Quantcast
Channel: Storage Expert Blog » Storage Services Catalog
Viewing all articles
Browse latest Browse all 3

The New Storage Services Catalog: Employing Storage Auto-Tiering and Policies

$
0
0

It has been in the works and long over-due and finally auto-tiering is here. It is perhaps one of the underpinning technologies, along with Virtual/Thin Provisioning of course, that change the science of Storage Management. Up until recently, Storage Service Tiers were build in one-to-one relationship with Storage Technology Tiers (see earlier blogs for the difference.) With the dawn of auto-tiering, the relationship between service and technology tiers becomes more obscure, yet, the ability to deliver on the described service tiers is higher, the efficiency gained is greater while the costs of doing so are lower.

While building the next generation storage infrastructure for data centers is a little more involved than before, it is simply an exercise of merging capacity engineering and performance engineering together to produce the appropriate mix of technology tiers required to meet the application needs. The intersection of capacity and performance engineering, which can easily be solved in a simple excel spreadsheet, optimizes your storage infrastructure and is likely going to save you several millions of dollars while keeping your customers happy as well. Add storage virtualization and call it a day (will talk more about storage virtualization at a later blog)

You cannot be expected to deliver the same performance and service guarantees for all applications (not unless you have an unlimited budget) but can provide customers with appropriate cost-effective service tiers to choose from. There is still a need for a Storage Services Catalog and the services defined and described are no different than before; they describe the peak input/output instructions (IO throughput) per application, data throughput (MBps) per application and expected response times (in milliseconds). What changes, however, is the reference architecture that adds a new layer of abstraction between service tiers and technology tiers, namely policies. Policies enable the delivery a high IO or data throughput without putting all applications on solid state drives (which can be very expensive) or even fiber-channel drives (that too can be expensive.) You can build your underlying technology with few simple technology tiers (e.g. an SSD pool, a Fiber Channel pool and a SATA pool). The appropriate balance between three simple pools of different performance capabilities in each frame can deliver the required performance without breaking the bank.

If you have not started down the path of policy-based tiering (using auto-tiering) you need to at least consider it and exercise in due diligence. The policies replace the old technology tier relationship to a service tier and make use of all technology tiers, in different combinations. For example, a Storage Service Tier 1 may comprise of up to 25% of capacity residing in SSD pool, up to 80% in a fiber-channel drive pool and perhaps 30% in a SATA pool. This forces 70% of the application to sit in faster (SSD or fiber-channel) storage at all times (which will reduce your response times.) Likely, your Storage Service Tier 4 may have a policy defining that up to 100% of the application can stay in a SATA pool while only up to 30% can be promoted to the fiber channel pool with no use of SSD capacity. There are of course few more parameters that will be considered which is the beauty of storage management science. Remember, you cannot go wrong building your new service tiers and aligning applications to them; even when you do, the fix is simple, fast and non-intrusive. Changing a misaligned application’s service tier may not even require a change window.

Building the technology to support this requires an understanding of the environment and either the possible concurrence of application IO and throughput or simply reading empirical data with peak front-end IOs and throughput loaded on different frames. This allows you to intelligently over-subscribe the SSD and fiber-channel pools and yet deliver peak performance at a fraction of the cost. It should be noted that over-subscribing the pools in this context is much different than thin provisioning over-subscription.

Whether you are aiming to contain your storage costs without containing performance or improve performance without blowing your budget away, auto-tiering is the way to go. Either way, your business customers will be happy if not ecstatic with their application performance (at least from the storage perspective) at the new price point. Remember to pass on the savings to them!



Viewing all articles
Browse latest Browse all 3

Latest Images

Trending Articles





Latest Images