I'm getting close to launching my first startup, and I've been planning on using AWS to host. But after this week's issues, I think I'm obligated to ask whether or not that's a good decision.<p>The AWS outage spawned a number of related blog entries, and several of them discuss overall issues with AWS.<p>My plan was to launch on a micro instance that hosts both web and database sitting behind an elastic load balancer (ELB), with a spare medium web-only instance ready to fire up to offload requests from the micro instance.<p>At this point in time, I understand that if an AZ or region (or any single critical construct) goes down, so will my site. For launch and the first couple of months, I'm fine with that, given the relative rarity of these problems.<p>My bigger concern centers around my app and how it fits into AWS.<p>My app is written using the Django Web framework and is backed by Postgres. The instance that hosts the DB is using EBS.
I'm comfortable with backing up the database using traditional dumps and EBS snapshots, and I can provision a new server and have it ready to go in short order.<p>What worries me is that AWS instances don't seem to perform like other servers, particularly when it comes to EBS.<p>Several of the posts I've read discuss strategies for dealing with the AWS shortcomings (limitations?), but most of them seem to require re-architecting a custom app, not using a framework like Django.<p>My site is not going to be super-highly trafficked, so I don't want to do more than I need to for launch (e.g. DB farm + web farm + cache farm). I'm just trying to figure out what I need in order to get off of the ground.<p>Does anyone have any good advice for deploying a "standard" framework-powered app?
If so, I'd greatly appreciate it.<p>P.S. If you'd be willing to help beta test the site in two to three weeks, I'd love to have some HN user feedback. If you're interested, please e-mail me at bretwalker -> gmail.<p>Thanks!
I run Django apps on AWS and two other VPS providers and I wouldn't walk away from Amazon because of this outage. Amazon have some great features that collectively aren't available as a package elsewhere.<p>My tip would be to host the db and app server on a single instance to start with (less things to go wrong), and assume that server will fail.<p>Have a warm standby either in another AWS region or (like us) with a completely different provider. And make sure you have several copies of your data available to restore if need be. (For this I use dirvish on yet another VPS provider).<p>Finally, test your failover / restore procedure on a fairly regular basis.<p>My email is in my profile. Feel free to ping me if you have any specific questions.
I think Quora's err page summed it up best - "We wouldn't be where. we are today without Amazon". There are lots of discussions about variable EBS performance from larger consumers around the web, i've yet hear smaller players get vocal about it (or vote with their wallets). So until you become a large consumer it's probably not worth being too concerned about it.