AWS Magento with Load Balancer

Question

I plan to use Amazon EC2 Server(s) for Magento. But I'm fairly new to AWS.

I know that I have to use Elastic Load Balancer (ELB) to balance load between two or more EC2-Instances. - That is important, because it's highly possible, that my main instance is having a loading peek 1-2 hours per day.

I can`t connect one EBS to two EC2-Instances, I know. But I have to have the very same data on both (or more) EC2-Instances. - One possible solution is to make a snapshot of Instance-1 and start it to Instance-2. But as the data can change really quickly (Cache for example, new products ...) it's maybe not the best solution, I think.

I heard that I can mount my S3-Storage to my instances and then use this as "global" storage, but as far as I know from different articles S3 is not quickly enough for a high-peek-storage-server.

Some facts by the way: this server is going to have 200-300 visitors per hour, but it can be 500-1000 too.

Conclusion: I need a Storage Server, that is quick enough to share a lot of data (images, js, css, php) and is mountable to more than one instance. How do I do this in a clever way?

Greetings Bubble

Answer 1

The new EFS service (NFS share) can give you a simple solution to what you are seeking to do, but its cost is high compared to the alternatives.

When you are dealing with multiple instances your instances should follow the shared nothing architecture. Meaning, any unique application data is not stored on the instance.

Application code can be stored on the instance, you should have a release process to update this automatically on an instance if its changed.

Cache data is something that can be regenerated, this ideally should be a memory cache like memcached.

Application data (Product images, etc) should be stored on S3. You can also serve from S3 (which offloads some of the work from your web server). I believe there are plugins for Magento to store images on s3.

Database should be on a server outside of web server instance. You may be able to use RDS to set this up quickly.