In our last entry we discussed using storage services as part of the hypervisor in a virtual server environment. In this entry we will explore embedding those services as part of a SAN infrastructure itself. In this deployment the storage services that we have come to count on are essentially part of the SAN switch instead of on the storage controller.One of the advantages of putting SAN services as part of the infrastructure are that they allow an independence from the underlying storage hardware. Since the management software is on the switch itself there is less concern over whose storage hardware is underneath. This sounds great when you draw it up on the whiteboard but the reality is that this approach has been tried for several years now and it has seen limited acceptance by users.
The lack of adoption has come for a variety of reasons. First the timing may not have been right. This capability was first rolled out four or five years ago but was going to require that the switches which make up the infrastructure be changed out, or at least upgraded, to be able to support the new storage service capabilities. Infrastructure, as we have stated many times, is very sticky and does not tend to get replaced very quickly. Second the software that provided the storage software was not very mature compared to it's counterparts on the array controllers themselves. Finally, from a user perspective there was great comfort in having the storage controllers provide those services. Habits are hard to break, especially if there is not a compelling reason to do so. If you were an IT maverick and wanted to move storage services off of the storage controllers then you were probably going to look and one of the mature storage software solutions that ran on their own storage appliances, not on the switch.
Fast forward to present day. Many SAN infrastructures have been upgraded to the point that they can support greater intelligence in the switch, but while the software that would be embedded here is good, it is not as robust as offerings now available in storage controllers. Storage vendors did not stand still, they continued to advance their capabilities too. As a result there is still not much demand for storage services to be embedded at the switch.
Infrastructure vendors should continue to focus on more basic blocking and tackling of storage services. Functions like encryption and WAN connectivity are good examples. Extending that to capabilities like volume migration to aide customers when they deploy a new storage system could also be helpful. There is also potential in adding functionalities like continuous data protection capabilities. It seems to me that the switch would also be a great place to embed SSD caching techniques like those we discussed in our automated tiering series.
By taking a more basic approach to providing services, I think that infrastructure vendors can be more successful than they have been in years past. This allows them to leverage capabilities that where executing them on the switch gives them a distinct advantage. It also causes less conflict with the storage suppliers that tend to be the largest supplier of their products.
Track us on Twitter: http://twitter.com/storageswiss
Subscribe to our RSS feed.
George Crump is lead analyst of Storage Switzerland, an IT analyst firm focused on the storage and virtualization segments. Find Storage Switzerland's disclosure statement here.