Thread

Hey guys, I’ve recently been using Amazon Elastic Compute Cloud (EC2), mainly to see just what it’s capable. In terms of LAMP development, it’s fantastic, and very cheap.

Has anyone else tried using it yet?

Elliot

It looks good, im not sure how you use it though. But the prices are cheap.

EDIT: just read your post, looks brill, is there a way of running it without ssh though?

is there a way of running it without ssh though?

No, not really… once you’ve setup your server (using SSH), you can save the current setup and use it again for other servers.

So, if you’ve installed an FTP server, or setup automatic svn updates, your new server should be ready to roll instantly.

But, if you’re not using SSH, I’d recommend getting into. It’s become almost essential to the way I work!

One of my favourite features is being able to request a static IP for your account, then instantly map that to any of your servers. And, you can change the mapping instantly.
So if your server is the ‘small’ one, and you need to scale up… create a ‘large’ one with the same config, them just push your static ip to your new server.

Job done,

// offtopic

Elliot, I like your new website design.

Can you also have multiple servers tapping into one S3 ‘bucket’?

I’m going to have to learn ssh.

WackyWebs, ssh is really easy to use. If you can do well with LAMP administration, ssh won’t stop you.

Can you also have multiple servers tapping into one S3 ‘bucket’?

I’m going to have to learn ssh.

Yes… you have persistent storage now.

So, on server 1, you create a persistent storage volume.
On servers 2,3,4,5… you get the idea… you mount that volume.

So, the ideal situation would be to have your public_html/ directory mounted by all your apache servers.

Then your master MySQL server performs the writes…
your slave MySQL server performs the reads…

And you take regular snapshot backups of your database to a persistent volume.

Any of the servers can go ‘down’ at anytime, and the data will never be lost.

However: Data stored ‘locally’ on a server is non-persistent. If the instance is terminated, the data is lost.

So if i shut down all my server (instances) then it would still all be there on the s3 service.

I’ll explain like this…

on your servers… you’ll have a standard linux directory structure.

and data stored in /mnt/* is data on an external device (short for mount)
so, if you mount your persistent storage on /mnt/my-groovy-data/

Anything saved in there will be persistent. Saved to S3 (but accessed as if a normal mount)
So, you can terminate all of your instances… and when you fire up a new one and mount my-groovy-data, all your stuff will still be there.

I’m not sure if this feature is live yet, i think it’s brand new.

Thanks for the incite into this. Im just trying to find the best way to load balance 3 AMI now.

To start with, you could try using a simple roundrobin dns

I have been using Amazon EC2 since it launched (early beta user.)  The pricing is actually comparable to some of the more traditional VPS providers such as Slicehost or Joyent.  Their pricing schemes are just different.  At base, EC2 is cheap but when you start to rack up the bandwidth then it is not necessarily cheaper than your other options.

I’m going to have to learn ssh.

There is really nothing about it to learn.  SSH is just shell access.  If you know the Shell environment on the selected OS then you know how to use SSH.

So if i shut down all my server (instances) then it would still all be there on the s3 service.

No, if your instance dies, you lose your data (reboots don’t kill your instance though.)  Instance death is relatively rare but you need to backup.  S3 is available and there are a number of 3rd party offerings which setup an easy storage option with S3 but your instance does not have this setup by default.  For example, I use a ruby rsync clone which copies all my important data to S3.

There is a persistent storage offering in the wings but Amazon has said this will not be available until later this year.

By the way.  If you want to try out EC2, there are easier ways than reading through all the Amazon docs to figure out how to setup an instance.  I don’t bother with the manual setups anymore.  Rightscale.com allows you to use their system to control up to 10 instances and they make the process brainlessly easy.

Hey John… I didn’t realize that their persistent storage was not released yet
Dammit, that would make EC2 the best solution ever!

@WackyWebs…

You could also use a ‘4th’ front load balancing server using mod_proxy_balancer
http://httpd.apache.org/docs/2.2/mod/mod_proxy_balancer.html

A couple months ago we were re-launching a high traffic website on a new platform. We were concerned that the servers we had in production wouldn’t be able to handle the load the new site would bring with it, and we couldn’t purchase more hardware at the time. So what we ended up doing was temporarily scaling out the site to our production servers as well as several EC2 instances. EC2 helped us manage our server load, and gave us confidence, while we worked out bugs and tuning.

We ran the EC2 cloud for I think 2 weeks. Because of the traffic we got and the instance size we used (the largest), it ended up costing us several hundred dollars for that time period.

So, yeah, it can be cheap for small things, but just like all the other AWS products it can be uneconomical for larger, higher traffic sites/applications.