The second type is the loosely coupled cluster where files stay intact and reside complete on one particular node in the cluster. Typically additional copies are written on other nodes for data protection. A master metadata index knows what nodes each file and its copies reside on. The user is simply pointing to a master mount point. This type of storage, while it can perform well, does not tend to get slotted into a high performance environment. Unlike the tightly coupled systems described above the performance of the node you are accessing is the key factor in what type of performance you get. It does however seem to be received well by cloud storage providers that are looking for competitive prices and capacity.
Both of these configurations offer a pay as you grow type of scalability that sounds appealing. Scale out storage makes a lot of sense for a wide range of applications but have done particularly well in file centric environments as well as storage for server virtualization.
A weakness of scale out storage may be the node itself. These nodes are typically off the shelf, 1U or 2U servers designed to hold storage. A simple configuration might be a two processor box with two or four cores and 4 drives for storage. As the Intel and AMD roadmaps continue to roll out we are seeing faster and faster CPUs not slower ones. Also the capacity per drive is becoming larger and larger. The concern, especially in the tightly coupled configuration, is too much CPU power per node and too much capacity. While you may be scaling processing power with storage capacity you may only need to scale capacity and you may end up with a lot of near idle processing.
As we discussed in our recent article "Top Criteria For MSP Storage" data centers are looking for ways to increase floor space efficiency. That means that every rack unit used has to provide the best possible balance of capacity and performance, something that scale out storage may struggle with as processing speeds continue to increase. There are steps that scale out storage solutions can take to increase this efficiency level. This also gives scale up storage solutions an opportunity to mix in some scale out techniques into their scale up systems to create that balance. We will explore how both sides of the aisle may address this in upcoming entries.
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.