Book Image

VMware Performance and Capacity Management, Second Edition - Second Edition

By : Sunny Dua
Book Image

VMware Performance and Capacity Management, Second Edition - Second Edition

By: Sunny Dua

Overview of this book

Performance management and capacity management are the two top-most issues faced by enterprise IT when doing virtualization. Until the first edition of the book, there was no in-depth coverage on the topic to tackle the issues systematically. The second edition expands the first edition, with added information and reorganizing the book into three logical parts. The first part provides the technical foundation of SDDC Management. It explains the difference between a software-defined data center and a classic physical data center, and how it impacts both architecture and operations. From this strategic view, it zooms into the most common challenges—performance management and capacity management. It introduces a new concept called Performance SLA and also a new way of doing capacity management. The next part provides the actual solution that you can implement in your environment. It puts the theories together and provides real-life examples created together with customers. It provides the reasons behind each dashboard, so that you get the understanding on why it is required and what problem it solves. The last part acts as a reference section. It provides a complete reference to vSphere and vRealize Operations counters, explaining their dependencies and providing practical guidance on the values you should expect in a healthy environment.
Table of Contents (28 chapters)
VMware Performance and Capacity Management Second Edition
Credits
Foreword
Foreword
About the Author
Acknowledgments
About the Reviewers
www.PacktPub.com
Preface
Part 1
Part 2
Part 3
Index

Storage counters at the cluster level


vCenter does not provide information for storage at the cluster level but vRealize Operations does, including counters such as IOPS, Throughput, and Latency. The main reason why you should not look at storage at the cluster level when working with classic arrays is that the cluster is a compute cluster; it is not a storage cluster, so the boundary a cluster provides for compute may not apply to storage.

Can you think of another reason?

The data at this level, like for the ESXi host level, includes all the local datastores. This can impact the overall result, especially those that give an average of all the datastores. If you have a cluster with 10 nodes that share five datastores, you will have 15 datastores in the clusters. The 10 local datastores will skew the total result, masking important data such as average latency.

For a view beyond disk and datastore, the datastore cluster is what you should look into.