Q&A with Dr. Allon Cohen of SANRAD - Page 2

By Dr. Allon Cohen (Profile)
Tuesday, November 29th 2011

VSM: How does this technology work?

AC: Significantly increasing ROI from FLASH within a virtual server host infrastructure, the VXL software provides caching services to multiple VMs over the host FLASH resources. Deployed as a virtual appliance on the virtualization host server, it distributes the FLASH resources on demand across the application virtual machines to maximize performance of key applications, such as Microsoft SQL, VDI and Microsoft Exchange.

Our VXL software offers an advanced policy-based caching technology. The caching efficiency increases over time as the algorithms learn data access patterns of key applications and customers.

VXL’s caching algorithm provides application-optimized strategies to decide what data to cache on the FLASH. The algorithm is transparent to the virtual machines, which are exposed to an accelerated virtual volume. With caching over highly available mirrored volumes, it ensures the same amount of FLASH is used to accelerate both copies of the data, doubling the efficiency of FLASH utilization. Meanwhile, no guest virtual machine agents are required due to support of nearly all modern operating systems including all variants of Windows and Linux servers via optimized virtual iSCSI connectivity.

VSM: How is it being used “on the ground”?

AC: In addition to maximizing FLASH utilization in general, VXL software has demonstrated significant improvements in performance of both Microsoft Exchange and SQL.

On Exchange, the combination of VXL acceleration and host-based FLASH has shown a 6.3x increase in transactional IOPS supported. Transaction IOPS for the test cases increased more than 6x (from 170 IOPS to 1083 IOPS). This increase in IOPS was accompanied by a large decrease in the DB read latency (from 20.3ms to 6.7ms) thus reaching access times 3x faster than the Microsoft usability requirements. Meanwhile, on SQL, using standardized queries based on the TPC Benchmark™H (TPC-H), the VXL decreased processing times of individual queries down to only 13% of the original processing time, a more than 7x increase in performance.

VSM: Where do you recommend FLASH to be “placed” to ensure maximum efficiency?

AC: We recommend that cache be placed as close to the virtual machines as possible, while retaining easy and efficient access to all the virtual machines requiring acceleration. This can be achieved either with host-based FLASH accelerated with a VXL virtual appliance or with network-based FLASH on physical appliances between the serviced hosts and any underlying storage arrays. That way, hot data is dynamically cached as it is transferred to the virtual machines and is ready for fast retrieval. Because the cache is distributed by the virtual appliance, it can efficiently service multiple virtual machines.

VSM:  How is your technology compatible with the cloud?

AC: Our technology is complementary to existing hypervisors, servers and SAN hardware. It increases their speed while ensuring high performance for key applications. Our solutions enable IT departments and cloud providers to deliver 'just-in-time' FLASH performance without needing to invest in FLASH hardware across the entire server infrastructure.