Volume throughput not improving after instance type change

Question

I changed an instance from r5.4xlarge to m5d.8xlarge expecting the disk I/O performance to increase. But I am not seeing any improvement on the volume metrics.  The volume is EBS and it shows that EBS optimization is enabled.
I have another instance that was launched initially as an m5d.8xlarge that actually shows significantly better throughput.  Both instances were originally from the same AMI.
My question is : when changing an instance type by stopping, changing instance type then restarting ,  does the root volume get upgraded also?

Accepted Answer

There are few things here. 
First both your EBS volumes and your EC2 instance have provisioned capacity. The EC2 instance has burst and baseline throughtput. You need to compare these with the throughput of the EBS volumes to see if your EC2 instance has sufficient capacity.

```
$ aws ec2 describe-instance-types --instance-types r5.4xlarge m5d.8xlarge --query "InstanceTypes[*].[InstanceType, EbsInfo.EbsOptimizedSupport, EbsInfo.EbsOptimizedInfo]"
[
    [
        "r5.4xlarge",
        {
            "BaselineBandwidthInMbps": 4750,
            "BaselineThroughputInMBps": 593.75,
            "BaselineIops": 18750,
            "MaximumBandwidthInMbps": 4750,
            "MaximumThroughputInMBps": 593.75,
            "MaximumIops": 18750
        }
    ],
    [
        "m5d.8xlarge",
        {
            "BaselineBandwidthInMbps": 6800,
            "BaselineThroughputInMBps": 850.0,
            "BaselineIops": 30000,
            "MaximumBandwidthInMbps": 6800,
            "MaximumThroughputInMBps": 850.0,
            "MaximumIops": 30000
        }
    ]
]
```

"Does the root volume get upgraded also?"

No, when you stop/start an instance you will keep the same volume.

```
aws ec2 describe-instances --instance-ids ws ec2 describe-instances --instance-ids i-0de519a76d9555c23 --query 'Reservations[].Instances[].[BlockDeviceMappings[0].Ebs.VolumeId]' 
[
    [
        "vol-0f7de28eb2b158b41"
    ]
]

aws ec2 stop-instances --instance-ids i-0de519a76d9555c23
aws ec2 describe-instances --instance-ids  i-0de519a76d9555c23 --query "Reservations[*].Instances[*].[InstanceId,State.Name,InstanceType]"  
aws ec2 start-instances --instance-ids i-0de519a76d9555c23

$ aws ec2 describe-instances --instance-ids ws ec2 describe-instances --instance-ids i-0de519a76d9555c23 --query 'Reservations[].Instances[].[BlockDeviceMappings[0].Ebs.VolumeId]'
[
    [
        "vol-0f7de28eb2b158b41"
    ]
]
```

Answer

The way to improve EBS performance would be to change the volume type.

Try changing form a GP2 to GP3

https://aws.amazon.com/ebs/general-purpose/

Answer

> does the root volume get upgraded also?

No. EBS volume is attached to EC2 instance via network(independent). By changing instance type, network throuput b/w EC2 instance and EBS volume can get better, but IOPS throuput of EBS is same.

To optimize throuput of EBS volume, the link as follows will help you.

https://repost.aws/knowledge-center/ebs-performance

Volume throughput not improving after instance type change

相關內容