--------------------------------------------------------------------------- SUMMARY OF AIAI/IRR INPUT TO COMPUTING COMMITTEE ON 14-DEC-99 --------------------------------------------------------------------------- 1. Access to suitably powerful compute servers from any desktop is requested by IRR students, and Unix system access from PC/Windows desktops is requested by AIAI. 2. An exploration of the issue of multiple site support is requested. 3. A means to allow connection of laptops by staff and students is requested. 4. There is a concern over the centralised nature of the controls on mounting systems on the servers accessible to Institutes. This leads to redundant mounting of software in personal or inappropriate areas for sharing. There is a need to mount and test software as part of our research that may be inappropriate for centralised and stable support. We should explore the idea of an Institute "experimental server" in a ring fenced "safety zone" under Institute controls that is not part of the stable service. 5. Support for PCs with Windows 98 (moving to Windows 2000) as well as Unix/Linux is requested by AIAI. 6. The ability to import, create, modify and export documents in up-to-date MS Word, Excel and Powerpoint formats is required (under Unix/Linux by IRR, and on Windows 98 in AIAI). 7. Improved tracking of problems or requests and public information about their status is requested. 8. An understanding of the nature of the duties and expected level of support in terms of both personnel and types of systems included and excluded from support is requested. See my previous suggestions for a framework and a process for this at http://www.aiai.ed.ac.uk/~bat/COMPUTING/policy-suggestions.html 9. Informatics is seeking to recover 1800 pounds per annum per researcher funded by external grants and contracts. I is proposed that this be centrally managed for the benefit of the research. A document proposal for how this will be used for the benefit of Institutes is requested. --------------------------------------------------------------------------- OVERALL --------------------------------------------------------------------------- Ken Baird - 25-Nov-99 Has written as Service Manager for AIAI and IRR to Head of Division (11-Nov-99) seeking clarification of level of support available to Institutes from Informatics computer officers. He also notes that 1800 pounds should be available per annum per externally funded researcher for a computing services charge that is available beyond the workstation capital costs included in grants. It has been proposed that this be administered centrally by the computing service for the benefit of the individual Institutes; an arrangement he is happy with, which he would welcome progressed and more details provided. Ken noted that responsive support within a short and reasonable period for some types of problem is a requirement for a good service. He commented that this is good at present due to the individuals involved. But it is unclear what the performance target is in general and what can be expected of those served. Ken also noted that support for both Unix and PCs is essential. He noted that MS products on PCs are in common use in AIAI for technical as well as administrative work and that support for these products is important. --------------------------------------------------------------------------- IRR --------------------------------------------------------------------------- Josh Singer - 12-Mar-98 Should move towards powerful compute servers which are automatically regulated to support PhD students. Bonnie Webber - 26-Nov-99 1. We should ask what Computer Office support we need that we are not getting now. Bonnie will mail IRR people to ask this and remind them to make their views known to Austin. 2. We need on-line access under Unix/Linux to a system such as StarOffice which can support the reading, printing, modification, origination and compatible export of Microsoft Office products (specifically MS Word, MS Excel and MS Powerpoint). 3. IRR wish to provide appropriate computing support to graduate students and is currently doing this through shared access systems that are sufficient in number so as not to cause unacceptable delays in access on demand. Assignment of individual machines to students will be considered if it is in the best interests of graduate programs in the Division. Bonnie Webber on 4-Dec-99 in response to Peter Ross on 3-Dec-99 Peter - Excellent points. I especially like your suggestions about - the public list of requests to COs, with an official DATED explicit response to each (I added "dated") - a mechanism for publicising improvements (how about an announcement page within the local Informatics site, since some of us have accounts on multiple machine configurations, since there's no other way of logging on at a different PHYSICAL site) - a list of services offered by COs, possibly with an a priori prioritisation (unless that's totally obvious). I don't have an opinion about a "safety zone" mechanism, since I don't fully understand its implications. Eventually, I would like to see a better solution to the multiple-site computing problem, as it looks like we're going to remain multiple sites for a good deal longer than we had hoped. Austin: Perhaps you can put this on a future agenda. Gordon Reid - 10-Dec-99 On behalf of the Maths Reasoning Group of IRR I would like to request that computing committee formulate a uniform Division-wide policy on the use of laptops connected to the University network. I believe at JCMB there is a 'ring-fenced safety zone' to allow for connection, but at South Bridge, mainly for practical reasons we can't do this yet. I have had requests from 2 students who want to be able to connect laptops to the network. Gordon Reid - 10-Dec-99 Maths Reasoning do have a separate filesystem for our own software installs, but it is no longer the case that lecturers and RAs have root access to install into it. We now use group permissions and shared user IDs for this. (but you do still need a CO sometimes which is where I come in.) It is not on a seperate secure network, as laptops with user as root ought to be, which seems to be the ring-fenced suggestion. I would also advocate that compute servers are a good thing. The question is should they be funded by the Division, Institutes or research grants? (our rolling funding puts us in a special position on this). --------------------------------------------------------------------------- AIAI --------------------------------------------------------------------------- John Kingston - 30-Apr-98 Need to provide good handling of MS Office documents Argued for incorporation of RTF as a supported file format in Informatics. Robert Rae - 23-Nov-99 Provided AIAI Computing Infrastructure Budget and Forward Look dated 20-Jan-97. Still largely valid. Noted continuing pressing needs are: better network which is flaky now colour printing good handling of MS Office documents Peter Ross - 25-Nov-99 Re Linux: I used to use Red Hat, now I run Mandrake, which is Red Hat with optimisations and extensions, and with KDE as the default environment on top of X. Plus it's very much cheaper if you actually go to the trouble of buying the CD set -- 29 pounds as opposed to 50+. Peter Ross - 3-Dec-99 A few thoughts about computing provision, in the form of a short discussion document; and one comment on PhD provision. (a) PhD provision: the phraseology in Nigel's document, that PhDs who take more than three years get to keep the machine, is unfortunate. Nothing should ever suggest that a university-bought machine actually belongs to a PhD student. (b) computing provision. First, it needs to be made clearer just what sort of services are offered by COs generally. Bug-fixing locally-produced software? Training new staff about Microsoft Office stuff? Customising mail environments? Hardware-specific backup and restore services? Ethernet monitoring? FTP and HTTP log statistics reports? etc. Second, it was not clear to me quite what was implied by the idea that grants which included explicit CO support (f/t or p/t) would be "direct-charged". Does this mean (a) the Division will auto-claim the money awarded in the grant for that purpose, or (b) COs will log time spent on a grant, and service runs out when money runs out so the grant-holder has to monitor all grant-specific usage of CO time. Third, re LFCS's request. There has been a long-running tension between academic and student desires to be able to install s/w themselves, quickly, and CO desires to prevent that in the interests of keeping the whole service stable. - On the one hand in ex-DAI the COs did all installation/upgrade work. The plus side: a stable service. The minus side: way behind, takes ages to get the latest stuff installed, especially since COs tended to be hyper-cautious and install new stuff in `option' mode (eg we have four different versions of Netscape in /usr/local/bin at the moment; none of them the latest release. We have gtar and gmake as separate programs rather than using these GNU programs as defaults, so some imported Makefiles screw up horribly. And so on). - On the other hand, in ex-DCS many people had root passwords and did installations. The plus side: current software. The minus side: many installers failed to realise the consequences of their actions and routinely fouled up other people's environments as as result, leading to lots of thrashing by COs. With this in mind, here are some suggestions: i) we need a public list of requests that have been made to COs, with an official and explicit response to each: e.g. - awaiting decision, because it affects X, Y and Z - will be installed by 31 Dec 99 - rejected, too hard to support because [...] ii) we need a good mechanism for publicising improvements. The `alert' mechanism is a loser; as you know, many people are unaware that Java 1.2 is installed (I knew cos I was the one who requested it!) iii) maybe we could have a `safety zone' to resolve the tension between users and COs -- that is, a set of machines on which at least some users have installation powers, but relatively isolated from the main systems. Inside the `safety zone' the service stability is not guaranteed, but users can try things out more quickly and as a result can gather the evidence to justify more widespread dissemination/installation by COs. Over the years, I have routinely installed more up to date versions of s/w within my own filespace (eg my private bin directory currently has 260 programs in it!) and I have typically used s/w for a while before urging COs to do a public install/upgrade -- eg, I have been using the Solaris version of MS Internet Explorer for several months already (and continue to use Netscape too). This methodology has worked OK for me. The `safety zone' idea might be a way to make such a practice more widespread? Clearly some things cannot be installed only within a zone (eg new automounters or other things with network-wide ramifications), but I still feel that the idea has some merit. It would provide COs with a test area, after all.