Amazon Web Services Arfchitecture and Costing

Question

I am extremely new to Amazon Web Services. I would sincerely appreciate any help on finalising the architecture and arriving at a costing schedule. I am working on designing a AWS based solution for a dynamic website that we are designing. To begin with I need to have 2 High CPU Medium Utilisation EC2 instances (both will act as web servers), 1 High CPU Medium Utilisation EC2 instance for my database which will be Postgre SQL and 1 High CPU Medium Utilisation EC2 instance to serve as a read replica for my database. I will also be having a considerable volume of static content like images, videos or just .doc files for which I am contemplating using an S3 bucket. So, my website will be a dynamic + static kind of a website. I am expecting a rapid exponential scale up of users from 0 to say for example 1 million in a year. Hence I will need to scale up my EC2 combos (as described earlier) according to traffic. I am contemplatng using a CloudFormation stack for rapidly scaling my deployment. Also, to efficiently route traffic I will be using a single ELB to start with. Also, I would want to vertically partition my database based on user id's. For example, user ID 1 - 2000 on one EC2 database instance users 2001 to 4000 on second EC2 database instance etc... I will be auto scaling only the web server EC2 instances while my database EC2 instances will have a 100% uptime

My questions are:

1. What should be my auto scaling strategy for the web server EC2 instances and how do I know what will be the monthly costing when the scaling is so dynamic. I mean is there any way to predict so that I can do a cost break even analysis?
2. Do all the EC2 instances (web server and database) necessarily need an EBS backing or will Ephemeral storage suffice? I believe that for the database EC2 instances I will need an EBS backing. What about the web server Ec2 instances?
3. Suppose I end up scaling up to 100 EC2 instances. Will just one ELB suffice or do I need multiple ELBs?
4. How do I analyse as to how many HTTP requests can one High CPU Medium Utilisation EC2 instance handle before a breakdown?
5. Can CloudFront be used to host this kind of a dynamic + static site or is it used only for static sites?

Please help me with these questions as I have no clue on cloud solution architecting...

Thanks... Vikram.

Answer 1

Here is a white paper we (RunSignUp) wrote about how we scaled on Amazon. Our use case was handling opening of an online race registration where 50,000 runners would want to sign up in less than 10 minutes. This shows how we configured everything including settings as well as some of the code we developed. We basically had the same issues as you and did not find anything that had a full use case of how to build a scalable app and how Amazon was used to scale with it. Hope it is helpful.

Answer 2

A quick attempt at an answer to get you started;

• #1 is too big to answer without writing a book, it has too much to do with the system and its design to answer in a generic way.
• #2 You'll definitely want backing for your database. The web servers on the other hand are probably better off with ephemeral storage, the more stateless/setup free they are, the easier to create/destroy instances dynamically.
• #3 As far as I know, there's no upper limit on the number of machines "attached" to a single ELB. Under extreme load, you may experience slightly lesser latency if you split your site over multiple load balancers. You'll find more info that you can map against your actual architecture here.
• #4 Only load testing of your system can tell you that.
• #5 From the Cloudfront page;

Amazon CloudFront can be used to deliver your entire website, including dynamic, static and streaming content using a global network of edge locations.

In other words, its architecture does not prevent you from mixing static and dynamic content on your site.