In a conversation I had a few weeks ago with Pillar Data's CEO, Mike Workman, we discussed his recent blog entry on the <a href="http://blog.pillardata.com/pillar_data_blog/2009/10/autotiering-of-data.html">"Auto Tiering of Data"</a>. In this blog he brings up several important considerations as vendors and users begin to examine automated tiering. One I'd like to elaborate on is QoS in Automated Tiering.

George Crump, President, Storage Switzerland

February 8, 2010

2 Min Read

In a conversation I had a few weeks ago with Pillar Data's CEO, Mike Workman, we discussed his recent blog entry on the "Auto Tiering of Data". In this blog he brings up several important considerations as vendors and users begin to examine automated tiering. One I'd like to elaborate on is QoS in Automated Tiering.Quality of Service (QoS) of course has been with us in networks for a while. We have it enabled in IP routers, we are beginning to see it in network interface cards and now we are even seeing it in fibre channel networks with NPIV. Automate tiering now brings this to storage, where active data is automatically migrated up to faster storage, like SSD or even DRAM and inactive data is gradually migrated down to slower SATA based storage. This is a great start and compared to the alternative should provide a performance boost in almost every situation.

Merely identifying active data and moving it to a high performance tier is a brute force method that should eventually give way to a more intelligent method where possible. Just because data is active does not mean that it should go on the fastest and most expensive tier of storage. In most automated tiering environments the fast tier is going to be a finite repository and it may be impractical to keep all the active data on that tier. Really active but non-important data could prohibit slightly less active but important when accessed data from ever making it to the high speed tier.

What is needed is a more granular QoS capability in automated tiering systems. The ability to exclude or include data by type or location for example. Eventually these systems need to learn who the requester is. If it is from a small number of users on a relatively slow network connection, leave the data on mechanical storage. If the requester is an application or a high number of users then move the data up to the performance tier.

For the time being Solid State Disk and DRAM are finite resources. You want to make sure that you are not only putting active data on these tiers but data that can actually take advantage of the tier.

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.

About the Author(s)

George Crump

President, Storage Switzerland

George Crump is president and founder of Storage Switzerland, an IT analyst firm focused on the storage and virtualization segments. With 25 years of experience designing storage solutions for datacenters across the US, he has seen the birth of such technologies as RAID, NAS, and SAN. Prior to founding Storage Switzerland, he was CTO at one the nation’s largest storage integrators, where he was in charge of technology testing, integration, and product selection. George is responsible for the storage blog on InformationWeek's website and is a regular contributor to publications such as Byte and Switch, SearchStorage, eWeek, SearchServerVirtualizaiton, and SearchDataBackup.

Keep up with the latest cybersecurity threats, newly discovered vulnerabilities, data breach information, and emerging trends. Delivered daily or weekly right to your email inbox.

You May Also Like


More Insights