The Smart Storage Hypervisor
Beyond Storage Hypervisor
ProphetStor Flexvisor® is the standard storage hypervisor that ships with ProphetStor Federator™ SDS solutions. Instead of adopting one of the many readily available storage hypervisors on the market, be it open source or proprietary, ProphetStor is determined to develop its own storage hypervisor as an integral part of the ProphetStor Federator™ SDS offering so it lives up to its “Prophet” name tag. We didn’t want to build yet another storage hypervisor just to virtualize the directly attached drives; we wanted the virtualized storage to be smart and adaptive so it can work with Federator to analyze the traffic pattern, predict the best-fit scenario for any timeframe, and automatically adjust the underlying storage composition (SSD+HDD) accordingly to maximize application performance and ensure a balanced storage infrastructure.
Scale Out with Commodity Hardware
Flexvisor transforms standardized commodity hardware into storage servers that provide enterprise-grade features such as pooled storage (eliminating the need to predetermine storage size), copy-on-right (ensuring data consistency), checksums (enabling data integrity verification and self-healing), rapid snapshot, writable snapshot, snapshot rollback, remote replication, volume creation from snapshot, and volume consistency group.
Elastic Resource Control (ERC)
ProphetStor storage hypervisor Flexvisor can fully leverage the installed server-side solid state drives to dynamically change the storage I/O throughput for an application. Take the above mentioned VDI situation for example, the application sees the same, say, 5TB volume, presented by Flexvisor during the two sample time frames, 5AM-7AM and 8AM-9AM. But underneath, the storage is accelerated by entirely different amounts of SSD cache.
During the peak hours, a much larger portion of SSD is given to the VDI volume to handle the high I/O demands. When VDI traffic decreases, the extra SSD capacity can be given to another application, say, ERP that can benefit from the boosted performance.
Flexvisor cannot only dynamically adjust the SSD/HDD ratio in a given provisioned volume, but also optimize the caching mechanism in the SSD. For example, “hot” data can be selectively kept on SSD to improve random read speeds, while random writes are always cached and serialized before they are written to the HDD. These functions are of vital importance to the general performance of a storage array and they can and should be easily incorporated into Flexvisor.
However, no matter how powerful and optimized a single array is, without Federator, the entire compute-storage environment cannot be optimized as a whole. Again, this is why ProphetStor must develop its own storage hypervisor so the entire federated storage system can act like a well-conducted orchestra.
Adaptive Control and Pre-Allocation
Major online retailers apply big data analytics to understand its customers’ behaviors and uses that information to recommend highly relevant products to the customers or even go as far as pre-shipping products that the customers are most likely to purchase to the nearest warehouse in advance, maximizing customer satisfaction and corporate profit at the same time. By the same logic, ProphetStor developed a proprietary traffic-modeling algorithm (patent pending) that utilizes fuzzy neural network technologies to analyze the applications behaviors in storage usage. The difference is, mega retailers have to analyze a huge amount of unstructured data while Federator only has to analyze a much smaller set of structured network traffic data, making its prediction much more accurate than Amazon’s. With Flexvisor, this crucial knowledge of the network usage patterns can then be used by Federator to intelligently and dynamically allocate, or pre-allocate, needed storage resources to areas that need it the most in real time.
Greedy Tenant and Noisy Neighbor
Federator allows the user to self-provision the storage but most users have a tendency to over-prescribe, grabbing more resources than what are actually needed. For example, if a user enters Federator’s control panel with building a web server in mind and three tiers of storages are offered, he would most likely pick the highest tier “just to be safe.” This particular mentality exhausts quality storage resources really quick. Another scenario would be that a user provisions a high-performance volume and actually uses its full capacity. The problem is, he’s watching streaming video, generally a low-priority task, and taking up precious network resources that could otherwise be allocated to other more important tasks. Federator’s monitoring and metering capabilities can analyze the traffic patterns and, when necessary, prioritize IOPS allocations on Flexvisor to feed mission critical tasks while starving low priority ones.