Back to posts

Posts

Choosing the right Instance Storage for Amazon RDS Instance Storage

Compare Amazon RDS gp2 and gp3 storage, including baseline IOPS, throughput, and migration considerations.

Introduction

Amazon RDS storage options illustration
Amazon RDS storage options illustration

Sometimes ago AWS just announced that a new storage type is supported in RDS. If you already using EC2 you must be familiar with this storage type as it has been released since 2020.

The storage type that RDS just supported is gp3. It has similar characteristics to the usual gp3 that you are already familiar with in EC2, it has a baseline 3000 Baseline IOPS. What makes difference with EC2 EBSis pricing, as in RDS both gp3 and gp2 have a similar price so there is no cost saving for migrating from gp2 to gp3.

Issues with Previous Instance Storage

As you already know when we only have a small size of data on RDS with gp2 we only got storage_size * 3 IOPS = Total, with gp2 formula for example if only we are using 100GB of storage we only got about 300 IOPS with a burst of 3000. If we need higher IOPS we need to change the storage type to provisioned IOPS (io1) and it’s a bit costly.

When to Use gp3

Usually, we need to check why our workload uses high IOPS first and after that, we decide which storage type is suitable for our apps. That’s the best scenario, but it takes some time to analyze and fix, so what we can do quickly is increase our RDS IOPS.

It’s recommended to migrate from gp2 to gp3 if the workload needs iops more than storage_size * 3 and your workload doesn’t need IOPS more than 3000 IOPS & 128MiB/s Throughput. Be aware that if you need IOPS of more than 3000 it requires a storage size ≥400 GiB.

Not all kinds of workloads need a gp3, before you migrate from gp2 to gp3 there is a need to analyze first your estimated/current throughput usage as gp3 has a baseline of about 128MB/s on the other hand gp2 can burst up to 250MB/s. If your storage size is > 400GiB it’s no problem in throughput as gp3 has a baseline 500MiB/s and IOPS of 12000.

Summary

The gp3 can help us greatly especially in cost if you are having a small size of data in RDS but needs higher IOPS than the default gp2 provided as in before you need to use io1. Before deciding to migrate you need to take note, especially of current throughput usage as gp3 has a default throughput of 128MB on the other hand gp2 can burst up to 250 MB/s (>170GB Storage size ).

If you have any questions or corrections/feedback please let me know directly in the reply box or mail me at [email protected].

Thank you :)


Originally published on Medium.