If you are super Sys Admin with lots of experience on EC2 then a lot of this might sound like motherhood. However you would be surprised how many people I have interviewed that claim to have extensive EC2 experience that do not understand some of these concepts.

1. If you are upgrading or releasing new versions of code regularly consider some form of deployment management software such as Puppet that will allow a basic EC2 Image to be launched and then the full configuration sucked down from your deployment server, this allows you to launch new instances with minimal configuration, it also stops you from having to update your images every time you put out a new deployment, it also increases the speed at which you can recover from a crash and the speed at which you can react to traffic peaks
2. Make sure from the start all your Application configurations allow for application components that could be separated out from each other to be easily separated onto different EC2 instances, its easier to do this up front than when you are in a hurry later
3. Backup to S3 costs nothing in data transfer (just storage costs and the processing overhead on your instance) therefore backup as often as is practical based on your dataset
4. Make sure you have 3 DNS with at least one geographically and provider independent of the other
5. Consider how you can create shared nothing architecture ie each instance once launched and populated is not dependent on any other to keep operating, each new instance can launch, get data as required from Master and the operate independently updating master as required and could continue to do so if master died, updating new master when it is replaced.
6. Dont waste time on small instances just start with large or extra large, they dont perform very well and cpu per $ pro-rate have limited CPU and bandwidth and IO compared to large and extra large
7. Its worth spending a few days on this site http://highscalability.com/ the case studies on major sites are very informative
8. You still need external management servers to monitor and give you a heads up if you have failures on your instances, unlike a Dedicated/Management Server in a good hosting company Amazon is not going to have one of their engineers go an reboot your instance if it dies, I open to correction but I am not sure they would even notice or ping you if there was a problem with one of your instances. You need to track and monitor this yourself and make sure your system can SMS or email you and that you have the ability to easily recover quickly. (another example of the same EC2 genius, suggested we put management alert system on the EC2 instance we were monitoring, after asking him to explain the process by which we might get an alert in case of instance failure we both agreed to part ways)
9. Load balancing is not as easy as you might imagine, especially if you are starting and stopping Instances regularly, the IPs for each new instance have to be inserted and removed from the load balancer dynamically, you cant rely on hardware style devices, you need to implement other linux software based systems.
10. Use Elastic IPs, this  is a new feature, previously you were given an IP for an instance from a pool, you didnt necessarily know what it was until it was assigned. This plays havoc with your uptime if you have a failure or if you want to move from one instance to one or more instances. (ie you need to put new IPs into DNS and wait for propagation) Elastic IPs basically allows you to assign an IP to your instance or move the IP from one instance to another as required and it only takes a few minutes ie no waiting 4-12 hours for DNS updates (as am example of Engineers not understanding this platform as much as they claim, I had one EC2 engineer guy go to shut down a production Instance on me without even thinking about DNS changes and the outage he was about to create)
11. ElasticFox and S3fox are great plugins that allow a marginally skilled guy like me to easily operate EC2 and S3 but be warned they allow you to do just about whatever you want ie delete, or reboot an instance reassign an IP to a different instance and launch all different instance types. Be careful or you will learn that EC2 storage is not persistant (yet)
12. Persistent Storage. Its important to note that although you get 1.7tb of disk space with an extra large instance, you cant rely on this for keeping your data. Think of it as a temporary work space that will be deleted if you close the instance or it dies. You must implement a means by which you either transfer all data, files etc to S3 or some other form of non Amazon storage as part of your regular daily operation or architecture so this could be writing to S3 directly (needs coding change),  or some way of syncing your data to an external server. EC2 is introducing persistent storage and it is in private beta, but its not generally available yet and from a backup perspective you still need to do work out how to get it away from Amazons infrastructure anyway as a backup.

• 0 Comments

Its been a crazy few months, I have been running Enikos.com for about 6 months now and we have managed to build a system that allows video owners and publishers to monetise and syndicate their content and earn revenue for it.

Very proud that we have been awarded 12th position in Australia’s top 100 Web 2.0 applications by Business Review Weekly last week.  http://enikos.com/enikos-voted-12-brw-top-100-australian-web-20-applications/

• 0 Comments

Hi

OK December is much better than November which I am happy about, given December in business to business land is a difficult month.

I managed to score a great consulting gig with a hot new online video company which I will blog more about shortly.

So the results for Dec are in

Consulting $15500

Google sales = $0

Website Advertising $2400

Online Advertising for others $500

CPA $60 (this was an experiment)

$18400

Again not bad, but not moving fast enough to the $85000 per month required to keep me in the EONetwork.

I would like some of my readers to get in on the act, set yourselves a challenge and join the Million $ challenge

• 0 Comments