Best practices for Infinispan P2P cluster on AWS

[anton.kozik]

Hi all,

What would be your choice regarding JGroups configuration (discovery protocol, transport protocol) considering 16 node Infinispan P2P cluster (at least 10 GB JVM heap space per node) hosted on EC2 AWS with very intensive utilization of Map/Reduce?

Please share your thoughts, possible pitfalls for creating such a cluster on AWS.

Kind regards,

Anton Kozik

[sannegrinovero]

Hi Anton,

this template is included in Infnispan specifically for EC2:

https://github.com/infinispan/infinispan/blob/master/core/src/main/resources/default-configs/default-jgroups-ec2.xml

Essentially it suggests to use S3_PING for Discovery. Wathever you pick for Discovery, it only needs to access the backing storage during start/leave of a node so it will not have an impact on Map/Reduce performance.

[anton.kozik]

Hi Sanne,

thank you for your response! This is just what I wanted!

Cheers,

Anton Kozik