Minutes of the IT strategy steering committee, 15/01/13

Present: Martin Hardcastle (chair: MJH), John Atkinson (systems manager: JA), Tim Gledhill (staff rep: TMG). Mark Thompson (staff rep: MAT), Hywel Farnhill (student rep: HF), John Barnes (PD/fellow rep: JB)

0. Introduction

The meeting opened at 11:00. MJH explained that HF had replaced Kieran Forde as student rep, as Kieran was now writing up.

1. Minutes of previous meeting

The minutes had previously been circulated and were approved. There were no outstanding actions.

2. System status report

JA gave a report on central IT facilities.

He reported that, as part of the cluster relocation, CAR had purchased a new storage RAID controller giving car 28TB of dedicated /data space. The server room E117D now only houses the CAR system kit and would be rearranged in a better layout. An upgrade to car-server would be scheduled at the same time. The upgrade would bring car-server up to FC16. Whilst upgrading does not address a specific security issue, it is favourable to upgrade the OS (as opposed to a fresh install) whilst FC 13 is now classed as "legacy". (Since the meeting, the relocation and upgrade has been carried out successfully.)

A number of staff had been relocated to the Innovation Centre (IC) as trailed in the previous minutes and these moves had been largely seamless. MJH reported that there were some networking issues between the IC and other parts of UH: a problem for client desktops when accessing /home directories on car-server. The problem has been reported to IH who state it is due to VLAN 900 being a legacy VLAN and the hardware connecting the IC to this VLAN being inadequate. Two problems were suggested: reconfigure IC kit to VLAN900 or wait until new networking hardware can be installed by IH. MJH reported that reconfiguring his PC to VLAN 900 had resolved the problem. MJH and JA would discuss the approach for the rest of the IC inhabitants. (Since the meeting, new networking hardware has been installed which solves the problem and so the decision is to retain the VLAN we currently use in the IC.)

JA reported that IH have now installed the two Gbit 48 port network switches purchased for the STRI. This enables all desktops to have a Gbit link to car-server and the /home and /data filesystems. This should increase transfer speeds and the reduced latency should improve users' interaction with the desktop. JA would produce a list of which ports are patched to the GB switches so that users can ensure their desktop is plugged into the correct port. (Done since the meeting.)

JA felt there had been no significant problems with the system and this was supported by the reps.

JA said that the yearly OS upgrade (to FC18) would take place within the next few months, after testing and updates had "settled down".

3. Cluster report

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

The cluster move to the De Havilland campus took place at the end of 2012 and was largely successful, with most of the cluster hardware being up and running again by the end of the week of the move, just before the Christmas break. One infiniband switch and one of the RAID controllers on the storage array had failed and would be replaced. (At the time of writing, the infiniband switch has successfully been replaced, and downtime is scheduled to replace the RAID controller.)
The head node and all compute nodes (except for the sandbox machines) were now running FC16, and the stability of the cluster was distinctly improved by the kernel upgrade.
The cluster continues to be well used, now including the extra CAR dedicated nodes referred to in the last minutes. Other recent upgrades included a new server dedicated to CAIR which takes some of the load away from the head node. CAR typically takes up 50-70 per cent of the cluster usage, and the vast majority of the usage is now by PAM users.
A renewal of maintenance on the original cluster (first 5 chassis) was now being negotiated with Dell.

MAT asked about usage of the SMP machines. MJH said that these were generally under-used with respect to the rest of the cluster, though some CAR users are using them as data crunching machines. They continue to be the most powerful single machines we have.

Reps asked about the effect of new staff on the cluster. MJH reported that he was not aware of any significant additional demands on the cluster that would be made by the new CAR/PAM staff.

Reps asked about system maintenance; MJH explained that most of this could be done remotely, supplemented very occasionally by visits to DH in person.

4. Cancellation and archive policy

JA reported that the CAR central systems are now close to holding more data for people who have left than for people who are still associated with CAR. The committee discussed a formal policy for the cancellation of accounts and the deletion of data owned by people who have left. They agreed that a policy was required and that the starting point should be that accounts are locked soon after a user leaves and that data should be deleted some period (three months was the consensus) after the leaving date. The committee agreed that given the volumes of data involved it was not practical to archive data; deletion was the only option. Exemptions to the policy needed to be justified (visiting fellows/lecturers/professors would be exempt).

Action: MJH to draft a policy based on these principles and circulate to committee members for approval, then forward to Janet/the MC.

5. Matters from reps/AOB

MAT asked if cutting-edge Starlink software could be installed on the CAR systems; JA reported that this had just been done.

6. Date of next meeting

The next meeting should be in early summer 2013, ideally before the end of the UH FY so that the committee could have input to spending plans.

Action: MJH to circulate a Doodle nearer the time.

The meeting closed at 12:00.