NorduGrid technical meeting

15-16 May 2006, Copenhagen

Minutes

Participants: Anders, Jakob, Aleksandr, Michael, Mattias, Balázs, Oxana (at the minutes)

CA

NorduGrid CA key extended by 1 year; every site owner and user has 1 week to apply for a new certificate.

0.6

The last of the old serie; after 0.6 start a new architecture, components. New s/w should initially include the old components and interfaces; aiming at smooth transition. 0.6 RC is needed, 0.5.x is not extremelly well tested. Data-proximity algorithm is missing, has to be added to 0.6.+. Aleksandr has something in progress, probably will be ready in 1 week. Detailed discussion to follow on Tuesday.

New projects

KnowARC

Brussels wants to see, in order of priorities:

  1. impact, esp. in industries
  2. service-oriented architecture
  3. a middleware based on standards, capable of working together with other standards-based middleware
Most tasks are computational-Grid related; plus extended Runtime environments; plus workflows. Job migration. New platforms (Mac OS, MS Windows, Solaris). Authorization ramework. P2P-like infosystem. Fault-tolerant storage (NIIF). Logging. Distribution (part of Linux distros). S/w should not be tied to interfaces, should be possible to change them.

NDGF

Common research computing infrastructure for Nordic countries based on Grid. Hosted by the NORDUnet. Based on ARC; must provide resources for Nordic researchers, primarily LCG. Must build a distributed LCG Tier1 (prime concern: storage; experiment-specific application interfaces; accounting towards LCG; T1 <-> T2 transfers). Does not have to follow standards - instead must meet the needs of existing applications (incl. proprietary stuff). Ev. MC production, distributed analysis. Long-term solutions are not a priority.

NORDUnet3

Budget enough for 2-3 PhD students; nothing left for summer schools; probably interactive applications and data management, depending on where the students will end up.

Manpower (development)

KnowARC: Lund 2, NBI 2, Uppsala 2, Oslo 2 NDGF: 6, distribution yet unknown Manpower won't materialize before September 2006. All teams will increase except of Lund.

A retreat in September (a week) is needed; participants from both KnowARC ad NDGF (NORDUNet3 umbrella?)

Worldwide projects

M/w: UNICORE, Globus, LCG/gLite, CONDOR

No standards, unclear who and how will influence them. Currently, imlementations define standards. No sense to follow any chosen implementation. On the other hand, there are clear trends and some of the main middleware projects expressed their wish AND started to work on common interfaces. See also the joint white paper by HP, IBM, Intel and Micro$oft.

Interesting Intel's analysis of GT4; reveals e.g. MDS4 problems:

Architecture

Web service interfaces: needed in addition to legacy interfaces.

Coding:

Components:

Functionality of a service must be complete; a service must be a largely standalone component.

Solution should be simple enough to create quickly "private" grids; be started by a single script; have a possibility to run without heavy-weight security procedures (auto-generated certificates, VOs)

May 16

Anders, Aleksandr, Mattias, Balázs, Michael, Oxana

Dependencies:

Should be described in the INSTALL file

System: glibc, bash, perl, libxml2, libltdl, libtool, autoconf, openssl

External dependencies: gSoap, VOMS, GPT, GLOBUS - openldap, certficates

Build: gcc, g++, lbltdl, for gSoap: byacc, bison, flex

Anders will see if gSoap can be linked statically, as the "special" NG version of it is required.

VOMS is patched, masaged and rebuilt, relocatable (def. in /usr/bin, different from gLite, can be a conflict), user "vomses" go to ~/.voms, system - to /etc/voms. Original sructure of having /etc/vomses/ is kept (also exists /etc/vomses *file*). Will be enabled by default for build.

Build README is in works.

FTP server reshuffled (again)

ftp.nordugrid.org
                  /repos
                        /...
                  /software
                        /arc
                            /nightly
                            /releases
                        /globus
                        /...
                        /VOMS

SRM proxy

Twofold task: T0 -> T1 writing; SRM proxy will help, Aleksandr works on it. ATLAS data access: read from 53 SEs at 9 sites, may take large amount of time to query each server, besides, it is unfeasibe to install SRM interfaces (dCache explicitly, it needs a clean system to be installed) at each of those. Aleksandr can think of writing an SRM interface for a GridFTP server, which will be lighter than dCache.

Testing

Log of tests (date, test, result, comment). Oxana to make a page, Anders to make a pre-0.6 tag (0.5.48, announced as a Release Candidate) this week, latest on Friday.