Tony's Blog Bytes

A recent Forrester report - “Do You Really Need A SAN Anymore?” - challenges the value for having storage area networks (SANs).  The title suggests a bold thesis and implies a new alternative to the status quo. However, it falls way short of the expectations created by the title. 

If I read the report correctly all it is really suggesting is to use direct-attached storage (DAS) with RAID instead of SAN-attached storage and put more of the data protection and management functions at the application layer.  With all due respect - this is a lot of sound and fury signifying nothing. 

Hu Yoshida discusses the Forrester report in a recent blog (Forrester says you don’t need a SAN) analyzing the challenges of traditional SAN approaches - and there is a real discussion to be had on the subject but - in my view - the Forrester report doesn’t offer a realistic or practical approach to solving the very real cost and complexity of data center storage. 

Let’s say you agree that DAS is better than SAN.  You can use an iSCSI or FC storage system as a DAS system and leverage its tried and true capabilities. And today you can get a pretty inexpensive but intelligent storage system to do this.

Let’s say you like the idea of having more storage-like functionality in your applications.  Why can’t you still use your SAN to work with these applications?  The answer is you can - these two things are not at odds with one another and can be complementary. 

However, think about all the applications that have to incorporate storage capabilities.  Some will never be able to achieve this.  Some might do an okay job of it but still fall short.  Some apps will be downright awful at it.  And all of them will be different in terms of capability, performance and reliability.  Additionally, what about all the home grown applications - does your company want to take on solving the storage problem at the application layer? 

The report also points out that there are more storage-related functions being implemented at the database and host virtualization level and suggests that these are better places for them to exist.  But databases and host virtualization solutions are not “applications” (e.g. Exchange).  Therefore you would need to have two other categories - database-centric storage and virtualization-centric storage. 

Databases are highly integrated with certain applications and therefore could be a smart place to put more storage-oriented capability.  And some databases do have differing degrees of storage-like functionality but are still connected to SANs.  But what about applications that don’t use databases?  Again, these non-database applications would then each have to individually support these capabilities. 

On the other hand host virtualization has no visibility into the applications that live above it and is actually “virtual infrastructure” - so it shouldn’t even be considered a part of the application-centric storage discussion.  Having said that VMware is a long way away from having best-in-class storage capabilities.  For example, the VMFS file system only supports 2 TB - which has limited applicability.  It is great for a server administrator that wants to do some testing but not appropriate for many (most) production environments.  Additionally, much of the value of VMware, Hyper-V, Virtual Iron and other host-based virtualization solutions is created when you have a shared storage environment.  If you take away the shared storage environment then you have a single physical point of failure. 

Unfortunately I think that Forrester missed the mark by a mile.  While I agree that SANs can be complex and costly - the value of sharing physical infrastructure - such as storage - will always have a lot of merit.