Short Update on Hardware Calculator for OpenStack

During the last couple of weeks we have been processing the feedback that we’ve collected from the initial launch of the Hardware BOM Calculator. Several vendors of both server and networking gear have contacted us to share their own recommendations on how to update our formula to be more accurate and we really appreciate the encouraging feedback so far.

In this interim release we’ve added the following functionality:

  • Information about Dell PowerEdge C Servers and about HP ProLiant Rack Servers product lines.
  • We began to filter all configurations which don’t have 10GB+ NICs (we don’t think that it’s a good idea to run cloud with 1 GB between server nodes).
  • We’ve polished UI a little bit, inspired by this article.

We’ve also realised that different configurations can be good for one thing and bad for other thing. This is not something that we can easily quantify, so we’ve decided to survey both vendors and those who are running existing OpenStack-based clouds to get their input and make this information publically available through the calculator.

In the upcoming release we plan to add the following features:

  • Tagging for servers (“good for” and “bad for”). Tags will represent qualities like density, performance, fragility, etc.
  • Recommendations on specific servers (including “Recommended by Mirantis” and “Recommended by vendor X” with experts’ comments).
  • Configurations for controller nodes (both non-HA and HA mode).
  • (most requested feature) networking gear considerations.

3 responses to “Short Update on Hardware Calculator for OpenStack

  1. One thought, I just tried putting in some details for the cluster I’m building, and it recommended either single servers, or pairs of servers.

    It might be helpful to be able to specify what sort of redundancy one expects. For us, we’ll be doing the storage using commodity hardware and redundancy in data storage is a must.

    Thus we’ve taken the option to move the back-end storage into dedicated boxes with Ceph, have a few dedicated management nodes looking after core services, then compute is a farm of relatively cheap (~AU$800 a piece) compute nodes.

Comments are closed.




Mirantis Cloud Platform
Automate Upgrades with Mirantis DriveTrain
ONAP Overview