So in general UCS is gathering some great traction. Along with Virtualization it's pulling storage pretty hard as well. One of the things that I continue to see is the underestimation of resourcing for storage in a virtual environment.
Let's now take a look at the eco-system of storage in a virtualization context with Cisco's UCS providing the vCompute platform. In this specific scenario it is entirely possible to achieve a very nice consolidation number.
But your storage doesn't need to be sweet ... right ? Wrong! Let's take an example. You consolidate 10 servers onto 1. One of the servers was VERY storage intensive but CPU cycles remained very low. This was due to a specific application that didn't like Storage latency.
But the decision that was made was to deploy these VM's onto an iSCSI connected RAID 5 array. This is a S/W iSCSI initiator connected storage device. You have 10 VM's and your storage sensitive VM is having massive issues. So what do you do? Where did it all go wrong. The planned storage implementation has enough disk space but the IOPS profile and connection choice is all wrong for the intended purpose. So how can it be fixed the budget is spent and the storage doesn't support online RAID level migration along with the fact that even if you could migrate the RAID level you would be completely out of storage. What can be done and how do we do it?
Well that's a case for another post. But if you ever end up in a position like this it's going to be a tough spot and it is possible that there might not be a good answer.
Talk to you tomorrow and I'll post about this again.
Let's now take a look at the eco-system of storage in a virtualization context with Cisco's UCS providing the vCompute platform. In this specific scenario it is entirely possible to achieve a very nice consolidation number.
But your storage doesn't need to be sweet ... right ? Wrong! Let's take an example. You consolidate 10 servers onto 1. One of the servers was VERY storage intensive but CPU cycles remained very low. This was due to a specific application that didn't like Storage latency.
But the decision that was made was to deploy these VM's onto an iSCSI connected RAID 5 array. This is a S/W iSCSI initiator connected storage device. You have 10 VM's and your storage sensitive VM is having massive issues. So what do you do? Where did it all go wrong. The planned storage implementation has enough disk space but the IOPS profile and connection choice is all wrong for the intended purpose. So how can it be fixed the budget is spent and the storage doesn't support online RAID level migration along with the fact that even if you could migrate the RAID level you would be completely out of storage. What can be done and how do we do it?
Well that's a case for another post. But if you ever end up in a position like this it's going to be a tough spot and it is possible that there might not be a good answer.
Talk to you tomorrow and I'll post about this again.
No comments:
Post a Comment