The ability to quickly resume operations after a severe disaster is critical for organizations of all sizes. Fortunately, disasters don’t occur every day, in fact they are actually pretty rare. It is their severity that makes organizations plan for them. The request to improve application response time on the other hand, is a daily demand, as is the constant pressure to reduce costs. The recent Amazon Web Services outage also taught us that disasters are often caused by innocent human errors that the infrastructure needs to be prepared for. DR Ready Storage needs to meet the day to day demands as well as help the organization withstand the worst case disaster.
The Problem with the Just Throw Flash at it Mentality
The first response to a request for more performance is to leverage flash media. But flash storage, without quality surrounding technology, won’t perform to its full potential nor will it be used efficiently. The software that drives the storage system should be flash aware so its performance can be fully exploited. It should also provide detailed analytics so performance sensitive applications are given priority. The problem with most storage analytics, if the storage system even provides them, is they are not granular below a volume or LUN level. This means in a virtualized environment it is impossible to differentiate and prioritize specific VMs. Deploying flash for the sake of just a speed increase by itself is like upgrading a race car to Formula 1 hardware standards but keeping the driver maneuvers at amateur level.
Another storage infrastructure challenge is managing growth. As the environment continues to scale most environments will need to expand their storage systems to meet either a need for more capacity, performance or both. Scale-up architectures require that the system’s controllers receive an upgrade, or add another unit, increasing management expense. Scale-out system address the scale-up shortcomings by aggregating capacity and performance from across multiple nodes but these systems often require three nodes to get started, tightly coupled federation among the nodes via a complicated clustered file system, plus they often don’t fully utilize the capabilities of each node, lowering efficiency.
IT professionals should look for storage systems that “scale-right.”. Meaning they can purchase the product as a scale-up system – one node to solve a specific problem – then gradually add nodes as more performance or capacity is needed but have those nodes loosely coupled, scale independently from compute, and all managed from a single management interface. Once again VM granularity comes into play here, facilitating the ability to move VMs between storage systems, automatically and non-disruptively based on policy.
A Role for Hybrid?
While the primary data center is becoming increasingly all-flash, there is still a role for hybrid storage, and there are advantages in working with a vendor that can provide both. A hybrid system is an ideal target for the DR Ready storage system, especially if the target’s sole role is disaster recovery. In this case it may make sense to use a hybrid system, with a small flash tier and a large hard disk-based tier. If there is a failure mission critical VMs, which will be recovered first, will automatically be promoted to the flash tier, ensuring excellent performance, even in the recovered state. One thing to note, the key factor here is nothing nostalgic about spinning disks vs. solid-state arrays. It all boils down to the balance between performance and cost, which take on different profiles, and when the crossing point might occur.
Recovery from a disaster is, of course, critical and DR Ready storage makes that so much easier to achieve. But at the same time, organizations have to deal with the day to day demands of the business and near the top of the list is the ability to provide excellent application performance and to be able select a storage system that can grow as the organization grow. However, the art to balance the performance requirements and disaster recovery needs should not be limited to just compromises. DR Ready storage should take into account the moving cross points of performance and cost profiles, and accept the change as the constant norm.
Sponsored by Tintri