0

I'm new to AWS and have been tasked to look at ways to optimize cost. I have gone through the best practices and strategies to do that:

  • Instance Right Sizing
  • Using Reserved Instances (RI) & Spot instances (SI) as appropriate (over On-Demand)
  • Auto Scaling Storage (Reduced Redundancy, Glacier)
  • Tagging strategies

However, I have specific questions around the instance type. Apparently, we have a r3.xlarge (On-Demand) instance used for demos (for new clients) with 240 GB EBS (i.e. 4 CPU, 30.5 GB RAM), therefore it's not used very frequently.

Questions :

1) Will replacing it with a t2 instance give any cost benefits (in medium to long term) without degrading performance (or any other adverse effects)?

2) If so, which t2 variant can be used (t2.xlarge, t2.2xlarge etc)?

Sorry for my ignorance if this is not the proper place for the question, any guidance would be much appreciated.

Thanks for your help!

user396603
  • 1
  • 2
    Possible duplicate of [Can you help me with my capacity planning?](http://serverfault.com/questions/384686/can-you-help-me-with-my-capacity-planning) – user9517 Jan 23 '17 at 13:24
  • Where you say 240GB EBS (ie 4CPU etc) it might help to know that EBS (elastic block store, ie network based disk) is not linked to CPU and RAM. EC2 instances have RAM and CPU, you can have any amount of disk you like, and you can have magnetic, solid state, or solid state with provisioned capacity. – Tim Jan 23 '17 at 17:53

1 Answers1

3

Replacing a r3.xlarge with a t2 node will save $. If the node is being used for occasional demos then the burst-cpu nature of t2 sounds like a good fit.

However, it is impossible to say anything about which one to use without knowing the details of the application. You might be running a static nginx server on that R3 and wasting 99% of the memory, or you might be running a large in-memory database and barely fitting.

With that, http://www.ec2instances.info/ is a useful tool for comparing the cost-per-hour and capacity of various AWS instance types. Look at your app (and its performance metrics / history) and switch to the size that best fits your application.

If you keep this instance running all the time, consider purchasing a 'reserved instance' to further reduce costs. A Reserved Instance (RI) is a billing construct where you commit to pay for an instance for a full year, and AWS gives you a discount. Note that it billing-only, so AWS will apply the discount to a matching node -- you don't have to reprovision anything, nor launch a special RI node.

Jason Martin
  • 4,865
  • 15
  • 24