Virtual storage I/O latency will impact VM performance because high read or write operations can cause performance issues to all the resources in the datastore. In order to resolve the issue, you need a common view into the virtual and the storage environments to help you identify the root cause of the issue.
To effortlessly troubleshoot I/O latency issues in your virtual environment, your virtualization management software should have a datastore I/O latency widget which lists all the top datastores with high read or write values. When you drill down to a latency issue within the datastore, you will be able to see several key indicators that show the current performance of the datastore. For example, IOPS, IO latency, disk usage, capacity, disk provisioning information, alerts that have been raised for that datastore, etc.
Monitoring the I/O latency metric is crucial. Generally, when you see a read or a write value which is greater than 20 millisecond, your datastores will more than likely experience latency issues. Drilling down into the datastore will show you a relationship or an environment map informing whether the datastore is affecting another resources’ performance. It’s possible that more VMs are hogging this storage resource causing latency issues. This occurs when VMs contend with each other for storage resources causing performance issues to other VMs and the storage system. When VMs have performance issues, critical applications running inside will also end up having a bottleneck. All this ultimately affects end users.
You can take the following measures to regulate I/O latency issues so your VM performance is not affected:
- Allocate I/O resources to VMs based on their workload.
- If you have a critical application or a latency heavy application, ensure that’s the only application running in the VM. Associate a VM that has a critical application running in it to a single datastore to prevent high I/O latency.
- If the VM is not utilizing the allocated storage resource, consider routing unused resources to other VMs which have I/O contention issues.
It should be noted that not all storage system performance the same. That being said, set appropriate threshold values based on the performance of each storage device to avoid latency issues. Manage your VMware datastore with high I/O latency issues easily. In the video provided at the end of this blog, vExpert Scott Lowe shows you how to drill down from a datastore. You’ll see how the datastore is mapped to hosts, clusters, and VMs in your environment and look at their dependencies. Once you have this mapped, it’s easy to find related resources that are either causing the problem or being affected by the problem. Leverage a virtualization management tool to avoid storage I/O latency issues.