Q: What level of performance consistency can I expect to see from my Provisioned IOPS SSD (io1) volumes?

When attached to EBS-optimized instances, Provisioned IOPS SSD (io1) volumes are designed to deliver within 10% of the provisioned IOPS performance 99.9% of the time in a given year. Your exact performance depends on your application’s I/O requirements.

Q: What level of performance latency can I expect to see from my Provisioned IOPS SSD (io1) volumes?

When attached to EBS-optimized instances, Provisioned IOPS volumes can achieve single digit millisecond latencies. Your exact performance depends on your application’s I/O requirements.

Q: Does the I/O size of my application reads and writes affect the rate of IOPS I get from my Provisioned IOPS SSD (io1) volumes?

Yes. For a given allocation of resources, the IOPS rate you get depends on the I/O size of your application reads and writes. Provisioned IOPS volumes process your application reads and writes in I/O sizes of 256KB or less. Every increase in I/O size above 256KB increases linearly the resources you need to achieve the same IOPS rate. For example, if you have provisioned a volume with 500 IOPS, that means that it can handle up to 500 256KB writes per second, 250 512KB writes per second, or 125 1024KB writes per second, and so on. You can use

Q: What factors can affect the performance consistency I see with Provisioned IOPS SSD (io1) volumes?

Provisioned IOPS SSD (io1) volumes attached to EBS-optimized instances are designed to offer consistent performance, delivering within 10% of the provisioned IOPS performance 99.9% of the time over a given year. For maximum performance consistency with new volumes created from a snapshot, we recommend reading or writing to all of the blocks on your volume before placing it into service.

Another factor that can impact your performance is if your application isn’t sending enough I/O requests. This can be monitored by looking at your volume’s queue depth. The queue depth is the number of pending I/O requests from your application to your volume. For maximum consistency, a Provisioned IOPS volume must maintain an average queue depth (rounded to the nearest whole number) of one for every 500 provisioned IOPS in a minute. For example, for a volume provisioned with 1500 IOPS, the queue depth average must be 3. For more information about ensuring consistent performance of your volumes, see Increasing EBS Performance.

Q: What level of performance consistency can I expect to see from my HDD-backed volumes?

When attached to EBS-optimized instances, Throughput Optimized HDD (st1) and Cold HDD (sc1) volumes are designed to deliver within 10% of the expected throughput performance 99% of the time in a given year. Your exact performance depends on your application’s I/O requirements and the performance of your EC2 instance.

Q: Does the I/O size of my application reads and writes affect the rate of throughput I get from my HDD-backed volumes?

Yes. The throughput rate you get depends on the I/O size of your application reads and writes. HDD-backed volumes process reads and writes in I/O sizes of 1MB. Sequential I/Os are merged and processed as 1 MB units while each non-sequential I/O is processed as 1MB even if the actual I/O size is smaller. Thus, while a transactional workload with small, random IOs, such as a database, won't perform well on HDD-backed volumes, sequential I/Os and large I/O sizes will achieve the advertised performance of st1 and sc1 for a longer period of time.

Q: What factors can affect the performance consistency of my HDD-backed volumes?

Throughput Optimized HDD (st1) and Cold HDD (sc1) volumes attached to EBS-optimized instances are designed to offer consistent performance, delivering within 10% of the expected throughput performance 99% of the time in a given year. There are several factors that could affect the level of consistency you see. For example, the relative balance between random and sequential I/O operations on the volume can impact your performance. Too many random small I/O operations will quickly deplete your I/O credits and lower your performance down to the baseline rate. Your throughput rate may also be lower depending on the instance selected. Although st1 can drive throughput up to 500 MB/s, performance will be limited by the separate instance-level limit for EBS traffic. Another factor is taking a snapshot which will decrease expected write performance down to the baseline rate, until the snapshot completes. This is specific to st1 and sc1.

Your performance can also be impacted if your application isn’t sending enough I/O requests. This can be monitored by looking at your volume’s queue depth and I/O size. The queue depth is the number of pending I/O requests from your application to your volume. For maximum consistency, HDD-backed volumes must maintain an average queue depth (rounded to the nearest whole number) of four or more for every 1 MB sequential I/O. For more information about ensuring consistent performance of your volumes, see Increasing EBS Performance.

Q: Can I stripe multiple volumes together to get better performance?

Yes. You can stripe multiple volumes together to achieve up to 75,000 IOPS or 1,750 MiB/s when attached to larger EC2 instances. However, performance for st1 and sc1 scales linearly with volume size so there may not be as much of a benefit to stripe these volumes together.