------------------DOCUMENTS------------------ Hardware: - pfjl0_details_specs_racks.mail - VS_BladeRack_Manual_RevI - ProCurve_Switch_2800_Series.pdf - ARCvault 24 User Guide SOFTWARE - Web-page: http://oceans.deas.harvard.edu/haley/HideyHole/cluster_software.html - rocks-usersguide-4.2.1 - pjh3_notes_conf_call_final.mail ------ CLUSTER SET-UP TODO: SOFTWARE TO INSTALL ------- - Cross mounting of disk: set up NIS on master server with NFS such that linux/windows desktops can see cluster disks. (Samba/CIFS for windows) - partitioning of disks in 1U so that one/some of them is/are used for backup. - software download from MIT for install on large cluster: Need to check with IS&T or simply download from Web, install on front-end and use rocks to copy. Someone needs to do a list of free/4pay software needed from MIT list - PVM: Check to see version included in Rocks works with schedulers -- Parastation, PBS, Score: to be installed Need to install parastation on top of Rocks (also Score) - parallel file systems: at least on main storage array (NFS and PVFS2). - OS can be changed mid-stream (user authentication) ---OS recommended by 2.006 student, which we should check http://en.wikipedia.org/wiki/Debian_GNU/Linux http://www.debian.org/ http://en.wikipedia.org/wiki/Ubuntu_Linux http://www.ubuntu.com/ DONE: - Software development kits: wait until we have better idea of needs. and have tried a few (free, demo licenses) - Firewalls (rocks default) - backup data: Oleg to contact MIT IS&T - Need to set-up rolling back-ups (search Pat's page http://oceans.deas.harvard.edu/haley/HideyHole/cluster_software.html - ask someone how to continue to copy files across tapes (Al Conte)? (amanda - see cluster_action_list.txt) - Surge protector: done, but we need to check it. - internal connect and configuration of switches for more than 1G connect (see Conway's emails). Switch-to-switch interconnects set-up, need testing. - We have Rocks (contains RHEL4). Contains: CentOS 4 update 5 (replace the OS Roll with the complete media set from any Red Hat Enterprise Linux 4 compatible distribution, including the commercial version). Schedulers: CONDOR (High throughput computing tools), grid Globus 4.0.4 (GT4) SGE Sun Grid Engine job queueing system area51 System security related services and utilites bio Bioinformatics utilities ganglia Cluster monitoring system from UCB java Sun Java SDK and JVM pvfs2 PVFS2 File System MPI installed