It’s not often that storage gets exciting. We have seen a lot of interesting disruptions in the storage field over the last couple of years with new hardware, and new software, which is designed to take the ever-present storage environment to the next level.
As flash inside the primary array became a common approach to add some tiering capability, the game changed a little. Along came all-flash arrays, which upended the concept of tiered storage by offering that a primary storage device can be all flash. This is interesting, but also wrought with challenge when it comes to pricing.
No matter how you slice the dollars per gigabyte numbers, all-flash storage has quite often proven to be more expensive in practice for many organizations. Some software has appeared on the landscape to attack this problem with server RAM, or with host-level flash cache. These have had a positive effect, but have also led many organizations to still dig in their heels to move towards this approach.
Quite often, the server-side flash options prove to be too expensive depending on the physical topology. It also opens up a future dependency on both software and server-side hardware. Hyperconvergence and hybrid approaches are interesting, but the barrier to entry still seems high depending on the implementation.
This led me to Cloudistics who have taken a new, and interesting approach to the very real challenge of accelerating storage, and reducing latency in your SAN environment.
I recently spoke with Cloudistics CEO, Najaf Husain. We talked about the very interesting opportunities in the market today around storage acceleration as well as lots about the important need to provide a disruptive approach without requiring significant change in the data centers of organizations today.
Wait, what? Disruption, without disruption? Exactly! Organizations are often wary of making moves into technology adoption that require the upheaval of physical environments, and as the all-flash, and software acceleration party is just getting started, how will Cloudistics take a different tack?
The Cloudistics Turbine Approach
Cloudistics Turbine is built to accelerate your SAN storage environment on VMware vSphere ESXi, Microsoft Hyper-V, KVM, or Citrix hypervisor platforms.
Rather than replacing your existing storage hardware, the Cloudistics team lets you quite simply drop in a Cloudistics Turbine appliance, map the current storage using the Turbine environment, and repoint the hypervisor to the new Turbine accelerated LUNs:
The resulting environment also gives you the flexibility of running as a write-back cache versus the traditional write-thru cache which requires a write confirmation back to the source storage platform before acknowledging the commit:
By embracing a sequential bypass capability, Cloudistics also lets the large sequential writes go straight to the SAN. This is done by the Turbine engine, which selects the cache utilization, based on disk load:
With a nicely designed UI and clean UX, Cloudistics has done well to present a clean view of the environment, all with a simple HTML5 interface:
Cloudistics Redundancy and Resiliency Approach to Safety
While the approach of injecting itself in the data path can often seem difficult for some to embrace, Cloudistics have provided a fully redundant physical and software platform.
Inside the physical Turbine appliance are a fully redundant pair of 1U units. Not only that, but each unit contains redundant power supplies, NICs, and flash arrays, with a RAID10 deployment to ensure both performance and resiliency of the file system across the platform.
NICs are bonded and provide high speed access to the environment on the four 10 GbE connections. Writes are done at 40 Gbps and reads are done at 4×10 GbE.
For fans of Infiniband (me included!), Cloudistics has used Infiniband to link the two units within the chassis to provide a high-speed channel between the devices.
I did ask about the failure risk introduced by being added to the data path. Cloudistics provides a storage LUN mapping which is identical to the existing mapping with a .TURBINE added. If there was ever a situation where the environment became physically unavailable, the hypervisor could attach to the original storage LUN presentations directly. Only in-flight data that was not able to have a completed write to the back-end storage may be lost.
Given the speed of the writes due to high-speed network connectivity, and the redundancy on the Turbine appliance, the risk is near-zero and would be no different than the risk of the originating storage LUNs failing. In other words, if something goes wrong with your Turbine environment, it is probably a data center wide interruption, which means that everything will need to be attended to.
Fighting the Cost Challenge with Storage Acceleration
Boasting a $0.35/GB price, the Cloudistics team have come to the storage acceleration market at an unparalleled price offering. There will be some questions about the ability to achieve that full per GB value across the entirety of the environments that utilize SAN storage, but compared to the all-flash offerings in the market, this is clearly a big boost in value for the purchase cost.
Operationally, there are lots of benchmarks that have been provided from openBench that show good results on SQL platforms, one of which illustrating a massive boost in density while greatly reducing the latency and queue depth on the platform:
Find out More About Cloudistics
To find out more, you can go to the Cloudistics website to get the openBench report and to contact the team to request a demo.
Najaf is a very dynamic individual, and his team has a storied past in every area of technology. It’s definitely going to be interesting to watch the Cloudistics approach to storage acceleration, and to see how companies can embrace their Turbine product suite.
Seeing is believing, so I highly recommend that you take a demo to see for yourself!