The storage networking market now is teeming with flash solutions. Consumers are probably sick to their stomach getting a better insight which flash solution they should be considering. There are so much hype, fuzz and buzz and like a swarm of bees, in the chaos of the moment, there is actually a calm and discerning pattern slowly, but surely, emerging. Storage networking guys would probably know this thing well, but for the benefit of the other readers, how we view flash (and other solid state storage) becomes clear with the picture below:
(picture courtesy of http://electronicdesign.com/memory/evolution-solid-state-storage-enterprise-servers)
Right at the top, we have the CPU/Memory complex (labelled as Processor). Our applications, albeit bytes and pieces of them, run in this CPU/Memory complex.
Therefore, we can see Pattern #1 showing up.
Pattern #1 states that the industry is pushing the processing of the applications closer and closer to the CPU/Memory complex for performance sake. This performance is usually measured with higher IOPS, higher throughput and lower latency, mostly notably lower latency.
Why? By reducing the “layers” or whatever (hardware, software, protocol, specifications, translation layers etc) in between, the lower the latency, the faster the applications are processed, resulting in faster response time, higher productivity and so on.
Also from the diagram above, another pattern is forming. We can see that different form factors are emerging. From the storage network, be it SAN or NAS, the “form factor” is the enterprise storage array. The storage array could be an all-flash or a hybrid, or even a PCIe-based flash card on the storage itself (like the NetApp Flash Cache), but it is really a black box to the CPU/Memory complex. It doesn’t really care where the data blocks are coming from, but as long as it reaches the CPU/Memory complex, it is fine.
Moving closer and beyond the cables of the IP or Fibre Channel network, we have server-side flash. This is the new battleground because we have been experiencing a hive of heavy duty activities, each trying to trump one another. Here at the server side, there are the usual 3.5″/2.5″ disk form factor of flash, the PCIe-based form factor flash and also the memory-DIMM slot form factor flash. There are others in the market as well, such as M.2 form factor as well as Express Bay form factor.
And this area get more convoluted with new introduction of specifications such as NV-DIMM, SATAe, SCSIe, NVMe and some proprietary ones as well. Every one of them has been touted to remove the shackles that has tied storage to the shackles of the spinning drive architecture.
(picture courtesy of http://www.loromgroup.com/research/img/nv_dimm.jpg)
The picture above shows the up and coming NV-DIMM architecture. As you may have observed, the entire storage networking industry is going through an important upheaval that is nonetheless, good for the industry.
Pattern #2, which is pretty obvious, is really about pushing the storage architecture closer and closer to the CPU/Memory complex.
However, since Pattern #2 is within the physical confines of the host, aka the server, we have to be aware of the limitations of the server. The physical limitations are obvious and the bullets we, the storage networking professionals, have used in the past such to shoot down Direct-Attached Storage (DAS) are the same. Scalability can still be an issue, although x86 servers with 60 x 3.5″ slots for hard disks, 6-8 or more PCIe slots are pretty mainstream nowadays.
3 things to note in a server-side storage are standardization on high availability and RAS (reliability, availability and serviceability) and hot-pluggability. There is certainly more proprietary physical hardware implementation and lack of industry standardization in this area when choosing a server-side storage.
I would probably call this Pattern #3 (for now) even though it is more of a negative one.
Pattern #3 is the lack of standardization when it comes to RAS, hot-plug and HA.
Necessity is the mother of invention, and therefore, to overcome Pattern #3, software is making all the difference. I would probably drop in the hot buzz word called “Software-Defined Storage” here because it is software that is overcoming Pattern #3’s limitations. VMware’s VSAN, PernixData FVP, Nutanix NDFS, VMware Virsto (when it is ready), Simplivity’s hyper-converged architecture, Virident FlashMax and vFAS, Datacore SAN Symphony-V and so many more that have been thrown at me in the last few months, are breaking down the barriers and limitations of a single physical server.
While standards are probably coagulating and taking shape at the server-side storage, both hardware and software, the industry simply can’t wait. Players such as Virident (through friends of mine previously with HDS Australia), are taking matters into their own hands and I must say that they have done the right thing to tie themselves expertly to enterprise applications such as Oracle RAC, Microsoft SQL Server, VMware VDI and MySQL.
Mind you, this is not anything simple. What Virident has done very smartly (while the entire server-side flash market is battling and jousting for their positions), is to intricately craft and deeply test these enterprise applications with their solutions.
For example, Virident’s FlashMax deep integration with the Microsoft Storport architecture results in enhancing the performance, stability and greater scalability of the Microsoft SQL Server. Virident also recently announced their certified performance-based architecture with Oracle RAC and again, it is their deep integration with Oracle SGA Buffer Cache and Smart Flash Cache that has resulted in a very high performance Oracle RAC architecture. They have a whitepaper out recently on accelerating Oracle database.
The consumers, through these Virident certified enterprise applications, are the ones that will benefit greatly for the work Virident has done so far, and I am sure there are more enterprise applications that will be part their portfolio of certified and proven enterprise solutions.
The one that impressed me about Virident is their boldness in vision. Amidst the chaos (and confusion) of server-side flash, Virident took matters to give what customer wants. Virident FlashConnect is another solution that gives a lot of value. In the page, FlashConnect has the vHA, vShare and vCache software modules
I don’t specifically have a lot of deep knowledge of each one of the software modules but go check out their website, because the pictures are pretty intuitive. Maybe in another future blog entry just on these software modules?
What Virident has done is what Pattern #4 is right now (although things will change).
Pattern #4 is server-side storage vendors are innovating to give their customers what they want right now – Application assurance.
These are the things I have been picking up along these few months. Flash is rocking the very core foundation of storage networking and it will redefine storage industry as a whole, and guys like Virident are leading the way.