IT/Datacenter & Super Computing News - Page 13

Flash Memory Summit 2014 - The list of software caching solutions is expanding rapidly as manufacturers, and software vendors, try to make their way into this expansive market. Deploying flash is great, but deploying it in a manner that accelerates existing infrastructure is even better. Very few will totally replace spinning disk in their datacenter deployments, so marrying the capacity of HDD with the performance of SSDs is the current go-to solution.

Cachebox wades into this large market with a new method of caching. The CacheAdvance software uses an ASM (Application Specific Module) that analyzes what is important to the application and provides seamless acceleration based upon application-centric performance profiles. The system has predefined knowledge of existing popular programs, but also adjusts to the specific environment.

One of the advantages of application-centric caching is that it doesn't accelerate any unwanted processes or applications, such as backup operations. This maximizes the capacity of the underlying flash solution. Users can use any block-level device for caching, so there is hardware independence from specific vendors and the system is also interface agnostic. PCIe, SAS, SATA, NVDIMMS, and UltraDIMMS can all play equally. Any block-level device is fair game. One particularly exciting aspect of the flexible design is that it will also allow for future storage technologies as well.

Flash Memory Summit 2014 - LiteOn isn't accustomed to the SSD limelight even though they have been in the OEM SSD business since 2008. In many ways they should receive more recognition in the SSD market. LiteOn is known for their ODD (Optical Disk Drive) offerings, such as CD and DVD players. However, they are also the brains behind the Plextor SSD brand, which gives them quite a bit of experience in the retail space with Plextor, and on the OEM side as LiteOn.

LiteOn is expanding into the datacenter with a new line of enterprise SSDs that are tailored for enterprise applications. They are developing both 2.5" and M.2 form factor SSDs with enterprise-centric features. LiteOn is currently working on a 1TB M.2 SSD with power loss protection that was being demo'd in their booth at the show. The SSD features a PCIe 2.0 x2 interface and blistering sequential speeds of 600 MB/s read and 520MB/s write. Jeffrey Chang, the Technical Product Manager at LiteOn, stopped by the booth to talk about the latest enterprise SSDs from LiteOn.

The N9S will serve the 2.5" market with capacities up to 2TB, and are also available in the small 1.8" form factor. LiteOn is initially using AHCI for their new M.2 SSDs, with NVMe solutions coming next year. Stay tuned to TweakTown for future product evaluations as the new SSDs become available.

Flash Memory Summit 2014 - It's always fun to catch up with DriveSavers, simply because of the interesting stories from the latest data recovery projects they are working on. We were lucky enough to take a tour of the DriveSavers labs in April, and we recently crossed paths with Chris Bross again at the Flash Memory Summit.

Chris detailed the recovery of some of Gene Roddenberry's (the creator of Star Trek) lost files, which were stored on 200 x 5.25" floppy disks. These very old disks held a number of files, which are not yet released, and there will be a public disclosure of the content in the future. This highlights some of the varied storage technologies that DriveSavers encounters every day. While staying up to date with helium and SMR HDDs, they also have to continue to extract data from old devices as well.

Chris also discusses the need for data recovery services for the datacenter, and a word of caution for those using cloud data storage. For instance, when Hurricane Sandy struck five datacenters were literally underwater, so users are cautioned to always keep multiple copies of their data. Of course, if you should lose data DriveSavers will be there to help recover it, and you can take a closer look at our tour of the DriveSavers facility here.

Flash Memory Summit 2014 - NVDIMM was a hot topic at the Flash Memory Summit, and Viking Technology is leading the way in the NVDIMM space with their ArxCis-NV solution. NVDIMM allows the use of DRAM memory as a persistent memory. In the event of a power loss the NVDIMM flushes data from the DRAM down to NAND that is embedded on the device. This allows users to leverage the ultra-fast speed of DDR3 (and DDR4 in future versions) as a block storage device, yet still recover from any power loss event with the data intact.

Users can also leverage the ArxCis-NV as a memory space with their software applications. One of the advantages of using an NVDIMM from Viking is that it can be integrated into industry standard x86 servers via the DDR sockets. The ArxCis-NV is a DDR3 240-pin JEDEX pin compatible ECC registered DIMM with an integrated SSD for data backup. The speed of the device far outstrips any 12Gb/s or 6GB/s products, even when they are aggregated behind a RAID controller or HBA.

The ArxCix-NV can be used for the most bleeding-edge applications, such as a write cache or buffer, and also boosting metadata performance. We have a device in the lab now, and look to these pages soon for a performance breakdown!

Flash Memory Summit 2014 - Avago Technologies recently acquired LSI and their SandForce division, and then scooped up PLX Technology as well. The merger with Avago isn't complete yet, so we couldn't discuss the details of the acquisition with Larry Chisvin, the VP of Strategic Initiatives at PLX Technology. PLX Technology is a right in the middle of the datacenter evolution with the advent of NVMe, and also the need for a new interface for converged server architectures. NVMe is currently a new interface that is still developing to offer multipath and fail-over features which will lean heavily on PCIe switches.

We visited PLX's office and took a tour through the ExpressFabric architecture, which can replace or augment networking infrastructure in the datacenter. The advent of hyperscale and cloud computing has spurred a re-imagination of datacenter architecture. Disruptive technologies and Open Compute initiatives are challenging the old way of thinking and delivering designs purpose-built for optimum efficiency. The emergence of software-defined datacenters (SDDC) and software-designed storage (SDS) requires new designs that allow total control of resource allocation. Using ExpressFabric's PCIe connections as a rack-level interconnect allows for new exciting datacenter designs that will pool server resources in separate chassis for optimum performance.

Integration with Avago will allow for access to a much larger company with a large IP portfolio to accelerate PLX Technology's initiatives.

Flash Memory Summit 2014 - Ray Pang, Director of Marketing at Skyera, stopped by our booth to discuss the skyHawk All-Flash Array (AFA). Skyera aims to bring the price of AFA down below $3 per GB, a similar price to disk-based storage subsystems. Skyera accomplishes much of the cost reduction by using their own design that doesn't rely upon traditional form factors. The end result is a remarkably slim 1U device that boasts up to 44TB of screaming flash storage.

skyHawk delivers 88 times more storage density at 1/10th of the power consumption of a like-priced HDD solution. Another key benefit of skyHawk is its converged storage architecture, which unifies block-based iSCSI and NAS file systems (NFS v3). The slim package provides up to 2.4 GB/s of bandwidth and 400,000 IOPS. skyHawk also employs compression and de-duplication technologies to boost the amount of use-able SSD space, but has base configurations before data reduction of 12, 22, and 44 TB.

Delivering all of this performance requires either 40 x 1GBe connections, or only 3 x 10GBe connections. The total package weighs under 20 lbs and pulls less than 350W, which is insanely efficient for a datacenter storage system.

Flash Memory Summit 2014 - As we reach the end of planar (2D) NAND scaling we will begin to see more disruptive products emerge with futuristic technology such as PCM and magnetic and carbon nanotubes, the only question is which will rise to the top as the premier after-NAND technology. These new technologies will initially emerge as complimentary technologies to the existing NAND used in SSDs. Saul Zales, the Chief Executive Officer of Contour Semiconductor, brought along a wafer of their DTM (Diode Transistor Memory)that is valued at nearly $10 million, due to its early design and rarity. DTM is a phase-change based memory (PCM)that reduces the mask and process costs by up to 65% compared to NAND.

The 4F2 self-aligned structure uses a vertical epitaxial diode as the select device. In the future, Contour Semiconductor could include magnetic or carbon nanotube storage elements. The wafer is built on the CMOS process with a cross-point array architecture that supports word, sector, and page-level erase commands, which are very similar to NAND's design and fucntion. This will allow the use of the DTM packages with NAND controllers. We were lucky enough to see one of the only DTM wafers in existence, and had a quick chat with Saul Zales in the booth.

What does all this mean? Up to a billion P/E Cycles, which would be nearly immeasurable endurance compared to current SSDs. Using this type of memory as a cache for less-durable NAND could boost performance and endurance exponentially, and eventually new storage devices could use this type of memory.

Flash Memory Summit 2014 - BiTMICRO's newest MaxIO SSDs pack a whopping 6TB of data into a half-length full-height PCIe form factor. The new MaxIO SSDs feature full data path protection and their proprietary DriveLight Management Software. The MaxIO series leverages Toshiba A19nm MLC NAND to provide 240,000 random read IOPS and 100,000 random write IOPS. Sequential performance weighs in at 1.1 GB/s read and 800 MB/s write. The MaxIO is exceptionally power efficient, requiring less than 30W active for up to 6TB of flash storage. The MaxIO also differentiates by running all of its processes, including ECC, on the SSD. Other alternatives utilize the host server to handle these basic flash management functions.

BiTMICRO tackles some of the inherent challenges of a large capacity SSD by separating their Talino controller architecture into a Split ASIC technology, placing the FTL (Flash Translation Layer) onto a separate processor. This allows the MaxIO to effectively manage the much larger LBA range of a high capacity device, while also enabling global wear leveling techniques to enhance endurance. Cramming as much capacity onto one device is critical in high-density applications that need the performance of an AFA (All Flash Array), but want to maintain the highest performance by keeping the SSD as close to the processor as possible. This helps to reduce any associated networking infrastructure. The drives are optimized for read-centric applications and features up to 1 DWPD of endurance during the five-year warranty period. We were lucky enough to speak with Zophar Sante and Stephan Uriarte about the differentiators and advantages of their new architecture in the video below.

As always, stay tuned for the latest updates on TweakTown.

Flash Memory Summit 2014 - The SSD industry is moving fast, and several companies with core SSD controller IP have been purchased by larger companies. As SSDs become more mainstream some of the smaller players, such as SMI, JMicron, and Phison, have become more prevalent in the SSD market.

We took a few moments to speak with Iri Trashanski from Marvell Technology Group about the state of the SSD controller industry. Marvell has one of the broadest lines of SSD controllers in the industry, and they recently announced their latest NVMe SSD controller. The NVMe 88SS1093 also features 3rd generation NANDEdge LDPC (Low Density Parity Check) technology. LDPC is the newest form of error correction that requires less space on the drive and also provides more error correction power. LDPC also allows for varying levels of error correction during different stages of the SSD life cycle.

The 88SS1093 features 4GB/s or 2GB/s endpoints, dependent upon host system customization, via a x4 PCIe 3.0 connection. The leaner command set and multiple queue/pair mechanisms within the NVMe stack will provide higher performance and lower latency in comparison to older interfaces.

Flash Memory Summit 2014 - Jim Handy from Objective Analysis stopped by our booth to talk about the latest and greatest in the storage market. Jim's 3D NAND series, which explores the rationale behind 3D NAND and delves into an in-depth breakdown of the architecture, has became a go-to source of information on 3D NAND. His informative series of articles can be found here on thememoryguy.com.

Jim also discusses the possibilities of future storage technology. Objective Analysis offers third-party independent market research and data for the semiconductor industry and investors in the semiconductor industry.