An Evolutionary Journey of Enterprise Storage for an SMB: Part 1
Welcome to my first blog in a series of three, discussing how we came to our decision about which storage solution was best for Evolution Recruitment Solutions, and why.
The start of the process
Choosing the right storage solution for your datacentre can be like navigating a minefield. Having recently gone through this process for Evolution, I wanted to share my experiences and highlight some areas which you should take some extra time to consider.
Firstly, not one size fits all. There are different solutions to the challenges you’re trying to solve, or make improvements upon, and probably numerous products, technologies, and ways to architect within each of these solutions as well. The choices are plentiful, so a good starting place is to analyse your environment, look at what you’re trying to achieve and within what sort of budget. Different solutions are more, or less, applicable depending on your environment.
Let me first describe our environment, which is probably a very apt time to use the well-known phrase, “a picture is worth a thousand words”:
This doesn’t constitute the entire infrastructure in Evolution’s datacentre, but it certainly makes up its core, for the most part. These two clusters house pretty much all the servers required to deliver Evolution’s business critical services, which is around 80 VMs. All these VMs are common place in a lot of business environments, Windows AD servers, Exchange email servers, SQL servers, file servers, and so on.
What I like most about Evolution’s datacentre is that it might not be the biggest deployment of datacentre hardware and technology; it’s quite a small deployment; but it’s a very neat, effective solution, which has been architected N+1 resilient. Everything has redundancy built in, and with VMWare as the virtualisation platform, this provides excellent flexibility and high availability for business services. It is quite rare to find a small business of Evolution’s size prepared to make such a relatively high level of investment in IT infrastructure, but it has rewarded the business tenfold on its investment.
The Dell EqualLogic storage array, purchased in 2011, had been a great servant over the years, but Dell made the device EOL as the business needed more storage capacity. In the past 6 years, business data had grown, the EqualLogic was full and we were regularly seeing disk latencies above ideal levels at peak IO times. So, these were the challenges; capacity, speed, an aging storage array and, let us not forget, budget.
Well, when it comes to which solution would be best, we started by defining what we wanted:
- Lots of storage capacity
- To have the flexibility to allow us to expand capacity easily
- Performance – our solution of choice needed to be blisteringly quick with low latency, particularly in our case for reading data, as well as offering us lots of IOPS. The more IOPS, the better!
- A cheap solution
When we look at our requirements, it’s clear that we wanted what we all want; the whole world, for free! Clearly we are a delusional species of animal, but let’s see what gets us closest to what we want.
There were a few other additional considerations which we needed to factor into the decision. Firstly, the cost for large capacity SSD drives was tumbling to the point that an All-Flash array was becoming a lot more viable, and I only see this trend continuing. The storage industry is changing, which is leading to the death of mechanical hard disks.
If we do look to introduce flash SSDs into the shared storage, then storage will no longer be a performance bottleneck; this will move to the network. The existing fabric iSCSI network runs at 1GB, so we should also consider upgrading the fabric switching stack to 10GB, so as to make use of the performance that flash SSD storage will deliver.
Be sure to check back next week for part two of this blog series, discussing a variety of solutions we considered – including the ones we ruled out straight away!