La Silla Observatory Science Operation Department DRL replacement to the w*off machines User's Requirements Prepared 2003-03-29T01:29:31 Sat ohainaut Reviewed 2003-04-15T11:58:07 Tue glocurto Released -INTRODUCTION Currently, the w*off (wg5off, w3p6off, w2p2off) machines are HP workstations that receive the raw data from the w*dhs machine. They are meant for the observer to work on his data. They feature a set of data processing tools and recipes. They are also used to store Service Observing data before their distribution, and to store technical data being processed by the various instrument scientists. This document describes the requirements for a Linux machine replacing the HP workstations. -REQUIREMENTS COMMENT (EPO): The needed VLT software (GASGANO, CMM, DataSubscriber...) has to run under linux well and in a stable way. All this VLT software should be officially supported by Garching||DMD||SWC on PCs with linux O.S.. Software should be maintained by ls-SWC, in a client-server configuration: a mirror script should read the configuration from the "server" and duplicate it to the other two (supposedly we will have 3 machines) "clients"; this system is already running in Paranal. REPLY: agreed -- HARDWARE * CPU: state of the art, typically a Pentium-4, around 2GHz clock COMMENT (RME): WFI we should have a system that can expand to a multi-processor (4) CPU. These are available on the market. REPLY: agreed on multiprocessor capability, but do we really need 4 ? Before installing more processors should be cheked whether the software (IRAF, MIDAS etc.) is able to run in parallel in more than 1 CPU. * MEMORY: minimum 512Mb, ideally 1Gb fast RAM COMMENT (RME, EPO) While for other instruments 512MB is fine WFI needs 1GB fast RAM. REPLY: agreed * DISKS: we consider here only the disk space for the users. The operating system etc are not considered. Volume should be the most important parameter, i.e. select IDE over SCSI. - a "/home" area large enough to host the home directories of the various users. These home dir. shall contain only "text", no "data", i.e. account configuration files (.tcshrc, .mozilla...), text and documents, local programs sources and executable. Some users are permanent (i.e. these account will be backed-up/modularized), some others are temporary (i.e. will be zapped and reinitialized periodically). 24Gb - a "/data" area for incoming data. It must be able to hold at least 2 full runs (typically 6 nights) of raw data and data processed by the astronomer, i.e. 6n * 5Gb * 4 (each frame existing in 4 steps of data reduction), i.e. minimum 120Gb for NTT/3p6. For 2p2, the equivallent figure (6n * 30Gb * 4 = 720Gb) might be beyond what is available on the marked. Select the largest, reliable disk available (240Gb?) COMMENT (JPR, EPO): we should consider a RAID array for WFI to reach the planned 720Gb (ask suggestions to NGAST ?) REPLY: agreed COMMENT (GLO): HARPS will record (on the basis of the commissioning 1 statistics) from 5Gb to 10Gb per night. A 240Gb disk volume is therefore requested. - a "/sciops" area for storage of and work on SciOp data, i.e. service observing, technical data, etc. It must hold at least one period of Service Mode (NTT/3P6: 70Gb) and leave ~70Gb of space for data processing, i.e. >~140Gb for NTT/3P6. For 2P2, the Service data are not stored locally, but the data processing requires more space, so similar (>~140Gb) disk should be provided. * EXTERNAL DISKS: the machine should have USB2/Firewire slots to hot-plug some external disks, such as - Instrument disks: each instrument shall have a ~100Gb disk for storrage of calibration plan data, work in progress, etc... under the responsibility of the Instrument Scientist. - Visitors' disk, in case they have a compatible disk. The operating system should be configured in such way that it recognises and mount automatically these disks, and permits the users to umount/remove them without root privileges in a documented way. COMMENT (GLO): We should have at least 2 USB2/Fireware slots per machine to avoid conflicts between users/visitors. * DISPLAY: large, high-res, flat screen. Dual display would be an advantage but is not critical. COMMENT (JPR): I would strongly support dual head systems. I note that the Dell Precision 650 with nVidia, Quadro4 900XGL, 128MB, VGA/DVI comes dual monitor capable. REPLY: dual display would be a nice feature, but is not critical. However it would be wise to choose dual monitor capabale platforms like the ones John is suggesting or MacIntosh (with linux & Scisoft). * CD/DVD writer. COMMENT (JPR): >> As there are several standards for DVDs (DVD-R, DVD-RW, DVD+RW...) it is >> important to check which standard the drive will support. These days it >> is possible to get drives compatible with more than one standard (though >> not necessarily from Dell), e.g. the Sony DRU-500A Dual RW DVDRW, see >> for example http://www.cdrlabs.com/reviews/index.php?reviewid=154 REPLY: agreed. COMMENT (RME): * DLT writer for w2p2off is required. REPLY: will this requirement stand also after we will have a central backup office in La Silla ? COMMENT (GHA): >> There should be a read-only CD/DVD reader and an CD/DVD burner (I would >> opt for the DVD burner since there will be no need for any upgrade, and >> astronomers can burn their data in one go if they have multiple nights. >> Same for the TiOs so means less work). The reason why there should be a >> separate read-only CD/DVD reader is to avoid having to dismount the CD >> when using >> xcdroast. If we have only one CD/DVD writer It is easy to have conflicts >> if the kernal tries to mount the CD, whereas xcdroast want to write to >> the CD. REPLY: I beleieve we could find a better solution than buy read-only CD/DVD to avoid conflicts in the system when using xcdroaast. A possibility would be to start xcdroast from a script which checks that all mounting is done correctly. COMMENT (GHA): >> There should be a scratch area for the CD images, or that xcdroast is >> configured to use /tmp or /data (this requires root permission). REPLY: agreed. COMMENT (GHA): >> About the CD/DVD burner/players -- if the computer will be physically in >> the computer room, we should consider using external (fireware or USB) >> CD/DVD burner/players which are accessable from the desk, and does not >> involve the user going to the computer area. This also may be configured >> as part of the auto-detection of external hard disk setup. REPLY: desk accessibility to CD/DVD burner is desirable but not critical. We should investigate possible solutions. COMMENT (GHA): >> I would advice against using the "ergonic keyboards". Also, the >> keyboards must be US standard. REPLY: agreed. -- SOFTWARE * Standard operating system (recommended by VLT, currently RH-7.3), with standard applications. This must include configured EMACS, VI LaTeX, dvips, a2ps xdvi, ghostview, acroread Mozilla xCDroast ADDITIONAL REQUESTS: Xemacs, jed (JPR, EPO) cdrecord suite (JPR) Spreadsheet (Staroffice ? Openoffice ?) (EPO, GHA) cdlabelgen (GHA) xv, xfig, (EPO) IDL (GHA, GLO) P2PP -noccs (GHA) CMM (JPR) lyx (see www.lyx.org) (ISA) REPLY: most of the additional software is free/shareware so, no problem. However it will have to be in cluded in the software maintenance list. IDL is a must for TIMMI2 (or we will have to redo an "offline pipeline"). * SciSoft updated to the latest available version, with local configuration (printers visible from the application softwares, IDL licenses, IRAF pipes, etc...) and patches (infamous /scisoft/saord/bin/access conflicting with LateX). * dhsSubscribe installed and configured to transfer all raw data from the w*dhs machine. In a next step, we may add a secondary dhsSubscribe for pipeline processed data. dhsSubscrive can be left with its "default" options (ie store the data in /data/raw/, DHS filenames, no post-arrival script) -- ACCOUNTS In addition to the various system related accounts: * sciops: used for installation and owner of sciops-related programs (eg MIDAS scripts). This is the equivallent of the nttops or e3p6ops accounts on the current w*off * astro: used by Sciop astronomers for daily work * visitor1, 2, 3: used by the visiting astronomers during their runs. These accounts will be zapped and re-initialized for each new visitor. These accounts could (should ideally be) the current lsusr* accounts delivered to the visitors on the off-line DRL system, PROVIDED that the access time to their home directory is much faster than currently. ADDITIONAL (set up) COMMENTS (GHA): >> The paths should include a /home/user/bin as default so that additonal >> programs can be installed by us. >> There should be a scratch area for the CD images, or that xcdroast is >> configured to use /tmp or /data (this requires root permission). >> It should also be decided whether KDE or GNOME should be run. At the >> moment the different drl machines are configured differently. REPLY: agree, but users should avoid to install unsupported software or they should ask the requested softwarto be entered in the supported software list. --oOo-- ### Gaspare Lo Curto - 3.6m telescope, La Silla ### ### European Southern Observatory ### ### Alonso de Cordova 3107 ### ### Casilla 19001 | phone:+56 2 463 3088 ### ### Santiago 19, Chile | fax :+56 2 463 3001 ###