Around the Storage Block
Showing results for 
Search instead for 
Did you mean: 

3PAR and Storage Class Memory: From Ludicrous Speed to Plaid



The same HPE 3PAR StoreServ architecture that delivers all-flash performance efficiently today will also enable the next step up in application performance with Storage Class Memory (SCM.)

Over the next five years, the most significant technology shifts in IT infrastructure technology (and the storage technology in particular) are expected in the areas of storage media and networking. In my last blog post on NVMe and need for ludicrous speed, I focused on one aspect of the upcoming change GoneToPlaid.jpgin storage media – the emergence of NVMe to replace SCSI as the interface of choice for accessing solid-state media. That topic wouldn’t be complete without an examination of the advances to the actual solid-state media itself – the emergence of storage class memory (SCM). So let’s do precisely that in the rest of this post and take a deeper look at how the solid-state media landscape is changing and how we are preparing for those changes with HPE 3PAR StoreServ storage.

Welcome to the brave new world of Storage Class Memory

In the spirit of my last post on NVMe, let me turn to Spaceballs once more (this is the last time, I promise) to explain what SCM is all about. NVMe with NAND flash took applications running at light speed to ludicrous speed. Combine NVMe with SCM and you get to the next level – plaid. (I am sure all of you Millennials remember this from your last Google search).

Here is the more serious and boring version with the standard memory hierarchy triangle. The main reason NAND flash has been a game changer is that it reduced the gap between DRAM (operating at nanosecond latencies) and mechanical storage media (operating at tens of milliseconds) by an order of magnitude. While NAND offers significantly better access time (90 to 100us), it still is an order of magnitude slower than DRAM devices operating in the 80ns range with considerably lower endurance.

standard memory hierarchy triangle.jpg

Enter Storage Class Memory devices

Storage Class Memory devices promise access times in the single digit microseconds and endurance similar to DRAM at costs closer to NAND flash devices. Couple such a device with NVMe and you are looking at the next big step in reducing application latency. Besides being used as regular block addressable storage, these devices can also be used as byte addressable making them apt for extending and replacing DRAM capacity in systems. 

The idea is that eventually the right SCM technology will collapse the whole memory hierarchy into a single pool that will serve as main memory and non-volatile primary storage. This will relegate slower NAND technologies to the archival tier – what is referred to as “universal memory.” While using SCM as a fast storage tier and displacing DRAM is expected to be reality in the next 2 to 3 years, “universal memory” will take much longer given the fundamental operating system and programming model changes it involves.

My view is that in external storage systems SCMs will appear as caching devices or as a tier of superfast storage above NAND flash. The advent of high density NAND flash drives (which are going to get denser with QLC) poses an IOPS/GB challenge that could potentially be solved with the right SCM technology.

The SCM landscape

Several non-NAND SCM technologies are in various stages of development. Some prominent examples are technologies such as Magnetoresistive Random-Access Memory (MRAM), Resistive RAM (ReRAM), Phase Change Memory (PCM), Memristor and Intel 3DXPoint. The last one of the list, Intel 3DXPoint, is expected to ship (as “Optane” SSDs) as early as next year and will be available as DIMMs at a later date. Samsung has also announced the availability of its own technology, dubbed Z-SSD, for next year as well. Western Digital has talked about 3D ReRAM as a promising SCM technology.

The flurry of activity and announcements in this space point to the importance of this emerging market that various estimates put at over $2B by 2020. So how are we at 3PAR planning to take advantage of this new technology?

The 3PAR view – it’s all about investment protection

3PAR StoreServ has a track record of adapting to media changes successfully and enabling our customers to take advantage of the latest technologies with minimal disruption. With its fine-grained three-layer virtualization, the 3PAR architecture can integrate new media types faster than competitive architectures that are purpose built for NAND flash.

Stay tuned for some exciting news on this very topic. It’s also important to note that SCM devices need an efficient, low latency interface like NVMe to deliver the promised performance gains and NVMe requires a media that can fully exploit its value.

To sum up, the key message that I want to leave you with is that the same 3PAR architecture that delivers all-flash performance efficiently today will enable you to take the next step up in application performance with SCM.

Aravindan Gopalakrishnan.jpg


Meet Around the Storage Block blogger Aravindan Gopalakrishnan, Product Manager, HPE 3PAR.

About the Author


Our team of Hewlett Packard Enterprise storage experts helps you to dive deep into relevant infrastructure topics.

June 18 - 20
Las Vegas, NV
HPE Discover 2019 Las Vegas
Learn about all things Discover 2019 in  Las Vegas, Nevada, June 18-20, 2019
Read more
Read for dates
HPE at 2019 Technology Events
Learn about the technology events where Hewlett Packard Enterprise will have a presence in 2019.
Read more
View all