Method for enhanced application performance consideration in high availability storage architecture like HyperSwap in network congestion and method thereof
Publication Date: 2018-Aug-10
The IP.com Prior Art Database
Disclosed is a method and system for providing an efficient data management of high performance storage tiers in multi-tier architecture for active-active recovery storage wherein two copies of a virtual disk are kept at two different sites within a single cluster by making it aware of direction change of the clustered system and other mentioned techniques.
Volume level IO throttling mechanism at storage controller has a map associated which records the IO operations and, based on IO performed, it detects whether the throttle limit is obeyed or not. In case it is detected that the IO is exceeding its throttle limits, then delay in IO processing is injected by throttling demon to control number of IO operation (same applies for bandwidth based throttling system).
As the IO processing delays are injected, the time required to complete the IO operation increases hence more latency. In the active-active high availability (HA) configurations, the host is connected to both the sites which are designed to provide HA solution. In normal circumstances, host performs IO on primary path detected by ALUA (Asymmetric Logical Unit Access) settings. Due to this the IO expectation of primary and secondary site are different, and hence different throttle limits can be configured to the pools on these two sites. At the time when hosts start sending more IO requests over non-preferred path due to network congestion or other network/host issues, storage cluster detects that preferred path is not active (due to say network unavailability, congestion etc) and switches the direction of the active-active relationship.
Once the relationship direction is changed, the method and system interacts with the direction change mechanism using an inbound or out of bound API (Application Program Interface) to replace the throttle values of secondary pool by the throttle values of primary volume copy which allows other site to receive and operate application IOs with SAME rate as it was before direction switch. If the system can afford the increment of throttle value of the secondary site because it is now primary serving volume copy and all the READ-WRITE IOs are served by this volume, then the throttling limit of both the volume copes across both the sites are fetched using storage API based system and will be switched to serve the IO.
The method and system also provides a way with an inbound or out of bound API where the throttle values are being identified for HA volume copies for primary and secondary copy which needs to be fetched from both the sites of storage system which can be part of single of multiple storage cluster entities. And overrides the same values upon each direction change of the relationship which ensures for the same throttle value each time for primary and secondary volume irrespective of the name and location of volume. This improves the overall time of replacing the IO throttle limits across the sites and hence will help improving user experience i...