ORIGINALLY POSTED 23rd April 2013
14,181 views on developerworks
For the last few years we have often talked about ‘flexible hardware’ options for SVC. While we have offered the internal SAS HBA and SSD attachment within the nodes, and a 10Gbit iSCSI/FCoE capable upgrade, we’ve never quite made it to true flexible hardware, with respect to Fibre Channel ports, and CPU / cache memory upgrades… until now…
Over the last couple of months we have release two RPQ’able hardware upgrades, specifically for the latest SVC CG8 nodes.
One common request for SVC is more ports… Fibre Channel that is. In parituclar when deploying Metro/Global mirror solutions or Stretched Cluster configurations, where users would like to dedicate ports for specific uses, such as inter-site links, and support more fabrics.
8 Port Node Upgrade
The first RPQ allows you to add a second 4 port 8Gbit Fibre Channel HBA in the second PCIe slot in each node. Thus, providing 8x 8Gbit ports per node. There are some restrictions on the usage of the ports, and today they are primarily provided for the above replication use cases. The rules are roughly that you can only have 4 ports zoned for local host access, 4 ports zoned for local disk access and up to 16 “paths” between any two nodes (local or remote). From a pure base performance perspective, I’ve seen this double the write bandwidth, even without having a remote cluster – so if you are running write MB/s intensive workloads, this can help too.
12 Core 48 GB Upgrade
The second RPQ allows you to add a second 6 core CPU and additional 24GB cache to your CG8 nodes. This is provided for Real Time Compression use. It allows you to add RTC volumes without impacting existing workloads, that is, the caveats on enabling RTC depending on your current CPU usage go away! (See the RTC redpaper for details). The additional cache makes a huge difference to RTC, allowing it to cache a lot more “to be compressed” or “already decompressed” data and improve random access performance, with worst case workloads by up to 20x and best case workloads by more than 3x. All while maintaining the original I/O profile running on the system.
For details on how to order these upgrades, speak to your IBM sales/account teams and get them to help submit an RPQ request. Both upgrades come with additional config and installation information to help make the most of the extra hardware.
Leave a Reply