Amazon’s Flexible Payment ServiceI totally <3 Amazon and now those Seattleisians have brought EC2 out of private Beta and into Public Beta! That’s right now all developers can sign up for their Compute Cloud goodness. I’ve had an opportunity, squandered of course, to play around with it. I hope to one day corner off some time to set something up on one of those bad boys. If you want to read about the awesomeness of EC2 check out this interview I did with GoPlan about their experience with it.

What’s even better news is that probably the biggest problem I saw with them has been addressed. They’ve added in new instance types, so instead of each compute unit always being a kind of underpowered server, they’ve added a “Large” and “Extra Large” instance sizes. That’s right, now you can supersize your compute unit. It’s all well and good to scale out by adding more machines, but I mean, at some point you just want fewer larger instances. Beyond that some uses really prefer to be on one beefy machine (like databases).

The instance types basically simulate a 1, 4 and 8 core machine. The RAM scales up, as well as moving from 32 to 64 bit and, strangely, increasing I/0 performance. The cost scales linearly with the number of cores but with the move from 32 to 64 bit and the better I/O it would seem to me that the larger instance types would be a better buy (if you need the horses) than several smaller ones.

Here’s what I wish that I don’t think is in the system right now… if you could up/down grade instance types. That is, the paradigm right now on EC2 is scaling out by adding more instances and then dropping them if you no longer need them. How much more convenient would it be if you could simply turn your 1 core machine into an 8 core one with 8x the memory? And then drop it back down! All in less time than it’d take you to go to a physical box and add more RAM. That would be pretty unbelievable.

Now that I’m mouthing off about things, another really great feature would be the ability to configure and mount an S3 based partition on the machine. That way you’d have the impermanent local storage you could write to, but also a super easy way to write to the permanent storage that S3 provides. It’d use all the tools one is generally used to using to shuttle files around, it’d be kind of like NFS. Now that’d be the bomb.

At any rate EC2 and S3 are grid computing done right. Amazon has always gotten it. And this just goes to show you.

← newer Apple SDK for iPhone (and iTouch) by February  ↑  Amazon’s One-click Patent Squashed older →

TwitterCounter for @nybble73