Minutes of the IT strategy steering committee, 30/04/12

0. Introduction

The meeting opened at 14:00.

1. Minutes of previous meeting

The minutes had previously been circulated and were approved.

Actions from previous minutes:

• The SC's recommendations for purchasing equipment from the rolling grants had been implemented.
• MJH had passed on the SC's comments about IT training to the CAR management committee. The feeling at the MC was that external training could be funded if it was relevant to a student's PhD programme (but not otherwise). Supervisors should approach Janet/Sean (as appropriate) with a proposal.
• The laptop policy now mentions mobile devices.

2. System status report

JA gave a report on central IT facilities.

He reported that the upgrade of Linux desktops to Fedora 16 is well under way, with over 50 upgrades completed. These upgrades take significantly longer than previous ones due to the new boot loader being shipped with FC16 - GRUB2. This required each Windows partition to be shrunk by 1 MB, which can take up to 3 hours. Progress continues...

The UPS has now been installed and works correctly. In its current state it has a 25% load. This will be even lower if/when the cluster is relocated to de Hav (see below). We estimate battery runtime to be of order 10 minutes.

The only issue since the last meeting was the failure of the /www filesystem on star-server (which remains unexplained). While inconvenient, it did provide a real test of the tape backups which enabled the full restoration of the filesystem.

3. Cluster report

MJH reported on the status of the cluster, as follows:

• The cluster, including the dedicated CAR nodes reported in the last minutes, was running well and being used extensively, with an increasing fraction of the use being by researchers in CAR.
• The extra storage bought from the remains of the rolling grant had been installed, together with some funded by PAM end-of-year funds and some new storage bought by CAIR. This means that there is 61 Tb of shared storage (CAR/CS/HH), 58 Tb of CAR-dedicated storage and 117 Tb of CAIR-dedicated storage attached to the cluster. With the 25 Tb of /data on the main system and 20 Tb on the smp machines, we are well past the 0.25 Pb mark in the server room.
• Downtime over Easter had been required to make space for the extra storage and to install the new RAID units, but otherwise there had been few problems.
• Although most of the racks are now full, a second CAR-dedicated chassis (192 cores) had been purchased from end-of-year funds. It should be possible to set this up for testing outside of the racks until a new home for the cluster can be found.

JA reported that preparations for the move of many PAM/CAR staff to the Innovation Centre (IC) were under way. IH had allocated an adequate address block for the new subnet that would be required, and one of the newly purchased gigabit switches would be used to provide fast connectivity to staff desktops. A separate DHCP server would be required in the IC but otherwise all server functions would be handled by existing machines. B/W and colour printers would be moved from the old Starlink room to the IC. The clearing of the Starlink room would also require the disposal of various spare IT equipment. The committee agreed that it would be important to retain some storage area for equipment not currently in use after the loss of the Starlink room.

MJH reported on the current plans for relocating the cluster. The existing server room is cramped and there is no room for expansion; the new development of E118 precludes any easy expansion of the room. After discussion of several options with Sean and IH it is proposed to relocate the cluster to the new IH data centre on the de Haviland campus. IH are discussing making 9 racks available to us there, which would triple the amount of space we now have, and they have the cooling capacity to support further expansion of the cluster as well. The move would take place either in September or early in 2013, and would obviously cost some money for professional movers and replacement of infrastructure shared between the cluster and the Starlink systems, as well as of the order of a week of cluster downtime. More information to be provided once we have a definite date. Committee members asked about the problems of remote administration: MJH explained that all cluster equipment can now be administered via web browsers, up to and including reboots and power cycling, so that there is no need for frequent physical access to the cluster.

5. AOB

There was no other business. The reps had nothing to report.

6. Date of next meeting

The next meeting should be in September 2012.

Action: MJH to circulate a Doodle nearer the time.

The meeting closed at 14:40.