By Richard Hogg - Presales Consultant, Proact IT UK.
As part of the ongoing evolution of datacentre platforms, a new breed of system has hit the point of product maturity and is now worthy of consideration for future deployment. We’ve gone from the disparate stove-piped systems of the 90’s (server + discrete storage), through the SAN/NAS era from the late 90’s onwards and depending on who you speak to, two or three rounds of x86 virtualisation and consolidation.
This shrinking of datacentre infrastructures has given system designers a new balance to strike between storage performance, capacity and host performance, particularly when planning around scale within the infrastructure. Do you buy a storage platform on day 1 that meets the possible year 3 or 5 capacity, or take the chance on a potential controller swap down the line. How do you manage the platform moving forward, do you look at disparate management tools for server and storage? Or do you need to deploy an additional overarching management layer?
There are a few products that take the integration headache away from traditional deployments, such as VSPEX, FlexPod, FlashStack and vBlock. These restrict the products used within the stack and in some cases provide a single point of management, but these converged product stacks are still delivering the physical infrastructure as discrete elements, leaving you with the traditional problem of initial size and scale.
Hyper-converged platforms take this to the next step, delivering infrastructure as building blocks rather than discrete datacentre elements, they bundle the compute, RAM and storage up into a single brick, cover it with a single management wrap and deliver a single, scalable platform. If you need more storage performance, you just drop in another brick, need more CPU across the farm, drop in another brick.
Whilst this delivers predictable scale, it does also come with some caveats. The main caveat is that you are now tied in to a particular manufacturer for your storage/compute for scale to remain within that single management wrap. If you go with the appliance-style delivery, you can end up with either too much compute or too much storage for your infrastructure as each node comes with an amount of RAM, CPU and Disk, you can only scale in these chunks. Some platforms can be delivered as a purely software-based offering, allowing you to create a platform which delivers more compute than storage, or create a storage-heavy cluster of nodes.
Essentially, there is no simple answer, you need to look at your current infrastructure refresh, a greenfield platform is much easier to determine, but with a rolling refresh, you need to decide on your breakpoints as to where you would cut over to these platforms in order for it to make commercial and operational sense.
Working with an independent vendor such as Proact, we can help you determine the cost-effectiveness of each platform choice, be it legacy components, pre-integrated or hyper-converged and build a roadmap for your infrastructure and business needs.