Here is my perspective, as a storage sales guy who has seen his fair share of industry trends come and go. The storage market is kind of weird these days. Depending on who you talk to, all data is going to be either all flash, in servers or in the cloud. I disagree, because many still think about storage deployments the same way they did in 2004, and believe me a lot has changed. Big, expensive, hard to use, expensive to get performance, massive capital out-lay every 3-4 years and migrations that are a real P.I.T.N (Pain.In.The.Neck – not sure this is the technical term!). That sounds very unattractive to me. In reality, companies can have the agile, cost effective for both performance and capacity, easy to use, highly available storage infrastructure they really desire. You just have to think differently about it. At EMC we have many technologies that accomplish this – and in the mid-range we have three unique technologies: VNX mid-tier storage, VPLEX SAN virtualization, and RecoverPoint replication (where needed). The trick is to let each of these technologies do what they do best without having to give up features of one for the other.
VNX: This is our unified mid-range storage that delivers the lowest cost per IOP AND lowest cost per GB. It’s essentially a hybrid array (support mixed drive types) with features like FAST Cache (extending traditional DRAM cache with SSDs) and FAST VP (automated tiering ) to allow for a very intelligent use of flash drives, while keeping the overall cost down (keep active data on high performing flash drives and inactive data on low cost drives) – let’s face it, everyone has large amounts of data that is not active, but certainly cannot be deleted or moved to a secondary storage device.
VPLEX: In a nutshell, it takes the LUN presentation from the array to a true tier-1 virtualization appliance. The true magic is the ability to have data from two physical locations, be it inside the data center, or in another data center be presented to a host. Pretty cool, and nothing else does this.
RecoverPoint: True homogeneous SAN replication. It really does not care what it is replicated from or what it is replicating to.
Let’s put this together in a very simple recipe to show how we solve real storage infrastructure challenges…and again, I am the sales guy, but this is what I am seeing out there:
- First, deploy SMALL, think 20-50TB increments, of VNX storage with high performing SSDs for FAST Cache and in the FAST VP pool as well as more cost effective larger drives in the pool to lower the overall cost/GB. We will treat each storage array as a “SAN node”, and will grow the environment by adding more “nodes”. When we do, we are adding more CPU, memory and capacity in relatively small increments. This is not unlike a scale-out NAS model We will not upgrade the arrays themselves, we are simply adding small increments of VNX storage as and when needed.
- Then, deploy these behind a VPLEX and you can accomplish many things; such as 100% non-disruptive data mobility between the array deployment, with results in non-eventful storage adds and migrations. This actually allows for this “node” based deployment to now make sense. One other very important capability is to create tier-1 storage from tier-2 storage…huh? We will take the data that is the “family jewels” which needs this type of availability and leverage the cache coherency of VPLEX and have that data on two of the storage “nodes”. If an array is offline for any reason, the host has no idea. This is not failover, but continuous availability inside the data center. “Oh wait, I have to double my storage!”. .Nope. Remember, we are only talking about the small subset of data that requires it, and, we are buying very cost effective mid-tier storage. Great win-win. BTW, this is licensed by the array itself, not by capacity. Very cost effective.
- Now, let’s replicate this. RecoverPoint sees all IO going to VPLEX from the host. RecoverPoint will replicate the IO for these LUNS from VPLEX, not the storage array itself. This is important, and makes this really easy and flexible over time. As data is moved around at the storage array level, RecoverPoint has no idea, and the replication sessions do not have to be set-up again….that is powerful. As storage is moved in and out for technology refreshes or balanced between storage “nodes”, there is no effort to maintain the replication relationships. On the other side, the storage can be mid-tier storage as well. Once again, very cost effective.
So, as you can see, the idea of cost effective, easy to use, flexible, highly available, replicated storage is a reality if we collectively think differently about how storage can be deployed.