HN Explorer

The "Cloud": who's really using it?

4 points by acl ↗ HN
The cloud buzz has reached a fever pitch. So who is actually using cloud computing, and for what?

To clarify, I'm talking about computation (EC2), not file stores (S3). Furthermore, I am excluding traditional VPS providers (Slicehost, Linode, etc -- don't get me wrong I love both of them). So cloud services would include, but not be limited to: EC2, AppEngine, Rackspace Cloud/Mosso, Heroku type of services.

So, are you using cloud services, and for what? Your whole infrastructure or just part of it? How big are you?

I'm also curious about auto-scaling. Of course auto-scaling sounds amazing, but when I think of the hard problems that need to be solved during the scaling process, most usually aren't solved by bolting on more hardware (virtual or not).

So who's really using auto-scaling? Does it work for you?

6 comments

[ 259 ms ] story [ 105 ms ] thread
byoung2
I'm using Mosso/Rackspace Cloud for a web design business. I wanted clients to be able to scale up without hitting bottlenecks or affecting other sites on the server.

It isn't perfect, but it has a few huge advantages. The biggest plus is that Rackspace handles all of the auto-scaling, so it is plug and play in that sense. You code it and load it, and if you need more PHP power or more database muscle, the system takes care of it for you. Great for blogs that may hit the front page of Digg (www.techspyer.com is a Wordpress install of mine that's on Mosso hosting and it works for that application).

joevandyk
Why exclude slicehost and linode? they have APIs for starting/stopping slices.
acl
Just trying to put somewhat of a box around what the "cloud" is. Especially since we're getting (I feel) saturated with the term. I put Linode and Slicehost in a non-cloud category because you pay by the month and the options for programmatic instantiation are limted. But yeah, I acknowledge that the lines get blurry.
sanswork
I started on Amazon grew to something like 30 servers on there doing a variety of tasks in a few groupings(Server group A has 5 servers all the same + load balancer for example). Used scalr.net to manage it and thought it was great.

I've recently started moving everything over to GoGrid(20+ servers there now) since I needed to be on the west coast for latency issues. I'm not as big of a fan of their services but it's ok for what it is. I still have a number of complaints though not as many as last time I wrote about it.

I also have a hand full of servers with rackspace cloud for redundancy but they get very little use at the moment.

Also our whole external infrastructure is now in the cloud. We have one server with softlayer and a mail server that are separate. In house we have another 15 or so servers for log processing.

We use autoscaling or rather did when with amazon using scalr and once I get my puppet setup running fully on GoGrid I'll be developing one there too. The problems I face can entirely be solved just by adding more hardware since all my systems are modular so it works well for me but your mileage may very.

Scalr support autoscaling by things like load/ram/bandwidth as well which is nice since a lot of times you need to scale out for different reasons.

agmiklas
We're using AWS EC2 mostly because they're the only hosting provider we were able to find that offered high-speed connections between their data centers. We use this capability to synchronously replicate every transaction to at least two data centers. The plan is that we should be able to cut over to another data center without losing any committed transactions.
jhammerb
SoftLayer's bare server API (http://softlayer.com/resources_api.html) and monthly pricing have suited our needs nicely. We've seen seen two problems (a porn site in the same datacenter sometimes gets DDOSed and the firewall in one data center is remarkably flaky), and they've been remarkably quick to respond to each.

The amount of tooling around the Amazon ecosystem is nice, but we mostly just use it for automating builds.