What is the throughput and IOPS limit for instance with EBS?

Question

For example, looking at the Amazon EBS–optimized instances an m4.xlarge instance has a dedicated bandwidth of 750 Mbps which is a maximum throughput of 93.75 MB/s, and maximum IOPS of 6,000.  
The question is, gp3 volumes for example have a baseline throughput of 125 MB/s and 3000 IOPS. If I attach multiple volumes to an instance. Does it means that the instance will not only handle one single gp3 volume with its baseline?.  
Or is it that instances can get more throughput and IOPS than the one specified in the EBS-optimized page, and that "dedicated bandwidth" is just an extra?  
  
Also, I have seen on    that the maximum IOPS for an instance is 260,000. But there's not a single instance that is close to that number. The maximum ones are 80,000 and we are talking about the biggest ones like m5.24xlarge, so that makes me wonder if the values on the EBS-optimized instances page are the throughput and IOPS limits for instances those instances or I'm misunderstanding the documentation.

Accepted Answer

Hello. The values in the table apply to the instance type. The bandwidth is shared by the volumes on your instance.  
  
If you scroll down in the table in \[1], the largest R5b instance supports 260,000 IOPS. There are also instances that support 160,000 IOPS, such as the largest C6gn and M6i instances, and the high memory instances (u-*tb1).  
  
\[1] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ebs-optimized.html#current  
  
I hope this helps.  
  
Julie

Answer

Okay, so I understand then that the values that are shown on the "EBS optimized by default" table, on the columns "Maximum throughput" and "Maximum IOPS", are then the maximum IOPS and throughput for those instances. And they are unable to go beyond that level. The "Dedicated EBS Bandwidth" is the only bandwidth EBS-optimized instances can get. And is **impossible** for them to get any shared network bandwidth for EBS (like non-optimized EBS instances). Is that correct?

Answer

Okay, so I understand then that the values that are shown on the "EBS optimized by default" table, on the columns "Maximum throughput" and "Maximum IOPS", are then the maximum IOPS and throughput for those instances. And they are unable to go beyond that level. The "Dedicated EBS Bandwidth" is the only bandwidth EBS-optimized instances can get. And is **impossible** for them to get any shared network bandwidth for EBS (like non-optimized EBS instances). Is that correct?

Answer

Okay, so I understand then that the values that are shown on the "EBS optimized by default" table, on the columns "Maximum throughput" and "Maximum IOPS", are then the maximum IOPS and throughput for those instances. And they are unable to go beyond that level. The "Dedicated EBS Bandwidth" is the only bandwidth EBS-optimized instances can get. And is **impossible** for them to get any shared network bandwidth for EBS (like non-optimized EBS instances). Is that correct?

Answer

That is correct. With EBS optimization, all EBS traffic uses dedicated EBS bandwidth and can't use instance network bandwidth. Without EBS optimization, all EBS traffic uses the instance network bandwidth.  
  
Julie

What is the throughput and IOPS limit for instance with EBS?

相關內容