[Noisebridge-discuss] could you host my media wiki pages, 2-4 months?
jna at retina.net
Wed Nov 9 18:06:01 PST 2011
On Wed, Nov 9, 2011 at 5:41 PM, Ronald Cotoni <setient at gmail.com> wrote:
> Huh? much of ec2 is not backed up. No ephemeral storage is backed up, at
> all, ever. It is ephemeral. That is the default instance store type.
Sorry, I think we're miscommunicating; I mean the non-running instances
here, which exist in S3. Not the running instance.
> And you are absolutely correct. The days of playing with hardware is
> almost done except when it is not. When I think of a cloud like it is
> described, I pay for an instance that NEVER goes down ever without me
> configuring anything. It just does it automatically. I pay for ONE
> instance and whether the datacenter along with whatever state it is in
> suddenly is destroyed won't affect me in the slightest.
Ah, I guess I've been working at Twitter for too long. When we design we
-expect- the instance to go away. Except now, when we don't use instances
anymore and everything is on our own bare metal because we have serious
performance needs that cannot be met by Ec2 or any cloud hosting anymore.
This isn't to say that we don't farm out tasks to ec2 from time to time,
but we don't depend on it.
> The other problem is an ec2 compute unit is equal to a 1.8 ghz xeon from 7
> or 8 years ago. That is slower than an Atom. Those lowendbox.com ones
> are significantly faster than that at a fraction of the cost. If you were
> designing an app/service in the cloud manner properly, it woudln't be tied
> to ec2 or any other single provider.
You should see Adrian Cockroft's reporting on EC2 and the variances
in performance in the same zone. It's very interesting.
If you would like to discuss some systems architecture, perhaps we should
> have a meetup at noisebridge on this. There are others around but why not
> make our own. A lot of up and coming Systems administrators
That would be interesting; I currently lead the security team here at
Twitter and prior to that lead the Operations team. I've keynoted various
operations conferences (notably, O'Reilly Velocity for a few years) and
have much to share on building very, very large systems (and the small ones
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Noisebridge-discuss