I am running the Magento 2.2.3 version on AWS EC2 currently on c5x.large the performance seems fine for me with proper page speed backed by CloudFront and Redis.

Due to cost optimization, I have decided to use the m4.large instance and saw the degradation in performance Magento page speed from 2.5 seconds to 6.6 seconds. I noticed the CPU usage with m4 large type instance it was going up during cache creation and was neutral at other time. I also noticed the cache flush operation from the Magento admin panel which took approx 3.5 min whereas in the case of C5 large it was taking 50 seconds to complete the same operation.

Is something wrong with my application or it's the cache operation that has a direct connection with my CPU? What will be the right Instance series to be chosen for Magento 2.2.3 on production?

Also, this was not the case with Magento 2.1.6, the cache flush operations are very normal even with t2.medium instances we used t2.medium for dev instances earlier.

Specifications: M4 large : 2 vcpu 8 GB RAM C5 Xlarge : 4 vcpu 8 GB RAM

  • Sounds like a caching problem. Is the full page cache configured properly? This type of performance degradation happens when code isn't cached properly.
    – Hunter
    Commented Oct 12, 2020 at 19:56

1 Answer 1


This question as received an answer on the Magento forum at https://community.magento.com/t5/Magento-2-x-Hosting-Performance/Ideal-EC2-Instance-type-for-magento-2-2-3/td-p/91990

You are not only halving the CPU but also moving from a faster CPU to a slower CPU and a faster hypervisor + hardware to a slower one.

Magento generally performs better when you allocate more CPU (and faster CPU) as well as Memory to it.

If you are running a live eCommerce store and conversion is important to you (which it should), I wouldn't go with weaker hardware and slower performance just to save a small amount of cost as it will cost you a lot more in the long run.

If the answer above satisfied the question may be the OP could close the question

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.