Recommend storage for this setup


(David Brierley) #1

Just looking for some recommendations really.

We have a lab setup currently built but once live will be scaled and we want to make scaling as easy and straight forward as possible.

We don’t have a SAN or separate storage network instead we have opted for the hyperconverged setup where each node will use it’s own storage and be made available to other nodes.

I am finding it hard to see which way would be best to achieve this. I have contacted storpool however as this is only in lab stage and it’s currently not big enough they don’t see it as being an appropriate fit (rightly so) so I am looking for another method to present the local storage of each node to other nodes as shared storage but without sacrificing performance.

The lab is only small as mentioned but will be scaling once we get a working solution. I have looked at other virtual sans such as starwinds but can’t see anyone using that on here (or are they ) , I see clustered lvm is also an option but would like to see what other people would recommend or if there is anyone with a similar setup .

Kind regards

David Brierley


(Vadim Tsaplin) #2

Hello.
I use simple SSH System and Image stores for about 350 VM’s and 50 hosts (nodes) in Enterprise IT infrastructure. Every node has:

  • own local System Datastore;
  • 2-4 Gbps Data Network connection, and now I going to upgrade our network to 10Gbps

Advantages:

  • Simple to deploy;
  • We can provide high disk load with low latency;
  • Simple to recover cloud infrastructure;

Disadvantages:

  • Quite long time for instantiate/deploy/migrate VM’s, but it’s not critical for us;
  • We can’t make live migration and native HA for VM’s , but we use different technologies for HA our services.
  • Yes, If we lost one node, we would lost all VM’s on this node. But my colleagues know about it and plan their services appropriately.

May be in future, we will start to use distributed storage like CEPH.