DSSD D5 – EMC’s quantum leap in flash storage with the rack-scale all-flash appliance

In their ‘Quantum Leap’ announcements, alongwith VMAX All-Flash storage EMC announced DSSD D5 flash appliance on Monday (29-Feb). And it is really a big leap in flash storage as they produce un-matching performance and capacity in small footprint. D5 is a ‘rack-scale’ flash appliance for today’s today’s and next-gen workloads, e.g; applications built on Hadoop.

emc-dssd-d5

 

10 Million IOPS and 100 Micro Seconds latency are real big numbers. Let’s dig in more to other highlights from D5 here in this post.

Continue reading

Hardware : DSSD D5 have a 5U module which with 2 redundant controllers and 36 drive slots. It can support a maximum of 144 TB of raw capacity (with 4TB flash modules). The system offers 100 TB of usable flash capacity in 5U. The system now supports 2TB/4TB (dual-ported, hot-swappable) flash modules and more in near-future.

NVMe : It’s ‘NVMe show’ with D5 as it helps the servers to have end-to-end PCIe NVMe access to the Flash modules. It is the key for providing shared flash storage via PCIe, by the NVMeF (NVMe over Fabric) in use in D5. The system supports 96 ports (48 ports per IO module) for host connectivity. i.e, 48 hosts (PCIe3 x4 lane availability) can be directly attached with redundant connectivity. As it is end-to-end NVMe enabled, there is no overhead of protocol translation.

DSSD D5 will act as a direct attached shared flash storage for varying workloads. It’s simply direct attached Flash with enterprise availability and features. NVMe  While supporting today’s database/data warehouse workloads it supports next-gen workloads including applications based on Hadoop also.

Availability : DSSD D5 is now available for order.

Additional reference : DSSD D5 data sheet,  EMC Press release

With FLASH, things are changing ‘in a flash’ !?

The data storage industry has gone through major changes in the past 5-7 years. And is still in the phase of changes. The user requirements has changed from “More GB’s or TB’s” to “More IOPS”. Introduction of FLASH drives been a catalist for these changes along with the other changes in hardware and software industries. FLASH as a Storage drive or PCIe memory is playing an important role now a days and that gave raise to many new technologies and products.

Speed

Continue reading

All-FLASH Storage Arrays : Many new storage companies started with AFA products and they managed to become the customer’s first-choices for the AFA arrays. As they started with AFA, they had their hardware and software designed/optimized for a complete-flash configuration. Now we are seeing the competitions between the existing players and the new comers to secure their place in the market. Their (the hybrid vendors’) success will be measured on how the well the Operating Environment (in-use for hybrid) is now optimized for an AFA configuration.

With the variety of FLASH memory modules available – NVRAM, PCIe FLASH, SSDs – the AFA’s are going Hybrid AFA’s also.That shows the degree of change from HDD-based storage to Hybrid-Storage to AFA’s and then to Hybrid-AFAs.

NVMe, 3D XPOINT and more… : The increasing demand for FLASH lead to many researches and many new technologies including Intel’s NVMe and 3D XPoint, the recent. FLASH memories made their entry as server-side PCIe FLASH initially and now storage systems as PCIe FLASH (Along with SSD drives) . NVMe was introduced as a Standard protocol for accessing PCIe FLASH. Read more about NVMe here

3D XPoint introduced a memory architecture with higher performance and low cost, which can be a mid-layer for DRAM and FLASH (can be up to 1000x faster than NAND and up to 1000x  the endurance) . 3D XPOINT combines the advantages of existing technologies like the performance of DRAM and lower cost and volatility of NAND. This new technology was announced by Intel and Micron in the mid of 2015.

A post mentioning 3D XPOINT will be incomplete without this architecture diagram. Courtesy : intel.com

A post mentioning 3D XPOINT will be incomplete without this architecture diagram. Courtesy : intel.com

With the stackable memory architecture 3D XPOINT have an improved density (more memory in less space). Data cells are read by varying the voltage being sent and this reduces the need of transistors. This is another reason for cost reduction. Faster cell switching and fast-write algorithm reduces the latency of IO in 3D XPOINT technology. Being comparatively new technology, 3D XPOINT will take few more months to have a real impact in the market.

Intel Optane technology of new Intel SSDs with high performance and high endurance is based on the 3D XPOINT technology.

SFD8 presentation from Intel had a lot about NVMe and 3D XPOINT. Looking forward to their presentation at SFD9 in March. They may have many thing new to discuss this time, 6 months is a long time.

SFD8 presentation videos from Intel are available here. More write-ups from Tech Field delegates are available here.

Here you can find more posts from me on SFD8.

Disclaimer: My travel and accommodation expenses for the SFD8 (Storage Field Day 8) event were paid by the Tech Field Day team. I am not requested to post about event content and I am not getting paid for any posts or for the time I spend during the event.

Understanding NVMe (Non-Volatile Memory Express)

SFD8 (Storage Field Day 8) gave us a great opportunity to listen from Intel about a few latest technologies on SSDs (Solid State Drives). The main discussions were on NVMe, 3D XPOINT, ISA-L etc… . I had to do some serious self studies for a better understanding as these topics were quite new for me. Here in this post I am discussing a few points on NVMe.

2000px-Intel-logo.svg

Continue reading

NVMe What & Why: Introduction of FLASH was big a milestone in the storage industry. A great innovation to the industry which was completely relying on the slow performance spinning disks. But even with FLASH, the way the media was being accessed remained the same – SATA and SAS. This became a big bottleneck for the high performance SSDs.

NVMe was formed as a standard protocol for accessing PCIe FLASH devices. It was formed by a group (NVMHCI group, later turned  NVM Express Inc.) of companies (Lead by Intel) in order to standardize the interface/protocol by which PCIe FLASH were being accessed. Before NVMe, each FLASH vendor was having their on standard of PCIe FLASH access. NVMe v1.0 was released in 2011.

logo-nvmexpress_tmNVMe architecture makes efficient use of the parallelism with SSDs. The result is of course high performance. Compared to SATA (which can have only 1 queue and can have 32 commands), NVMe can have 65536 seperate queues which can have 65536 commands per queue. Below table from Wikipedia will give you a comparison in more detail.

AHCI_vs_NVMe

Thus with NVMe, vendors can leverage the existing PCIe technology in producing a high performance and highly efficient FLASH modules.

intel.web.480.270

Image courtesy : intel.com

PCIe FLASH vendors are having NVMe support with their new products in the market. The product acceptance depends highly on how good they are in supporting the NVMe admin and IO command sets.

Intel SSD 750 series (in pic) is an example of PCIe FLASH with NVMe support.

 

Operating Systems : Most of the common Operating systems are now supporting NVMe. Windows 8.1 and Windows server 2012 R2 support NVMe and NVMe drivers are available as updates in Windows 7 and Windows server 2008 R2. NVMe drivers were made available for Linux and is added into the Kernel version 3.3 . NVMe is supported by other O.Ses including Solaris, OS-X, Chrome OS, OpernBSD and FreeBSD.

Additional Reading : There are many good posts available on internet on NVMe. Posts from PCWorld and flashstorage deserves a special mention. You may also find more resources from NVMe website and you will find the Wikipedia page also very useful.

SFD8 presentation videos from Intel are available here. More write-ups from Tech Field delegates are available here.

Here you can find more posts from me on SFD8.

Disclaimer: My travel and accommodation expenses for the SFD8 (Storage Field Day 8) event were paid by the Tech Field Day team. I am not requested to post about event content and I am not getting paid for any posts or for the time I spend during the event.