You want to run what?, on what? on what? (Amazon style)
I found instructions for running Windows images in QEMU on Xen on (presumably) Linux. Phew.
This reminded me that I've been meaning to post about Amazon's new grid offerings s3 and ec2.
I get Amazon's s3 storage-on-demand -- I think it's great and I'm actually using it to store older database backup files for a production server. IIRC, Amazon is charging a tenth what our hosting provider charges for local SCSI/RAID disk space.
The s3 tools are way cruder than I'd like (even jungledisk's webdav->disk mapping is a bit hacky), but I see the potential.
I'm having trouble wrapping my head around EC2, though. I'm spoiled by desktop virtualization, and my clients are starting to offer quick virtual hosting of servers on their intranets. So, when/why would having a server that runs for a fraction of a day/month/whatever make sense?
I know it's exciting because it's new and all, but I just don't get it and I'm not sure who the target audience is. I suspect I'll have to wait until folks start building apps on top of EC2 before it gets on my radar.
Update: Ten minutes after I posted this, it did occur to me when/why you'd want to choose EC2. Amazon is in the process of rolling out something similar to VMWare's Virtual Appliance directory with the difference of offering a rapid deployment/hosting option.
These appliances are somewhat interesting to imagine running on a per/minute basis, but considering how easy it is to download to your desktop it's not too great. It begins to become really compelling once you move beyond individual machines and into the concept of what's now termed "Lab Management" - when you have a testing environment that consists of many machines acting in tandem, and you need to build up the cluster, test it, and tear it down and restart it , many times and in many different configurations. Covering your test matrix for distributed applications/SOA is hard, and Lab Management is ridiculously easier than the alternative.
Lab Management will remain inside the enterprise, but the value and direction of EC2 is in applying the testing-focused Lab Management concept to production user-facing applications: what I like to call the Downloadable Data Center.
Why is this cool? Well, consider the difference between your typical startup and a mature web enterprise: to really run a web hosted application according to best practices, you should have
- staging setup
- production setup,
- hot standby, DR plan
- version control repository/bug tracker
- integrated authentication
- distributed file system
- load balancer
- firewall/intrusion detection
And no one does initially because it takes a lot of time, money and expertise to put all these pieces in place. But what if you could have it all initially and it didn't cost an arm and a leg? The idea of a vendor (like, say Novell or RH) pre-provisioning all the machines required to pull the above off, and offering them via the Amazon EC2 Control Panel is quite compelling. Imagine the options:
- Statefull Firewall with mod_security? Check.
- Dedicated Image Server pre-configured with optional Akamai CDN support? Check.
- Web analytics reporting server? Check
- Offline bi/olap database with real-time replication? You get the idea.
Each check on that control panel is the equivalent of days or weeks of work on your hand-rolled data center. Accordindly, enabling composable application clusters like this would be worth much more than $60-70/mo/machine. EC2 ISV's like Novell would clearly be due some royalty or rent on top of the EC2 grid-dial-tone costs, in addition to their traditional subscription/support revenue.
In the near term, it's hard to imagine shifting from a hand-rolled mini data center to a commodity one because of all the weird hacks that evolve. You could call this "The ActiveGrid Challenge" since that product, which sniffs around the same space, has essentially identical adoption barriers.
Over time, as developers become aware of the prefab/click-to-deploy options and start coding to them, however, it's hard to imagine it not taking a good chunk of the hosting marketplace.
To give credit where credit is due, Marc Andreeson had this idea with his LoudCloud startup. For the small fee of $100K/month, they would professionally provision and maintain your data center. I liked the idea, but thought that although it sounded like a nifty escape valve for harried CTO's, they'd probably run of of customers pretty quickly at that price, and they did. I'm betting the same fate does not await EC2 much more mass-market effort.