Ask Your Question

Are there distance limitations in physical OpenStack architecture ? [closed]

asked 2014-01-15 23:50:15 -0500

epochuni gravatar image

We are getting ready to implement a very large private cloud with OpenStack. I am a hardware engineer focused exclusively on supporting the hardware of this OpenStack cloud. I was looking for documentation and / or advice related to distributed infrastructure. Our first severs for block, object.. etc. are being implemented in one data hall in a colocation site. We obviously expect to grow the cloud over time. The colocation site however cant guarantee us space in the future will be anywhere near this gear. Growth segments could be hundreds of meters away, but not more than say 300 meters in context of using multi-mode fiber with Ethernet or Fibre-Channel. Everything I have been able to read doesn't lead me to believe this will be an issue and having exact locality is not an issue. The only functions I thought could be affected would be cinder based functions as block storage can be finicky as distance increases. So the questions are

  1. Are there recommended or best practice principles around distributed pieces of the same private cloud ?
  2. If distributed are there specific settings such as a time-out value which need to be changed to accommodate this ?

Any help from the experts would be greatly appreciated. :-)

edit retag flag offensive reopen merge delete

Closed for the following reason the question is answered, right answer was accepted by smaffulli
close date 2014-01-27 17:42:56.970598


This sort of questions will lead to a debate which is not the sort of questions well served by this site. For debates please use the Operators mailing list

smaffulli gravatar imagesmaffulli ( 2014-01-24 19:33:55 -0500 )edit

2 answers

Sort by ยป oldest newest most voted

answered 2014-01-24 19:36:35 -0500

smaffulli gravatar image

Distance influences things. The current best practices are documented on the OpenStack Operations Guide and Cloud Administrator Guide.

edit flag offensive delete link more

answered 2014-01-25 02:19:00 -0500

tim-bell gravatar image

The CERN cloud is split between Geneva and Hungary (around 1,300 Kms). We use cells to maintain localisation of message queue traffic so that the Budapest centre has a cell with around 600 servers and the Geneva site has 3 cells, each with over 700 servers.

We have not seen significant impacts of distance using this approach although we are still working to tune ceilometer configurations as this can be quite latency dependent over the WAN lines and the keystone service is in Geneva.

edit flag offensive delete link more

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2014-01-15 23:50:15 -0500

Seen: 373 times

Last updated: Jan 25 '14