The Cloud Won't Be Like ILM

The best thing about comments to blogs is that sometimes they require a response that is so long it becomes a blog itself, or in the case of DanB's latest response to my SNW Blog, it can become two or even three. It was a great comment and underscores some of the mistakes that we in the analyst/media make when it comes to a subject like cloud storage. Dan's first point and the focus of this entry was that clouds sounded like the next ILM (Information Life Cycle Management). By that I take it that he meant we are guilty of over-hyping, much like we did with ILM.

On the hype part, I will plead guilty as charged. I have been talking about cloud a lot but I would like to make a case that the hype is deserved and cloud is not going to be like ILM and fizzle out. Further, it is worthy of the large majority of the conversation it generates. As former Houston Oiler's coach Bum Phillips used to say "It ain't bragging if its true."

First, let's look at what went wrong with ILM. In my opinion the commotion around ILM was the introduction of first PATA and then SATA based hard drives. These drives dramatically drove down the cost of building large capacities of storage. The problem was at the time they were relatively slow and their reliability was an unkown. They quickly were segmented for use as backup and archive applications. Both use cases had what can only be described as bumpy starts. At the start, backup software did not do a great job of supporting disk, basically treating it as tape. Virtual Tape Libraries (VTLs), added cost and complexity to the equation. It really wasn't until NAS based appliances with dedupe functionality built in that the market really began to take off. That allowed the disk backup market to grow while complexities were resolved around VTL and eventually lead to backup virtualization.

On the archive side, vendors tried to wrap up the idea of moving old data to a SATA pool of storage. All done under the banner of ILM. I remember being involved in storage assessments, we could show that 8TBs out of 10TBs of data was old and could be moved to less expensive storage. Problem was that not many IT managers wanted to buy 16TBs plus (for growth) of SATA storage to move 8TBs of data to it when they were not really out of disk space on the original data. While 80 percent of their actual data was old, they also often had another 50 percent or more of free space that could still be allocated.

Second, in the early days of ILM we were without mature file virtualization and archiving applications and certainly the concept of automated tiering was unheard off. That has, of course, changed and the tools are now in place. File virtualization as we describe in 'What is File Virtualization' is very viable but vendors for obvious reasons are remiss to call it ILM anymore. These tools are here now for cloud storage, and it is well suited to backup and archive types of data. Data movement is a mature and now, if you want it to be, is an automated process. These tools alone don't make cloud storage viable. Cloud storage has more interesting business case and greater utility than the original ILM based strategy. All of which we will cover in our next blog.