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

HPE 3PAR Storage Class Memory: Turbocharged Application Acceleration Is Here



Here’s the challenge: delivering high performance from a small number of high capacity SSDs. Enter HPE 3PAR Storage Class Memory, the cutting-edge turbocharger for the 3PAR StoreServ all-flash engine.

Recently, I blogged about the changes that NVMe and Storage Class Memory are poised to bring to the IT infrastructure market with the 3PAR point of view on those technologies. So I couldn’t think of a better way to bring it all home than talking about how 3PAR uses both NVMe and SCM to deliver the next big step in application acceleration. I am talking about HPE 3PAR Storage Class Memory, the exciting new technology we previewed at HPE Discover London 2016.

Before we dive into 3PAR SCM, I want add that a technology preview is meant to showcase how HPE 3PAR StoreServ storage can leverage next-generation solid-state technologies and protocols.

Turbocharge your all flash system 

3D Cache_turbocharge_blog.jpg

Formula 1 racing represents the pinnacle of automotive innovation and with each passing year, the cars get lighter, more powerful and faster. One would think that F1 cars would be using some monstrous 7-liter 12-cylinder engine to get them to speeds over 200mph. Surprisingly, the current generation F1 engine is a 1.6 -liter 6-cylinder unit that uses a big turbocharger and whole host of other technologies (these engines are all hybrids interestingly, which will make you look at your humble Toyota Prius a little differently I am sure) to deliver all the power these cars need. The trend is to squeeze more out of a small power plant as efficiently as possible – and you will see this in the latest road going cars as well. I think the trend is no different in the storage world– high capacity, cost effective SSDs are shrinking storage footprint dramatically. The challenge is delivering high performance from a small number of high capacity SSDs. Enter HPE 3PAR SCM – the cutting edge turbocharger for the 3PAR StoreServ all-flash engine.

3PAR SCM works as an extension of DRAM cache and combines HPE 3PAR’s intelligent caching algorithms with Storage Class Memory (in this case Intel® Optane) and NVMe to deliver extreme application acceleration with HPE 3PAR StoreServ All Flash storage systems. Come to think of it, 3PAR SCM is designed to accelerate flash itself. That is remarkable if you consider the fact that flash took off (and is still extensively used) as a caching layer and now we have devised ways to accelerate flash itself. History does repeat itself, doesn’t it?

The premise is very simple:

  • Intel Optane SSDs provide access times that are a fraction of standard NAND flash.
  • NVMe reduces overheads and access times even further.
  • 3PAR caching places data intelligently such that DRAM cache misses are being served out of Intel Optane.

The end result is a significantly lower latency profile.

Let’s talk numbers – around 50% and 80%

So how much lower latency are talking about? Our tests have shown an improvement of over 50% in read latency for a volume with 3PAR SCM turned on compared to one without 3D Cache. As you can see from the 3PAR StoreServ Management Console (SSMC) screen shots in Figure 1, latency for the volume with 3D Cache is 0.23ms compared to 0.54ms for the non 3PAR SCM volume. What is even more impressive is that the very tight latency curve barely moves from the 0.23ms for the duration of the test – almost 95% of the read IO is serviced under 0.25ms.

EDITOR'S NOTE: the name original used in the diagrams for 3PAR SCM was 3D Cache.

3D Cache Figure 1J.jpg

Note: Read service times comparisons for volumes with and without 3PAR SCM enabled.

A very useful side benefit of lower latency is higher IOPS and as you can see from Figure 2, the IOPS for the volume with 3PAR SCM is almost 80% higher (85,600 IOPS) compared to the volume without 3PAR SCM (47,741 IOPS).

We believe small block random read workloads can benefit the most from 3PAR SCM. Also as mentioned earlier, 3PAR SCM based acceleration is the answer to the problem of falling IOPS/GB for all-flash systems configured with relatively fewer large capacity SSDs (like the 15.36TB)

3D Cache Figure 2J.jpg

Note: The test set up was a 2-node 3PAR 20000 class all-flash storage system with 8*15.36TB SSDs and two Intel Optane AIC SSDs, one in each controller node. The workload used for the 3PAR SCM and non 3PAR SCM volumes is a small block (4KB), random with 85% Read/15% Write mix. The volume named OTHER is running a 16KB, random workload with 40% Read/60% Write. Overall workload is >200k IOPS.

Focus on latency

What is most impressive is the latency profile of the volumes that are accelerated by the Storage Class Memory. As you can see from the histogram below, 55% of data is services in 125 microseconds or less and more than 10% serviced in 62 microseconds. If you compare this to the volumes not accelerated, you can really see the shift in service time from the blue bucket (250 microseconds or less) to the grey bucket (125 microseconds or less).

3D Cache Figure 3J2.jpg

What’s next?

As I mentioned in the beginning, the performance results you are seeing here is a preview of the 3PAR SCM technology. This is cutting-edge technology actively worked and will be productized in a future date. The key point that I want to leave you with is here is that 3PAR SCM is the latest proof point for how fast the 3PAR architecture can adapt to new media technologies. Intel Optane is the latest in solid-state memory technology with very different characteristics from NAND flash and 3PAR is one of the earliest adopters of that technology. Finally, I am sure you have a very good reason to watch this space for more details in the coming months.

Disclaimer: This document contains forward looking statements regarding future operations, product development, product capabilities and availability dates. This information is subject to substantial uncertainties and is subject to change at any time without prior notification. Statements contained in this document concerning these matters only reflect Hewlett Packard Enterprise's predictions and / or expectations as of the date of this document and actual results and future plans of Hewlett-Packard Enterprise may differ significantly as a result of, among other things, changes in product strategy resulting from technological, internal corporate, market and other changes. This is not a commitment to deliver any material, code or functionality and should not be relied upon in making purchasing decisions.


Aravindan Gopalakrishnan.jpg

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

0 Kudos
About the Author


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


When it will be planned to be available for customers ?

As I understand that's good for read-intensive loads, but how about for write-intensive loads ?


Anton - we can't speak for Intel and generally don't share dates about futures publicly. It's targeted to be use as an extension of read-cache so there's no information about write performance. 


I'm guessing the SSD will be internal to the cintroller, PCIe attached?


@B_Davis You are correct that the 3D Cache will be PCIe attached via the 3PAR controller. 

Nov 27 - 29
Madrid, Spain
HPE Discover 2018 Madrid
Learn about all things HPE Discover 2018 in Madrid, Spain, 27 - 29 November, 2018.
Read more
See posts for
HPE at 2018 Technology Events
Learn about the technology events where Hewlett Packard Enterprise will have a presence in 2018.
Read more
View all