Amazon EC2 etc

[Steve21]

Hey Guys,

 

Bit of a silly question but has anyone dealt with services such as amazons servers much? (Expanding servers based on current connections etc).

 

Looking at setting up a bit of a side project, and seemed a nice idea to be able to use one of these companies, rather than a dedicated server, if it means anytime I need more processing/space etc etc it'll just automatically add it for me.

 

Few silly questions,

 

a) Can you run anything on these servers? Do they act as normal windows/linux etc boxes? Confused about the whole running thread/states etc it mentions on the sute

b) How do services like databases etc work across multiple "boxes", will it just replicate them, or will these stick to one "box" just the other processing get spread out?

c) Does the expansion/shrinking happen automatically, or would you need to request it to happen at certain times? e.g. If 1million suddenly hit a page, will it grind to a halt until you manually ask it to inrease space?

d) Is there anything you can't run on the boxes (legal stuff ofc ) but in terms of connections/ports/blahblah?

e) Any downside to using these? Assuming it'll work out dearer for small processing, but better as it grows?

 

-Edit f) as another point, Can you really only access both EC2 and rackspace cloud over SSH/terminal? Or once you set it up can you remote in as normal?

 

Just looking at options, and getting bogged down in the millions of pages of EC2 documentation Can't decide if it's the right way to go, or dedi server better.

 

Thanks,

Steve

Edited June 16, 2012 by Steve21

[dhicks]

has anyone dealt with services such as amazons servers much?

 

I've investigated it in the past, and I think I understand the basic idea of how Amazon's cloud services are intended to operate, and it's important to note that Amazon's definition of "cloud" is probably different to some other providors. A virtual machine in Amazon's cloud starts up with no dedicated storage and is expected to have a mechanism to connect to appropriate storage nodes. Pre-packaged solutions exist so that you don't have to create your own from scratch, but Amazon cloud processing nodes aren't the same kind of computing node as the virtual machines you probably run on your local school system. Amazon's definition of "cloud" strikes me as the proper one, and other providors are simply offering server VMs that happen to run remotely.

 

If you want a practice, I think Amazon now offer a basic starter account for free so you can get your head around how this works. Also, you can install local cloud systems that are compatible with Amazon's API, so you can move processing nodes between the two.

 

How do services like databases etc work across multiple "boxes", will it just replicate them, or will these stick to one "box" just the other processing get spread out?

 

That is for you to decide. If you're used to SQL databases then it's probably going to be easier to assume that you'll keep your database on one VM and add processing nodes as needed, with any scaling of database capability being done by assigning more processors and RAM to the central database VM. For a standard page-serving website that seems to be the agreed way of doing things, that's why you can see database servers costing £50,000 upwards - it's still cheaper to pile resources on to one chunk of hardware than pay people to redesign your database (or design a non-SQL database in the first place). If you have something more fancyily computational in mind then you'll hopefully have an idea of what you're doing anyway, but I'd have thought it'd take a while before a (reasonably well-written) small "side project" hit the point where it was using more database processing power than is available via a single server.

 

Does the expansion/shrinking happen automatically, or would you need to request it to happen at certain times?

 

If I understand correctly, you can add / remove nodes manually or through an API controlled by whatever software you write, so you could have a pretty simple script watch out for spikes in processing, storage or bandwidth and add as appropriate. Bear in mind that those extras cost you money, and a bug might result in more resources being added than you intended, so you might want to keep a very close eye on any automated process and build in some kind of maximum protection.

 

What sort of project are you aiming for?