Hacker News Reader

Comment by seunosewa | Hacker News Reader

So their "no noisy neighbours" claim is essentially incorrect. (Poor IO due to shared storage is the main performance issue with virtual servers.)

View on HN

When you use a remote blockstore like they are as your storage, you also generally are always booting via PXE, and so far have a provider controlled kernel.

I used to work on Rackspace OnMetal: http://www.rackspace.com/cloud/servers/onmetal

https://journal.paul.querna.org/articles/2014/07/02/putting-...

In OnMetal this not the case -- you get local storage, you have your local kernel that you can re-install or do whatever you want to -- and if we have a control plane outage, you can reboot -- there is no dependency once your server is created on a provider hosted PXE server.

OnMetal isn't cheapo-ARM servers, but it is possible to build a "bare metal" cloud, which is truly bare metal, like you would normally get, and no remote block stores for your root FS.

Do you think it would be feasible to provide a bare-metal cloud made of cheap ARM or Atom servers while giving each server its own independent SSD?

Yes, most definately.

OpenStack Ironic[1] can do this today, given you had right hardware. Making that hardware "mutli-tenant" is not a trivial project, mostly because the firmware stacks are still very closed, but it can be done. Top of Rack switches are also another area of complexity/pain, but also feasible to overcome with a little engineering.

[1]: OpenStack has its complexities, but the OnMetal project has upstreamed everything feasible we can.

These folks seemed to tried that approach, but gave up on Ironic for a number of reasons:

https://www.packet.net/blog/how-we-failed-at-openstack

True; I think it is reasonable project to take on if you are willing to staff developers... if you are trying to build a "bare metal" public cloud, having developers on staff is going to be a prerequisite for quite some time going forward.

Reply on news.ycombinator.com