It's around $65 a month to run an Instance on their $10 Billion facility. Each instance is the equivalent of a system with a 1.7Ghz x86 processor, 1.75GB of RAM, 160GB of local disk, and 250Mb/s of network bandwidth. You can scale up to 1,000 or more servers in just minutes and pay only the minutes you use per server which is .10 cents.
MediaLayer, LLC - www.medialayer.comLearn how we can make your website load faster, translating to better conversion rates for your business!
The pioneers of optimized web hosting, featuring LiteSpeed Web Server & SSD Storage - Celebrating 10 Years in Business
I tested it for the first time yesterday and it is pretty remarkable. The main problem I see right now is that the cloud instances are designed around the assumption that they are temporary. They don't offer basic features like static IPs, or for that matter, the ability to have multiple IPs on a single instance.
And if you issue a shutdown command your instance (server) is gone forever -- data and all. So any service designed to use EC2 would best use the S3 storage service to permanently store data.
I would totally use it if they provided an easy way to load balance among the instances, and made it easier to expand and contract my compute cloud. My current application is completely scalable and I would love to have a working solution where I pay for extra instances by the hour.
As it is now, there is a peak time in the week were I need about 10 servers for a few hours. This would just be a couple of dollars with EC2, but, over a thousand dollars for real hardware.
However, right now, you can't rely on each instance, and it's manual work for me to add instances. There is no easy way to load balance among my instances, and instances are unreliable.
the unreliable part would be fine if it automatically recovered.. but, if the instance goes down, you'd have to restart it, at which point it would have a new IP and DNS setting.
I also wish there was some competition in the pay per hour instances... however, places like thegridlayer decided to go with monthly pricing.. completely destroying the advantages of instant scaling.
From their AWS license: "If your Application is determined (for any reason or no reason at all, in our sole discretion) to be unsuitable for Amazon Web Services, we may suspend your access to Amazon Web Services or terminate this Agreement at any time, without notice."
You can script simple applications to scale compute clouds on demand if load is high, low, or an instance fails. That does doesn't seem too difficult. An amazon user suggested Webmin to setup and manage the clusters.
The difficult part for me is persistent storage and lack of IPs. There is an application called S3InfiniDisk that supposedly allows cached storage onto S3. For my simple need where I would only need file access and storage, this seems ideal for a central storage solution. I tried experimenting with S3InfiniDisk, but couldn't get it to write to my S3 bucket.
The other problem is no dedicated IPs. I have my main namebased members, and some folks who use domain names. Nobody uses SSL on this service, except for my main domain. The fact that there is no IP permanence poses a problem for me currently.
There isn't much ease of management of EC2 instances and S3 at the moment. And I don't think Amazon wants to get into that business. I think they are expecting enterprising organizations to come up with the tools that will make EC2 and S3 powerful for the masses (e.g. JungleDisk). I give them tons kudos, because they've essentially invented a market.
While I have seen the simplistic scripts to add and remove servers, I haven't seen anything that
really ties it in to a workable load balanced solution. Besides having to create and run your own load balancer, you'd couldn't run it within ec2 itself, because it's too unreliable (dns wise) and running it somewhere else would make things much slower.