Sorry Bill, there is a new question burning in our minds today. There seems to be a lot of buzz lately about tiering your data storage and who can and who can not, why and how but not alot of people are talking about when to tier your storage. Netapp has indicated they are not as concerned with a tiering approach and this is true for the IBM XiV product. Others like 3par and IBM' SONAS has it built in for clients to move... [Continue Reading]
Sorry Bill, there is a new question burning in our minds today. There seems to be a lot of buzz lately about tiering your data storage and who can and who can not, why and how but not alot of people are talking about when to tier your storage. Netapp has indicated they are not as concerned with a tiering approach and this is true for the IBM XiV product. Others like 3par and IBM' SONAS has it built in for clients to move data from one pool to the next. But how does one gauge this old standard of giving the best to the most demanding and the least to the dregs of our storage footprint?
Tiering can be based on performance needed from the client/application or the length of time and frequency of use. Some vendors will come out and say we treat all data the same and can shift needed resources to more used areas of the subsystem. And some allow you to create pools of storage to allocate cycles for just that application. The main difference is what happens when a system is over subscribed, do you have a guarantee without the pools that your application will always get the set performance it needs. Archival tiering allows you to move data that has a lower frequency of being accessed to lower cost (large and slow) disk and then to tape. The movement from pool to pool is based on either rules or policies set by the administrator based on date or time. This is a bigger issue with NAS data than SAN due to the nature of NAS files.
An indication of when to tier is the size of your storage system. Is it worth to create three tiers of storage for a 5 - 10 TB storage system? Probably not and there are simple ways of isolating storage for higher performance that has been tired and true. If your storage doesn't have built in tiering, you can use isolation of drives to increase the amount of performance to an application. You can also use higher amounts of cache like the N series PAM cards. This can decrease the latency response time and improve your application performance by adding additional read cache.
A larger system that has 100 TB and up would be ideal for tiering based on performance. As your storage grows there is data that needs to be on fast disk and data that can live elsewhere. Think of your storage as a tool chest of wrenches, screwdrivers and sorts. As you get more tools in your tool chest, you will want to keep those used more frequently in the top where you can get to them quickly like the trusty screwdriver that does both Phillips and flat head screws. But the tools that used less can be in the shelves below, sorted by either size or frequency of use.
Tiering data may be important to you as your build out your system and maybe you need to implement it day one. With the growth in the digital media, whether you are taking pictures for a marketing campaign or producing a new digital movie, we will see data storage grow ten fold in the coming years. I suspect tiering will be needed more for these projects as their data plat form scales out quickly and the smaller storage units will be used as secondary units in field offices or retail stores. Either way, you will need to evaluate whether to tier or not to tier based on your storage needs today and in the coming future. Would Shakespeare believe in tiering? Only if it sold more tickets for his play maybe...