Flash storage is often the solution of choice for accelerating business applications. But with flash popping up all across the data center in various places – server-side flash, array based flash and even network based flash – organizations may be unwittingly creating a new and potentially very costly form of infrastructure sprawl – storage performance sprawl.
Watch our On Demand Webinar "Stopping Performance Sprawl Develop A Single Point of Performance Management" and get an exclusive white paper.
Storage Breaking Point
Many organizations have successfully consolidated their server environment through virtualization and are realizing the benefits of lower data center costs. But unfortunately, due to the ease by which virtual machines can be spun up and deployed, there have been unprecedented demands placed on legacy storage environments that have pushed them to their breaking point.
Enter Flash, the storage I/O “cure-all”. Many businesses have invested in Flash across every conceivable form factor to help drive storage performance for their physical and virtualized server workloads. The problem is different forms of Flash storage drag along with them different points of Flash performance management. An administrator may need to manage server-side Flash resources with one management interface and an array based Flash resource with a completely different interface.
In addition to creating management complexity, there is no way to share or load balance these resources as storage I/O conditions change in the data center. If IT planners thought that islands of conventional storage was a financial sinkhole, it’s an order of magnitude worse when high cost Flash capacity goes heavily underutilized.
Agnostic Flash Management
One way to combat storage performance sprawl is to integrate a layer of software intelligence that is flash hardware agnostic. This intelligence would leverage sophisticated cache algorithms that would move data, at a block or file level, between slow spinning disk and high performance flash. The software should also be capable of managing flash storage resources regardless of where they are located in the environment. But beyond the mere management of these resources, the flash software layer would allow organizations to leverage multiple forms of flash to better enable non-disruptive virtual machine migrations. For example, hot application data residing on a server-side Flash PCIe device could be mirrored to a Flash resource installed inside an alternate host or inside a hybrid storage array.
Implementing a single point of storage performance management through cache acceleration software technology also enables businesses to implement a tiered caching strategy. This means that hot application data can progressively be promoted up a Flash “food chain” based on the read and write I/O demand being placed on certain data sets.
For instance, a heavily accessed Oracle database table may initially get promoted up from spinning disk to a PCIe Flash device inside a host. Then as that data became increasingly hot, the caching software could promote it up into DRAM resources inside the host for even faster response times. This helps to ensure that the right data is promoted up to the right resource at the right time, without requiring manual intervention.
Adopting a single point of storage performance management delivers several other key capabilities which we cover at length in this on demand webcast that Storage Switzerland hosted with Intel. By registering for the on-demand playback, you will gain access to over 100 Storage Switzerland on-demand webinars.
Intel is a client of Storage Switzerland