Virtualization and the Storage Network

Share it on Twitter  
Share it on Facebook  
Share it on Linked in  

There's an interesting article by storage consultant Henry Newman on Enterprise Storage Forum explaining in detail what I've been trying to get across for over a year now: too much server virtualization too quickly could wreak havoc on your existing storage network.


Newman spells out very clearly the dangers of trimming a 1,000-server farm to, say, 100 without upgrading even a relatively fast 1 Gb Fibre Channel infrastructure. Assuming you're using dual-port HBAs, you're faced with a 10-fold decrease in available ports, which means that even if you upgrade to 4 Gb Fibre Channel, you're still seeing total server bandwidth drop from about 195 GBps to 78 Gbps.


Even worse, he calculates that this scenario will likely see a 3.6-times increase in IOPS, which will be tough to meet even with a 10-fold increase in both CPU performance and storage capacity.


Newman offers a number of methods to address the situation, urging that you pay particular attention to how your file system will allocate data from multiple I/O requests.


However, the underlying point is that things like IOPS and CPU performance do not necessarily scale 1:1. It's going to take a bit of work to figure out exactly how changes on one side of the data center will affect the others.