DC P3700, the 1.4TB Micron P420m and the 2.2TB HGST FlashMAX II
The first page of results will provide the 'key' to understanding and interpreting our test methodology.
Benchmarks - 4k Random Read/Write
4k Random Read/Write
We precondition the 1.6TB Samsung XS1715 for 15,000 seconds, or four hours, receiving performance reports every second. We plot this data to illustrate the drives' descent into steady state.
This dual-axis chart consists of 30,000 data points, with the IOPS on the left and the latency on the right. The red dots signify IOPS, and the grey dots are latency measurements during the test. 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 that high-granularity testing gives our readers a good feel for latency distribution by viewing IOPS at one-second intervals. This should be in mind when viewing our test results below. This downward slope of performance only happens during the first few hours of use, and we present precondition results only to confirm steady state convergence.
Each level tested includes 300 data points (five minutes of one second reports) to illustrate performance variability. The line for each OIO depth 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 Samsung XS1715 jumps to an early lead with a scorching average of 752,476 IOPS at 256 OIO (Outstanding I/O), the Intel DC P3700 averages 467,055 IOPS, the HGST FlashMAX II averages 350,532 IOPS, and the Micron P420m averages 724,958 IOPS. The XS1715 also leads the majority of the lower loadings, establishing a lead early and expanding through the remainder of the measurement window. The competition is not close in the 64 and 128 OIO columns.
Our Latency vs IOPS charts compare the amount of performance attained from each solution at specific latency measurements. Many applications have specific latency requirements. These charts present relevant metrics in an easy to read manner for readers who are familiar with their applications requirements. The arrays that are lowest and furthest to the right exhibit the most desirable latency characteristics.
The XS1715 dominates this test with a beastly average of 660,000 IOPS at .2ms, and the nearest competitor is 462,000 IOPS from the P3700.
The Samsung XS1715 averages 116,692 IOPS at 256 OIO, Intel DC P3700 averages 147,846 IOPS, the FlashMAX II averages 119,662 IOPS, and the P420m averages 98,304 IOPS. The XS1715 places third, but has a clearly defined performance envelope rivaled only by the FlashMAX II.
The Intel DC P3700 and the Micron P420m outpaced the XS1715 in latency-v-IOPS measurements with a 4k write workload.
Our write percentage testing illustrates the varying performance of each solution with mixed workloads. The 100% column to the right is a pure 4k write workload, and 0% represents a pure 4k read workload. Mixed I/O is a constant reality In VDI and other intensive applications, resulting in the I/O blender effect. The XS1715 has an obvious propensity for read-centric workloads, but also scores in line with the other competitors in heavier write workloads, with the exception of the Intel DC P3700, which takes the lead.
Latency measurements during the mixed workload testing reveal increasing latency as we head into heavier write mixtures. The XS1715 operates within a tight latency envelope, which is indicative of predictable performance, but latency falls behind the DC P3700 and FlashMAX II in mixed write workloads.
Benchmarks - 8k Random Read/Write
8k Random Read/Write
Many server workloads rely heavily upon 8k performance, and we include this as a standard with each evaluation. Many of our server workloads also test 8k performance with various mixed read/write distributions.
The Samsung XS1715 averages 384,690 IOPS at 256 OIO, the Intel averages 275,367 IOPS, the FlashMAX II averages 207,859 IOPS, and the Micron P420m takes a slight lead to top the chart at 387,785 IOPS. The XS1715 again displays impressive performance scaling in the middle ranges as well, dominating the 16-128 OIO segment of the test.
The Micron P420m took the performance crown under the heaviest workloads, but the XS1715 separates itself from the crowd in our Latency-v-IOPS chart.
The Samsung XS1715 averages 54,588 IOPS at 256 OIO, the Intel DC P3700 averages 66,082 IOPS, the P420m hits 79,633 IOPS, and the FlashMAX II tops out at a speed of 59,279 IOPS.
The Samsung XS1715 and Micron P420m are closely matched under heavy load, and the Intel DC P3700 and FlashMAX II take the lead.
The XS1715 exhibits the same affinity for read-centric workloads in this test, but also performs well in heavier write workloads. The Intel DC P3700 continues to lead in mixed random write workloads.
We note the same latency characteristics observed in the 4k mixed random workload testing.
Benchmarks - 128k Sequential Read/Write
128k Sequential Read/Write
128k sequential speed reflects the maximum sequential throughput of the SSD and is indicative of performance in OLAP, batch processing, streaming, content delivery applications, and backup scenarios.
The XS1715 displays solid sequential read speed of 3,006 MB/s at 256 OIO, the Intel DC P3700 averages 2,392 MB/s, the HGST FlashMAX II averages 2,664 MB/s, and the P420m leads with 3,186 MB/s.
The XS1715 delivers great performance vs latency in the sequential read testing.
Sequential write performance is important in tasks such as caching, replication, HPC, content delivery applications, and database logging. The Samsung XS1715 averages 1,399 MB/s, the Intel DC P3700 scores 1,818 MB/s, the HGST FlashMAX II weighs in with 1,091 MB/s, and the P420m brings in the rear with 647 MB/s.
The Intel DC P3700 continues to provide tremendous performance in sequential write workloads, but the XS1715 also delivers very low latency and solid performance metrics in our testing.
The XS1715 delivers good performance across the board.
Database/OLTP and Webserver
This test consists of Database and On-Line Transaction Processing (OLTP) workloads. OLTP is the processing of transactions such as credit cards and high frequency trading in the financial sector. Databases are the bread and butter of many enterprise deployments. These demanding 8k random workloads with a 66 percent read and 33 percent write distribution bring even the best solutions down to earth.
The Samsung XS1715 averages 113,627 IOPS at 256 OIO, the Intel DC P3700 averages 132,238 IOPS, the HGST FlashMAX II averages 121,240 IOPS, and the Micron P420m averages 102,265 IOPS.
The XS1715 falls between the FlashMAX II and the P420m in the OLTP workload.
The Web Server workload is read-only with a wide range of file sizes. Web servers are responsible for generating content users 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 hosting the website.
The Samsung XS1715 leads this test from 8-128 OIO, again displaying its great performance scaling. The P420m takes the edge at 256 OIO with 184,845 IOPS, but is followed closely by the XS1715 with 181,062 IOPS. The Intel DC P3700 averages 138,073 IOPS, and the FlashMAX II averages 110,747 IOPS.
The XS1715 provides superb performance in this test.
Email Server and File Server
The Email Server workload is a demanding 8k test with a 50/50 read/write distribution. This application is indicative of performance in heavy random write workloads.
The Samsung XS1715 averages 93,238 IOPS at 256 OIO, the Intel DC P3700 averages 107,768 IOPS, the FlashMAX II and P420m are in a dead heat at 256 OIO, with an average of 93,771 and 92,416 IOPS, respectively.
The Intel DC P3700 easily provides the best IOPS-to-Latency ratio in this heavy random write workload.
The File Server workload tests a wide variety of different file sizes simultaneously with an 80% read and 20% write distribution. The wide variety of simultaneous file size requests is very taxing on storage subsystems.
The Samsung XS1715 averages 93,379 IOPS at 256 OIO, the Intel averages 109,585 IOPS, the HGST FlashMAX II averages 91,820 IOPS, and the P420m averages 70,574 IOPS.
The advent of the NVMe specification is revolutionizing the way the industry utilizes flash storage, and its forward-looking design guarantees its impact will extend to future non-volatile memories. Numerous complimentary standards have sprung up to extend the benefits of the optimized interface to externally accessible ports, and even as a network fabric. The development of the NVMe Over Fabric standard is already underway at the Fibre Channel Industry Association, and will side step the performance-inhibiting SCSI translation layer to deliver end-to-end NVMe performance outside of the box.
NVMe Over Fibre is a promising future advancement, but other ingenious standards have already addressed crucial supporting roles for server-side NVMe deployments. The SFF-8639 connector and ExpressBay slots extend PCIe's high speed and low latency to the front of the server. The immediately recognizable benefits of SFF-8639 equipped ports are accessibility and hot-swapping capabilities. Another overlooked advantage is scalability. There are a limited number of PCIe ports in the server, and the ability to aggregate four NVMe devices onto one PCIe port increases the amount of performance and storage density available to the server.
Samsung leverages their own NAND, DRAM, and a customizable PMC controller, to deliver maximum performance, in a surprisingly small package. The Samsung XS1715 makes a big contribution to density with an amazing top capacity of 1.6TB in a small 2.5" 15mm SSD. NVMe provides impressive performance and a standardized platform that is compatible with in-box drivers in all major operating systems. This ensures compatibility and eases qualifications requirements.
In our testing, the XS1715 offered leading random read speeds that even surpassed HHHL competitors. The XS1715 also offers superior scaling under lighter loads, which will be beneficial in bursty environments. The XS1715 has an affinity for random read workloads, but also offers solid performance in random write workloads. The XS1715 scored well in our mixed workload testing, though, it did experience higher latency in mixed random workloads than some of the competing devices.
The Micron P420m posts amazing sequential read performance, but lower sequential write speed. The Intel DC P3700 takes an opposing stance, with leading sequential write speed, but the lowest sequential read speed. The XS1715 offers surprisingly balanced sequential performance that is impressive in both metrics. In mixed sequential testing, the XS1715 continued to provide solid performance.
A particular strength of the XS1715 was its tight latency and performance envelopes in all of the tests. The XS1715 did not lead the pack in our server workloads, with the exception of the read-centric webserver profile, but offered plenty of performance for most workloads.
The purchasing decision for any component consists of a complex mixture of factors largely determined by workload requirements, compatibility, and pricing. The Samsung XS1715 offers expected enterprise-class features, such as a five-year warranty, data path protection, power loss protection, a two million hour MTBF, and competitive endurance specifications.
One area that is a clear differentiator is its impressive density and innovative form factor. The ingenious design merits the TweakTown Best Features Award for the 2.5" NVMe SSD segment.
|Quality, Design, Build and Warranty||95%|
|Power Consumption and Efficiency||92%|
The Bottom Line: Samsung's XS1715 NVMe SSD delivers incredible performance and storage density in an amazingly small 2.5" package with up to 1.6TB of capacity. Competitive endurance specifications are paired with power loss protection, a two million hour MTBF, and data path protection.
PRICING: You can find products similar to this one for sale below.
United States: Find other tech and computer products like this over at Amazon.com
United Kingdom: Find other tech and computer products like this over at Amazon.co.uk
Australia: Find other tech and computer products like this over at Amazon.com.au
Canada: Find other tech and computer products like this over at Amazon.ca
Deutschland: Finde andere Technik- und Computerprodukte wie dieses auf Amazon.de