4 TB Ram server,architecture help



I am looking to index around 3TB of documents. I have been given an incredibly powerful server with 4TB of ram and 160 cpu cores. I want to make sure that elastic uses this hardware to its best.

My current layout is virtualising two elastic instances in kvm. The master node is using a tmpfs ram disk to hold the entire index in memory. The secondary node is using niofs on disk to account for when the systems are rebooted. Each elastic JVM is allocated 30gb of ram within their VM. Will it be as simple as that to get the best from elastic?

(Mark Walkom) #2

With that hardware, it should make a good difference!

(system) #3