Fibre channel continues to be the protocol used by most virtualized server implementations. In fairness it was the only supported protocol by VMware for the first few releases. It has with it a long history of use, with that comes an answer to almost any question you might have because of the broad support base. As we discussed in a previous blog "NIC QoS" there is a growing need to provision bandwidth at the virtual machine level. Fibre channel suppliers are moving quickly to provide the ability to prioritize storage I/O bandwidth on a VM by VM basis. Additionally the fibre channel community is pushing the adoption of FCoE, which may be the ideal protocol/interface for large server virtualization deployments.
iSCSI is the IP block alternative to fiber channel. While some in the iSCSI community will disagree with me, it is an ideal protocol for first SAN deployments, especially those motivated by a server virtualization project. Most SMBs are going to have an IP infrastructure first, and it is easier to tap into that infrastructure. iSCSI can deliver most of the capabilities of fibre channel, as long as those environments don't need the bandwidth throughput of fibre.
iSCSI also may have a cost advantage if the environment can use software based iSCSI initiators which allows the customer to use standard ethernet cards. In many cases, customers will deploy iSCSI on a separate network, which eliminates some of the cost savings but does provide a dedicated storage I/O path.
Finally some server virtualization solutions support NFS as a protocol to boot VMs from. First, even if your a Windows only environment, you can boot a Windows VM if you decide to use NFS. The upside here is that this also means you can use a NAS, which means it runs on standard IP like iSCSI. It also means that you can manage much of the virtualized environment like you do a file server. This can make the learning curve even easier. The downside is some of the advanced functionality provided by server virtualization software is not always supported and you can not boot the virtualization host from NFS. This means you are going to have to mix in one of the other protocols.
Which one is for you? It comes down to a decision of what protocol will give you the performance you need at the lowest cost and easiest manageability. In our next entry will discuss storage performance and what that really means to you. For now conventional wisdom is that fibre is the fastest protocol with the lowest CPU affect and iSCSI and NFS are tied for second in almost every test.
In the end the protocol you are most familiar with has a big advantage in your consideration and most protocols can be tuned to deliver much of the performance you need, although be aware that the tuning can drive the cost of the solution up. Manageability of the storage infrastructure is going to be affected by your prior familiarity but also by the level of training and tools that your supplier and organization will invest in you to learn a new infrastructure.
Track us on Twitter: http://twitter.com/storageswiss
Subscribe to our RSS feed.
George Crump is founder of Storage Switzerland, an analyst firm focused on the virtualization and storage marketplaces. It provides strategic consulting and analysis to storage users, suppliers, and integrators. An industry veteran of more than 25 years, Crump has held engineering and sales positions at various IT industry manufacturers and integrators. Prior to Storage Switzerland, he was CTO at one of the nation's largest integrators.