Zephyr Is Now In Production

From Template
Revision as of 16:08, October 26, 2012 by Amiot (talk | contribs)
Jump to: navigation, search

As you may know, TRACC has acquired a new cluster computer from Atipa Technologies. It is called Zephyr, and it is a 92-node system with each node having two, AMD 16 core, 2.3 GHz, 32 GB processors. The system has 120 TB of Lustre-based user storage (see Zephyr and Phoenix Cluster Configurations). Zephyr, has now moved from an early user test mode to a production mode. We have installed all of the latest software versions that we think our users will need as well as some older versions that will allow you to run programs that are not compatible with the latest software releases.

Many of our users that have participated in our early user testing of Zephyr have reported faster run times on their codes. With higher core counts (16 cores per processor as compared to 4 cores per processor on Phoenix) and faster interconnect speeds (40 Gbps Infiniband as compared with 20 Gbps on Phoenix), well-tuned codes should see significantly shorter run times.

We have been updating this Wiki to help you get started running on Zephyr, and you can always get help by contacting us at TRACC-help@anl.gov. Also, we would appreciate it if you would send us email telling us about your experiences running on Zephyr.

There are still a few issues that remain with regards to Zephyr. As examples, we do not yet have data backup implemented; TRANSIMS Studio is not yet implemented; and we are still working on a couple of Ls-Dyna scripts. These and other issues are, or shortly will be, discussed in more detail in this Wiki. See "Linux Cluster/Using The Cluster/Zephyr Outstanding Issues".