Redundancy and performance is extra important on storage, so today we were at the DC brainstorming this.
We ended up on decision to build a big cluster of small machines, and keep on upgrading those on a as needed basis. This will give much better redundancy as things are relying on many more individual machines and scale out is easier on our budget.
This doesn't mean any changes on the current offerings, just how they are going to be provisioned in a month from now. After this initial expenditure this allows us to put much more budget upfront on the nodes and then weekly storage upgrade routines.
We are targeting a year goal of 120+ HDDs on the cluster with 40+ SSD caches, with steady throughput capacity of 40Gbps+, likely to be upgraded to quite fast to as high as 400Gbps, depends upon the needs.
Also, we are going to experiment sooner than expected with local SSD caching, and purchased a lot of mSATA SSD drives for testing purposes.
Our IP Block has been now assigned, and tomorrow people are working towards getting things online.
Friday, July 5, 2013
