Automatic request analyzer for QoS enabled storage system
Svetlana Lazareva, Ilia Demianenko · 2015
The advantages of Shared Storage can be diminished when heavy applications devour most of the storage resources. With traditional QoS, any minor change in storage usage leads to either inefficient use of resources or the need to change QoS settings manually. It is essential that business-critical applications be attended to first. Through the present approach, storage will manage QoS settings efficiently so that the most critical applications reach their proper performance level, and resources are not wasted for less important tasks. A storage administrator prioritizes applications according to their level of importance. The intellectual module analyzes requests sent from clients by means of supervised machine learning techniques based on the well-known ensemble of decision trees Random Forest, identifies clients that use business critical applications and optimizes QoS for the clients. When data flow within the storage infrastructure no longer fits the pattern, the algorithm alters QoS settings accordingly and continues to monitor activity of clients.