CERN/FOCUS 2001-006
12-November-2001
Present:
J.Altaber, P.Anderssen*),
D.Asbury, J.Boucrot, T.Cass, M.Cattaneo (Secretary), M.Ernst, F.Etienne*),
F.Gagliardi*), R.Gokieli,
F.Hemmer, H.-F.Hoffmann, V.Innocente, P.Jeffreys (Chairperson), J.Knobloch,
M.Marquina, H.Meinhard, S.O’Neale, M.Pimiä, K.Safarik, J.Shiers, A.Silverman,
P.Vande Vyvre, I.Videau, W von Rüden
Invited:
J.P.Baud (representing H.Renshall), P.Buncic (representing A.Sandoval),
R.Cashmore*), D.Duellmann,
D.Heagerty*), A.Hirstius
(representing A.Norton), M.Lamanna, I.Papadopoulos (representing A.Grant),
G.Poulard (representing N.McCubbin), T.Smith*)
Apologies:
M.Delfino, B.Gobbo, A.Grant, L.Mapelli, N.McCubbin, A.Norton, H.Renshall,
A.Sandoval, E.Valente
Absent:
D.Jacobs, M.Kienzle, W.Lerche, J.May, M.Mazzucato, L.Robertson
*)
part time
1.1
Consideration
of Agenda
1.2
Minutes from last meeting and matters
arising
1.3 Chairman’s comments
2.
New FOCUS terms of reference
a.
Chairman
b.
CERN Director for Scientific Computing
c.
IT Deputy Division Leader
3.
Reorganisation of IT division (W. von
Rüden)
4. Mass Storage issues
a. User Experience with Castor (J.Boucrot)
b. Status of migration to Castor (J.P.Baud)
c. Charging policy and plans for 2002 (J.P.Baud)
5. Experience and feedback from running experiments
a. NA48 (A.Hirstius)
b. NA49 (P.Buncic)
c. COMPASS (M.Lamanna)
d. HARP (I.Papadopoulos)
6. SUNDEV status and plans
a.
User feedback (V.Innocente)
b.
IT division plans (T.Smith)
7. Update on ongoing IT activities
a. Wireless LAN deployment status (P.Anderssen)
b. Security update (D.Heagerty)
8. Actions from last meeting
9. A.O.B.
The Chairman began by apologising for the
change of date for the meeting and hoped to convince the membership that its
postponement to All Saints day was worthwhile. The agenda, including last
minute changes made the morning before the meeting, was approved without
comment.
The minutes had already been approved by E-mail. There were no further comments.
There have been varied discussions about
the future of FOCUS. The Chairman does not want FOCUS to terminate through
apathy, and was pleased that there would be two opportunities to discuss its
future. He hoped to have a full discussion at this meeting, with final
agreement at the December 6th meeting.
The two main items on the agenda are good
examples of what we have considered to be central FOCUS business. CASTOR, LSF
and data storage are generic services that affect a large cross-section
of experiments. For example CASTOR has been a regular item on the FOCUS agenda,
and we are now at the stage of the project cycle where we are looking at
production issues.
It would therefore be useful to bear in mind this meeting’s agenda when discussing the future of the committee.
The Chairman distributed a draft revised FOCUS mandate, intended to be a starting point for discussion, and wanted to hear the views of people around the table. An original proposal had been made by IT Division and was discussed at COCOTIME. The new draft has attempted to take into account the feedback from COCOTIME, from E-mails by Richard Gokieli and Helge Meinhard, as well as from numerous discussions.
Hans Hoffmann explained the current financial difficulties at CERN. A paper proposing to save 20M CHF from the Materials budget has not been accepted by the funding agencies, who have asked the management to look into more drastic cuts in the CERN programs – this may lead to an external review of CERN. In this context, the future of FOCUS depends on its relevance to LHC computing: is FOCUS only concerned with computing for current (Fixed Target) experiments? If so, what happens to FOCUS if there are drastic cuts in the Fixed Target program?
The Chairman wished to clarify that he is proposing a role not necessarily restricted to the Fixed Target program, but rather an overseeing role over the services provided by IT. This is in contrast to the SC2 committee, which looks at the longer term planning of the LHC Computing Grid project.
Over the coming months there will be a major reshuffling of programs at CERN, affecting as little as possible the LHC programme. Hans encourages FOCUS to discuss, and to go along with the Chairman’s proposal. Given the coming changes and the emerging role of SC2, he suggests that the mandate be re-discussed in a year from now.
Following Hans’ comments, the Chairman read out his draft revised FOCUS mandate, highlighting that FOCUS would continue its activity with a somewhat reduced scope, and a possible reduction in membership. This mandate will have to be reconsidered at the end of 2002 (ACTION).
Wolfgang von Rüden pointed out that FOCUS concentrates on IT services for the research sector, but it must not be forgotten that IT also provides services for the whole laboratory. He thinks that FOCUS is important and would like to see it as the place where activities are reviewed, according to a well-defined calendar fixed well in advance. He also would like to see more input from the experiments. He considers that when a committee such as FOCUS is well accepted by the people it advises, they will follow the advice. The Chairman acknowledged all the points made by Wolfgang.
Steve O’Neale asked whether the budget over-runs for the LHC machine, experiments and computing would be addressed separately, or as one. Hans Hoffmann replied that the member states understand that the ultimate goal is the physics results, and that all three components are necessary to achieve that goal. The physics ambitions of the LHC experiments will be discussed in the LHCC, so as to review scientific issues not addressed by the computing review, such as the additional trigger rate from b-triggers, or the expected complexity of heavy ion events physics. The sharing of computing costs between CERN and the member states, possible savings and other funding possibilities must all be explored, for example by increased CERN participation in EU projects.
The chairman asked Helge Meinhard and Richard Gokieli if they were happy with the modified mandate. This was the case, it being understood that the proposal to merge FOCUS and COCOTIME had been withdrawn. Richard suggested that the wording of point 5) was too rigid, and proposed changing this to “FOCUS plans and operates across a medium range timescale, typically two years”. The change was accepted (ACTION) – the justification for a two year time scale is that it corresponds to a “generation” of change in IT technology.
Pierre Vande Vyvre asked whether the SC2 committee would oversee activities such as data challenges. Wolfgang von Rüden replied that SC2 would take some time to become operational, and in any case has a longer-term view. Detailed operational matters will continue to be handled by FOCUS and COCOTIME, taking also into account non-LHC activities. Hans Hoffmann stressed that FOCUS is concerned with services provided by IT to the experiments, including the LHC experiments.
The Chairman proposed to wait until the next meeting before final agreement on the revised mandate. On the assumption that FOCUS continues as a forum concentrating on services, a new chairman has to be found, and the help of the experiments is solicited for this search (ACTION). Ideally the chairman would be young (in his or her 40s), from an external institute but familiar with CERN and at least one of its experiments, and able to devote 3-4 days, 4 times a year to this activity, for a 3-year term. Hans Hoffmann underlined his wish to change generation. The post is an opportunity to influence IT policy at CERN.
Proposals for candidates are requested by 16th November, after which a shortlist and recommendation will be presented to the DG, who will hopefully decide in time for the next meeting.
Wolfgang presented the new structure of IT Division. A reorganisation was
necessary to absorb the new people joining the division in the context of the
LHC Computing Grid project, and to replace people who have left the divisional
hierarchy.
Jacques Altaber remains Deputy Division
Leader in charge of Infrastructure and Operations, comprising the existing US,
CO, IS and CS groups, and a new Product Support group (PS, headed by Alan
Silverman) which incorporates the
engineering product support activities of
CE group and deals with all external products including CVS, the PTTools
service and SUNDEV. IS group absorbs the PC technology section of the former TS
group.
Wolfgang von Rüden becomes Deputy
Division Leader in charge of computing for Physics, comprising the existing DB
and API groups, and three new groups issuing from the former PDP and TS groups.
Data Services (DS, headed by Harry Renshall) is responsible for CDR, Castor,
file systems, disk and tape services. Architecture and Data Challenges (ADC,
headed by Bernd Panzer-Steindel) is responsible for R&D in farm technology
and components, Linux OS and software services (formerly IT/PDP/AS section),
and the WP4 activities. Fabric Infrastructure and Operations (FIO, headed by
Tony Cass) absorbs the TS group activities (including PC sales) and the physics
production services (formerly IT/PDP/IS section).
Chris Eck has been appointed to the new
post of Resource Manager, responsible for purchasing, contract handling,
dealing with SPL Division, as well as contracts with the member states sending
manpower to the LHC Computing projects. Chris will also serve as IT/DI group
leader, replacing Jean-Clause Juvet.
Jacques Altaber commented that IT has
been very active trying to consolidate activities, and is pushing for further
consolidation in order to accommodate the CERN-wide personnel reduction and the
pressures on the budget.
Vincenzo Innocente was surprised to see
responsibility for SUNDEV in PS group, while the other PLUS services are in FIO
group. This is because there is expertise in Solaris tools coming from the
former CE group; it was also pointed out that the asymmetry will disappear with
the phasing out of the RISC platform, which it is hoped to accomplish much
earlier than the end of 2003 as originally planned.
Fabrizio Gagliardi would like to
understand how activities in the three LHC computing projects are coordinated
with and supported by the IT groups. A meeting of the group and project leaders
is foreseen to address these concerns.
Helge Meinhard asked for clarification on
the split of PC procurement activities between IS and FIO groups. IS group is
responsible for drawing up technical specifications and for validation and
certification of new models, which are then delivered to CERN customers by FIO
group. Helge was worried that IS group might focus on Windows certification and
neglect Linux – Jacques Altaber assured him that the certification activity is
hosted by IS because the Windows certification is more complex, but that the
Linux experts will be strongly involved in the process, as is already the case.
Steve O’Neale would have liked to see
programming language and compiler expertise consolidated in one group. Wolfgang
replied that this is the case, existing expertise is currently in API group,
but it is recognised that more expertise is needed in this area and is a
priority for future hiring.
Jacques Boucrot described Aleph’s heavy use of CASTOR in 2001, for the data migration from Redwood tapes (6 Tbytes of existing data), and for the final Monte Carlo production (16 Tbytes of new data). The MC data produced at outside labs was copied to CASTOR at CERN. The CERN production consisted of 200,000 jobs submitted in 6 months with an (unacceptable) LSF failure rate of 15%, which required the development of a sophisticated tool for job submission and book-keeping.
General impressions concerning CASTOR are very positive, in particular the ease of migration of Aleph software to CASTOR. The main criticism concerns the lack of monitoring which makes it very difficult for non-experts to understand problems. Another problem is an insufficient allocation of tapes, which led 4 times to a complete blockage of the production. Other positive points are the performance of the Linux machines, a huge progress in AFS stability, and the fast and efficient help of the IT experts.
Unfortunately, due to the recurring LSF problems and problems with earlier versions of the CASTOR software, this huge production was very manpower intensive, much work is still needed to obtain a well monitored, user friendly system free from failures.
Steve O’Neale confirmed that Opal had similar experiences, and stressed that early problems with CASTOR are unrelated to the recurrent LSF failures. Wolfgang von Rüden agreed that the level of failures is unacceptable, and believes that many of the problems have disappeared in the latest version of LSF, which is ready to be deployed in coordination with the experiments.
Jean-Philippe Baud reported on the current status of CASTOR. The latest release is very stable and fixes all but one of the known bugs. 350Tbytes of data are stored in CASTOR, in 3.5 million files. One problem is that the system was designed for large physics data files, but many users are storing small files in Castor (mail folders, Netscape cache!)
The SHIFT compatibility mode of CASTOR has enabled experiments to continue to manage tapes themselves (without the need for a copy) but is not recommended for new applications. Next spring the SHIFT compatibility part of CASTOR will be split from the HSM part. The former will be frozen at Easter (no new bug fixes or developments), while the HSM part will continue to evolve.
The Redwood migration is about 50% complete. It has now been put on hold while waiting for much cheaper media (expected middle of next year) and to avoid overloading tape drives during the move of tape silos to a new building.
Migration of HPSS to CASTOR is also well underway. A script “h2c” allows individual users to migrate from HPSS to CASTOR. Migration should preferably be completed before end 2001, and by March 2002 at the latest.
Jean Philippe believes that charging is necessary to optimise the utilisation of hardware resources. The system can do some optimisation itself, but much optimisation requires action by users (e.g. avoid small files (<10MB), avoid large directories, use large disk pools). The system implements no restrictions, but can be inefficient if incorrectly used. Quotas or a fair share algorithm may be needed to avoid users flooding tape queues.
The proposed charging formula takes into account the space occupied on tape (2CHF/GB, possibly dropping to 0.8CHF/GB next July, after hardware compression and excluding “deleted” files), tape I/O activity, and tape device time (to favour lightly loaded disk servers). The weights of the latter components in the pricing formula must be agreed with the users.
Many recent complaints from users have been due to a shortage of tape drives. 2 new STK silos are being installed and a robot will be acquired to test LTO and Super-DLT technology, and help with data challenges. A tender for new drives is being issued.
Future development plans for CASTOR include support for files larger than 2GB, access to Castor data through Grid middleware, a CASTOR file-system (i.e. access to CASTOR files without requiring RFIO), improved monitoring and statistics, support for DVD and self describing tapes (i.e. possibility to read tapes without access to Castor catalogue).
The Chairman suggested that ideally the system should not be optimised by users being required to adjust their activities, and asked what steps are being taken to optimise automatically. The monitoring system is active, so it can take some decisions. In general, the strategy is to optimise if and where there are problems.
The Chairman also asked what charging is supposed to pay for (what fraction of the full economic costs). It covers media costs, plus an incentive for good practises. It is not intended to cover the cost of tape drives and robotics.
Helge Meinhard pointed out that stage disks are not paid directly by the experiments but by COCOTIME (i.e. by IT money anyway), so the charging algorithm is just an incentive to experiments to ASK for more disks, not to pay for more. It was suggested that it would be better to have a centrally managed disk pool. Jean-Philippe agrees and thinks it is that management of the pools by the experiments is the wrong approach.
David Asbury suggested that the number of tape mounts should also be in the charging algorithm. Jean-Philippe prefers to let this be optimised by the system.
Wolfgang von Rüden asked whether limits should be imposed on this very open system. Steve O’Neale objected that this is not really possible, as there are always reasonable exceptions, but Vincenzo Innocente warned that leaving the system open is asking for abuse, e.g. to bypass the AFS quota; this is particularly true if Castor becomes a file system. If file size is the problem, a lower limit could be imposed.
Jean-Philippe suggested that there could be two categories of files – user files (e.g. n-tuples) with lower limits on their size, and production files with no limit, accessible in write mode only from specific accounts. Helge Meinhard welcomed the file system access to files, but suggested that this should be read only by default to avoid misuse by non-experts. Access rights could be an alternative solution; it would be a shame to force users to store n-tuples via a multi-step process.
Helge Meinhard suggested that a thorough investigation be made into everything that is necessary to be able to store files larger than 2 GB, as CASTOR is probably not the only limiting factor.
Andreas Hirstius expressed interest for the DVD developments and asked when this would be available, as this would be very interesting for import/export. A prototype will be available before Xmas, but it was stressed that there are no plans to turn this into a service, due to manpower constraints. It is necessary to understand the relative importance and priority of this request. Also, CERN has no DVD equipment; the development is done outside CERN.
Finally, Wolfgang announced that two more people will join the CASTOR team in January, bringing the total to 5 FTEs. Jean-Philippe will make a series of presentations on CASTOR in January.
FOCUS thanks Jacques and Jean-Philippe
for the informative and encouraging talks. It is pleased to recognise the good
performance of Linux and AFS. It notes the substantial improvements in CASTOR,
in particular the easy adaptation of the Aleph software, and the failure rate
of 15% with LSF. FOCUS is keen to see the developments foreseen for CASTOR
addressed aggressively, and hopes that the extensive work on LSF will lead to
an improved service. FOCUS supports the goal to charge for managed space, and
to allow the experiments to optimise the provision of this space for the user.
It is pleased to note the increased staff resources invested in the CASTOR
project (5 FTEs from next year).
Andreas reported on the performance of the NA48 CDR system, which in 2001 produced 80TB of raw data, with a total bandwidth to tape of 30MB/s. Of 22 Dual PIII CPUs, 20 were borrowed from LXSHARE, as well as one disk server. Experience with these shared resources was very good. Experience with tape drives was, however, mixed. The drives were reliable and the bandwidth to tape sufficient, but there were sometimes long queues waiting for a drive, and the disk buffering could only just cope with the data rate. This was solved by dedicating tape drives to NA48.
Predrag explained that NA49 has a dedicated farm used since 1994 for its CDR and production. In 2001, however, a new data-taking configuration meant that additional resources had to be found. The additional LXSHARE resources made it possible to complete the backlog of year 2000 data processing, and to complete the year 2001 processing only a few days after data taking. He stressed that this was successful largely because planning for use of the LXSHARE resources started a long time in advance, with formal interaction with IT division in the preparation phase allowing solutions to be found efficiently, particularly concerning early problems with CASTOR. He also highlighted the need for efficient monitoring, which NA49 achieved by writing a fully automated stage-submit-recovery procedure. Overall, NA49 are satisfied with the LXSHARE model and the level of support.
COMPASS had foreseen the use of up to 100 CPUs from the shared pool, but eventually used only 20. These 20 however were essential in the last phase of data-taking. Conversely, COMPASS had not foreseen the need for any additional disk servers, but problems with its SCSI disk servers in August meant that an emergency allocation had to be made from the shared pool. This led Massimo to conclude that the sharing model for resources is fine both for scheduled peaks in requirements (e.g. large productions) and for just-in-time emergency use. In addition, the shared resources were also used (as scheduled) to make large scale tests of concurrent access to Objectivity databases stored in Castor.
Massimo also reported unsatisfactory and confusing behaviour from LSF (up to 30% of lost jobs), major improvements in CASTOR since 2000, and problems with data retrieval, where the insufficient number of tape drives meant that not all data could be retrieved for detailed monitoring. He concluded that software is developing in the right direction, but needs more improvement.
Ioannis reported on the first data-taking run of HARP. The computing was based on standard components both for the data flow hardware and for software. The data flow was rather stable, with only occasional CASTOR problems fixed with software upgrades. A major problem for many physicists was the large access times to tape resident data (up to 5 hours reported), and the fact that no tape drive was dedicated to HARP; next year the reconstruction production will be run in parallel next year, so this problem will have to be addressed. The database and analysis software support was excellent.
The Chairman asked IT division to comment on the availability problems with tape drives and the need to dedicate them to experiments. Jean-Philippe Baud replied that 28 drives were available for production, and that they had to be dedicated to high rate (30MB/s) experiments, otherwise they would lose data. Allocation of drives to lower rate experiments is not affordable, and would lead to poor access times for all, due to the limited pool available. A possibility would be to buy more drives, for example LTO.
Helge Meinhard pointed out that tape services were discussed in COCOTIME. All drives were used at 100%, with no contingency. Possible solutions are to buy more drives, and to better understand usage. The latter was done by COCOTIME, there is potential for improvement and COCOTIME has the commitment of the users concerned to alleviate the problem. COCOTIME wants to be convinced that drives are used optimally before proposing expenditure on new drives.
Wolfgang von Rüden felt that this year was particularly bad due to drives being committed to the Redwood migration, and to the tape silo move to a new building. New purchases next year should improve the situation.
FOCUS thanks NA48, NA49, COMPASS and HARP
for this feedback. It recognises the crucial role of robust job monitoring and
accounting in production analyses. In addition to the comments made under the
previous item, particularly with respect to LSF, it notes the problems
encountered with the large access times to data on tape, and asks IT to propose
a solution for the March meeting of FOCUS (ACTION)
Vincenzo briefly reviewed the justification and the tools required for a second development platform in CMS. The migration to SUNDEV first required a migration of the software to Solaris 2.7 and CC 5.x. This took longer than expected both for the CMS software and for the IT/DB and IT/API software. A first scheduled date for migration to SUNDEV in June was missed due to problems with LSF. CMS require a standard environment for all categories of users, and were not ready to migrate to a machine that was felt to be a validation machine for the OS itself. The green light for the migration was given in August.
Tim described the hardware configuration of the new SUNDEV service. The 8 machines are currently allocated to the interactive service (5 machines), a build service (2 machines) and one machine of operating system support. This configuration has been available to experiments since April, with 60 distinct users in recent months. Discussions with CMS have taken place since June, to iron out problems on both sides – the go ahead for the migration was given in a CMS/IT meeting just minutes before this presentation!
Two additional machines are available, but require Solaris 2.8 – these are not yet in production status. More of these machines have been agreed by COCOTIME for 2002, based on experiments’ estimates of their needs. Solaris Support service machines will be added to disentangle OS support from the service.
Vincenzo was pleased to learn about the
new Solaris Support machines. He hopes that commissioning of Solaris 2.8 will
not be as painful as it was for Solaris 2.7. Steve O’Neale pointed out that a
background batch service would be useful on these machines – night time batch
queues will be opened.
Pal reported on progress with the wireless LAN pilot service. The installations in conference rooms have proven useful and much in demand, though there are problems with the power supplies of the base stations. More seriously, there are concerns about the security and privacy of wireless communications.
A tentative plan for the near future is to allow access only to DHCP requests from registered MAC addresses. Unregistered PCs will only have access to the service Web page explaining how to register. PCMCIA cards will be distributed by stores and charged to the users. A manufacturer of base stations will be selected, but their deployment will be constrained by the financial resources available.
The Secretary asked when users could expect the pilot to become a service; no firm commitment was made, due to manpower constraints, but Easter 2002 seems a possible target date. Alan Silverman requested a web page describing the settings required to use the service – he had to change his settings while travelling and found it very difficult to return to the CERN settings.
Denise made an update of the security
incidents so far this year, in particular the recent break in to PCs running
old unpatched versions of the Linux OS. This is a potentially very serious
problem since there are large numbers of such PCs on site. She underlined the
importance of promptly applying all security updates. She then went on to
describe the project to strengthen the firewall, which was first introduced in
June but had to be rescheduled to this week due to instabilities. One new
possibility is to selectively disable high numbered ports, following
discussions with affected parties. The possibility remains to disable all high
numbered ports in an emergency.
A long discussion ensued over the Linux
security incident. Jürgen Knobloch questioned the necessity to fully reinstall
these PCs – this is the only safe option, as one cannot know what an intruder
might have modified. The Secretary expressed concern about the logistics of
updating the large number of PCs affected, in particular since many systems are
under the control of individual users. Massimo Lamanna asked whether an attempt
had been made to contact individual users – the list of affected PCs can be
monitored from the network scan information. Denise feels that such mails would
be largely ignored, and prefers to go through the management chain, but it is
clear that a strategy is needed in the long term to enforce compliance with
security requirements. Helge Meinhard pointed out that some specialised
installations are not trivial to be redone; this is not helped by the fact that
the standard installation simply assumed the standard desktop hardware
configuration.
Jürgen Knobloch asked whether something
might be learnt from other sites. It was felt that the priorities and access
policies of different labs are different, but there are certainly ideas to be
investigated, in particular when users request new functionality; for example
one-time passwords could be an option for giving remote access to Windows
files. Vincenzo Innocente suggested requiring ssh instead of telnet for
connections from other sites, but this would limit access from portables also.
It was also argued that ssh does not solve problems, since CERN passwords can
still be sniffed at remote sites.
The discussion was curtailed due to lack
of time.
Minuted/Section
|
Action
|
Who
|
Status
|
|
08/06/2000-18/6 |
Request experiments
to specify requirements for large shared LHC computing test-bed |
M.Cattaneo, P.Jeffreys |
No longer
responsibility of FOCUS. CLOSED |
|
03/05/2001-21/3 |
Decide on need
for a central CVS server |
Experiments |
Included in mandate
of new IT/PS group. CLOSED |
|
03/05/2001-21/5.1 |
Plan annual
Security report at FOCUS 25 |
Secretary |
|
|
27/06/2001-22/2 |
Prepare new
terms of reference for FOCUS, in conjunction with HEPCCC |
M.Delfino, H.Hoffmann, Secretary, Chairman |
See these
minutes, section 2. CLOSED |
|
27/06/2001-22/2 |
Plan file
sharing discussion at a FOCUS meeting in 2002 |
Secretary |
|
|
27/06/2001-22/4 |
Review Linux
strategy and developments at FOCUS meeting early in 2002 |
IT Division, Secretary |
Planned for
FOCUS 25 |
|
27/06/2001-22/7 |
Prepare
deployment plan for Wireless LAN, to be presented at FOCUS 23 |
IT Division |
See these
minutes, section 7a. CLOSED |
There being no other business, the meeting was closed at 18:00.
Minuted/Section
|
Action
|
Who
|
Status
|
|
03/05/2001-21/5.1 |
Plan annual
Security report at FOCUS 25 |
Secretary |
|
|
27/06/2001-22/2 |
Plan file
sharing discussion at a FOCUS meeting in 2002 |
Secretary |
|
|
27/06/2001-22/4 |
Review Linux
strategy and developments at FOCUS meeting early in 2002 |
IT Division, Secretary |
Planned for
FOCUS 25 |
|
01/11/2001-23/2 |
Revise FOCUS
mandate at the end of 2002 |
FOCUS and CERN management |
|
|
01/11/2001-23/2 |
Modify wording
of point 5. of the proposed mandate |
Secretary |
|
|
01/11/2001-23/2 |
Propose
candidates for post of FOCUS chairman, by 16th November. |
Experiments |
|
|
01/11/2001-23/5 |
Prepare a
solution to the long delays for accessing tapes experienced during 2001, to
be presented at FOCUS 25 |
IT Division |
|