IBM Spectrum Scale and IBM Elastic Storage System Network Guide
An IBM Redpaper publication
Published 17 February 2021
IBM Form #: REDP-5484-00
Rate and comment
Authors: Kedar Karmarkar, John Lewars, Sandeep R. Patil, Sandeep Naik, Kevin Gildea, Rakesh Chutke, Larry Coyne
High-speed I/O workloads are moving away from the SAN to Ethernet and IBM® Spectrum Scale is pushing the network limits. The IBM Spectrum® Scale team discovered that many infrastructure Ethernet networks that were used for years to support various applications are not designed to provide a high-performance data path concurrently to many clients from many servers.
IBM Spectrum Scale is not the first product to use Ethernet for storage access. Technologies, such as Fibre Channel over Ethernet (FCoE), scale out NAS, and IP connected storage (iSCSI and others) use Ethernet though IBM Spectrum Scale as the leader in parallel I/O performance, which provides the best performance and value when used on a high-performance network. This IBM Redpaper publication is based on lessons that were learned in the field by deploying IBM Spectrum Scale on Ethernet and InfiniBand networks.
This IBM Redpaper® publication answers several questions, such as, “How can I prepare my network for high performance storage?”, “How do I know when I am ready?”, and “How can I tell what is wrong?” when deploying IBM Spectrum Scale and IBM Elastic Storage® Server (ESS).
This document can help IT architects get the design correct from the beginning of the process. It also can help the IBM Spectrum Scale administrator work effectively with the networking team to quickly resolve issues.
Table of contents
Chapter 1. IBM Spectrum Scale introduction
Chapter 2. Network planning and best practices for IBM Spectrum Scale and IBM Elastic Storage System
Chapter 3. Implementation recommendations
Chapter 4. Network monitoring and troubleshooting
Appendix A. GPFS lease configuration variables effect on expel timing flows
Appendix B. nsdperf command examples
Others who read this publication also read
Follow IBM Redbooks
Follow IBM Redbooks