A Method for Improving performance of Disk Intensive I/Os in High Availability Cluster Servers
Publication Date: 2016-Oct-06
The IP.com Prior Art Database
Disclosed is a method for Improving performance of Disk Intensive I/Os in High Availability Cluster Servers.
Insights that could be used to address the problem: 1. Is it possible to improve the performance of the I/O on Master node of a HA cluster, by making use of the resources of the passive nodes of the cluster? (1.a) Consider that the application I/O is disk intensive (examples: media encoding or editing large blocks of photos)
Page 01 of 4
A Method for Improving performance of Disk Intensive I /Os in High Availability Cluster Servers
Context of the problem:
A group of independent servers (nodes) working together as a single system, to provide high availability of the services/applications for clients, are known as High Availability cluster servers.
In case of the failure of a cluster node hosting an application, the application migrates to some other node in the cluster without any disruption to the access of the application by the end users, thus providing high availability of the application hosted on a cluster server.
HA-cluster can be 32 nodes, each of which either actively running an application (active) or waiting to take over when another node fails (passive). Failover makes a passive node to active node.
In a HA cluster server, the resource utilization of the passive nodes is negligible. A passive node's scope is limited to:
Node joining/exiting the cluster
New Master selection
Configuration changes (shared storage added/deleted)
Hosting the application and serving the application specific requests is responsibility of the Master node (active node) only.
Is it possible to improve the performance of the I/O on Master node of a HA cluster, by making use of the resources of the passive nodes of the cluster?
Consider that the application I/O is disk intensive (examples: media encoding or editing large blocks of photos)
Method for Improving performance of Disk Intensive I/Os in High Availability Cluster Servers:
For the disk intensive applications on the Master node, it is possible to use the primary memory/RAM of the passive nodes of the HA cluster as replacement to the disk access.
This is beneficial in 3 folds:
Accessing the RAM of a passive node via the cross cable of the cluster would be more efficient to the disk access via SAN.
o Since Crossover cable provides a direct connection (not being routed to the network), the speed must be much faster than the Ethernet speeds.
The processing power of a passive node could be used, which is more powerful than the controllers at the disk.
The performance of the RAM of the passive node would be more efficient to the performance of the disk storage.
Page 02 of 4
Working of the Solution Method:
In past, lot of work related to cluster server (not HA clusters though) configurations and global cache across the different nodes of the clusters is done.
Global cache mechanisms require hardware support
Global cache mechanisms require file system level support
Global cache mechanisms are talked in the context of cluster nodes sharing workloads (not the HA clusters)
There could be multiple possible ways of implementing the solution method described.
One easy and interesting way of implementing the proposed solution method is through the use of RAMDISK
Below is a way to implement the proposed solution method of using passive node's RAM for the disk intensive applications on the Master node:
Create a RAMDISK on a passive node (RAM...