Need to provide server sizing specification for ElasticSearch

With the addition of ElasticSearch in 4.6.* by defualt this new software runs on the same server as Bitbucket Sever but there is no information about how much Memory, CPU, Disk, and Network resources are required to use ElasticSearch and Bitbucket on the same server.

Servers have been optimized to balance JVM and scm-throttle to get the most from the sever but without some guidelines for ElasticSearch this could bring the entire environment to a halt

Further with Bitbucket Datacenter ElasticSearch has to be run from a separate server. Enterprises have to plan far in advance to acquire the additional hardware needed and they need to know:

• How many CPU's should the server have?
  • Is Elastic search performance improved with additional CPU's?
• How much Memory should the server be allocated?
• How much Disk Space is Required?
  • Does the disk space have to equal the amount of disk space consumed by the total of all repositories

I was able to find this https://www.elastic.co/guide/en/elasticsearch/guide/current/hardware.html should this be considered the official source of ElasticSearch hardware requirements?

Reading this document I believe that is is saying:

• CPUs are not super critical
• 16, 32, or 64GB of RAM instances are common

Note: Because ElasticSearch knows nothing about Bitbucket and is used for indexing and searching of almost anything, there is no mention of hard disk capacity planning that is useful to Bitbucket Server admins.

If the above information is correct, it seems that even a small instance of Bitbucket Server on a Linux server with 8 CPUs, a Bitbucket JVM of 2048m, Git operations 8400m (if using the default scm-throttle), 16000m for ElasticSearch, and 2048m for base OS operations means that you need a minimum of 29GB of RAM.

Official guidance is needed