Transcript presentation source

High Performance Computing
Center at JINR
1997
LCTA “Status Report” Release 1, 1997
Telecommunication systems:
* External communication channels (INTERNET);
* High-speed JINR Backbone;
Systems for powerful computations and mass data processing:
* General High-performance server;
* Clusters of workstations of JINR laboratories and experiments;
* Computing farms (PC-farms);
Data storage system:
* File servers system based on AFS;
* Mass storage system;
* Information servers and database servers;
Software support systems:
* System for application creation and maintenance;
* Visualization systems.
Route Server
Auxiliary
Network Manager
Network Manager
155 Mb/s
ATM Switch
155 Mb/s
10-100 Mb/s
12
10-100 Mb/s
12
12
Statistics for Convex
2381
554
523
1465
2922
1160
38
LCTA
LNP
BLTP
LHE
LPP
FLNR
FLNP
Statistic for VAX Cluster
1761
39
149
284
LCTA
FLNR
555
26
LNP
BLTP
LHE
LPP
Scalar-Vector system
CONVEX C3840
Operating System
ConvexOS 11.5.0
Number of Processors
4
Performance (peak)
200 MFlops/1_CPU
Memory
2 GB
Disks
48 GB
Ethernet, SCSI interface, DA T, 9-trk tape
Scalar system Hewlett-Packard S class SPP2000
Operating System
SPP-UX 10.20
Number of Processors
8
Performance (peak)
800 MFlops/1_CPU
Processor Type
PA8000
Memory
2 GB
Disks
64 GB
ATM, 100BaseT Ethernet
Software:
Batch processing, NQS, C, C++, Parralel Scientific Lib, Fortran,
Message Passing Interface, Parralel Virtual Management,
Parralel Application Development Tools
Data Management System
D-class server + ATL 2640 Automated Tape Library
D-class server
Operating System
Number of Processors
Processor Type
Memory
Disks
HP-UX 10.20
2
PA8000
512 MB
36 GB
ATM, Ethernet, CD, SCSI
Software: HP OpenVeiw OmniBack II, HP OpenVeiw OmniStorage
ATL 2640
Number of cartriges
Capacity of the cartrige
Total capacity
Number of drives
264
20 GB/40 GB
5.28 TB/10.56 TB
3
A Computing Farm to Solve the Problems of Simulation
and Homogeneous Physical Data Mass Processing
The main goal:
to construct at JINR the PC-farm with a new possibilities of
physical experiments simulation, mass data processing and analysis
The main result:
there will be the real possibility of equal in rights participation
of JINR physicist in the data analysis of the large international
experimental programs in high energy physics
The main requests:
CPU Time (on Pentium 200)
>109 s/year
RAM Memory on processor
up to 128 Mb
HDD Memory
200 - 500 Gb
Mass Storage System
4 - 12 DLT
The reason of choice:
A most simple and cheap solution for very special problems
in high energy physics
Customers:
Laboratory of Particle Physics
Laboratory of High Energy Physics
Two stages of the Project realization:
1. PC-farm segment
(2 years)
32 Pentium Pro 200 Processors (4 - for interactive using);
128 Mb per processor; 200 Gb HDD memory (RAID and
SCSI); 4 DLT; 100 Mbit Ethernet Switch with 16 ports
OS "Solaris x86 v.2.6" or Linux
Andrew File System (AFS);
Network Queue System (NQS);
Fortran, C++.
The total cost of this stage is 187,600 USD.
Different financial sources are needed to realize this stage of the project:
the Russia Ministry of Science and Technologies subprogram
"Perspective information technologies" fund, budget of JINR,
special funds of physical collaboration.
2. Full PC-farm:
(3-4 years)
128 Pentium Processors (8 - for interactive using);
128 - 256 Mb per Processor; 500 Gb HDD Mary;
Mass Storage System with 12 DLT;
new Ethernet Technology; new Software Technology.
An estimation total cost of this stage equipment is about 260,000 USD.
The additional financial sources and funds are needed.
SUN/SPARC CLUSTER at JINR as OS SOLARIS
ENVIRONMENT SERVER
Cluster ultra.jinr.dubna.su is working under OS Solaris 2.5.1
Site-licenses for JINR on Fortran F77-4.0 and C++-4.1 provide
all JINR specialists complete conditions in OS Solaris environment
for their work including the usage of current versions of CERNLIB
Site-licenses on F77-4.0 and C++-4.1 are available
for any JINR hosts working under OS Solaris
The latest versions of many FSF/GNU products widely used in JINR
are installed at cluster in a proper way and can be used for
installation at any other JINR hosts working under OS Solaris
CMS CLUSTER at JINR
Hardware:
3 SPARC-stations ( 140 Ultra SPARC station and two
SPARC-stations-20)
Disk Space:
24 GB
Software:
OS Solaris 2.5.1, C-4.0, C++-4.1, F77-4.0 compilers
Number of users: 37
The computational environment is the same
as at the CERN CMS cluster (cms.cern.ch)
CMS cluster at JINR supports both the tasks
of simulation and data processing
MAIN DATA BASES
1. "Physics Information Servers and Data Bases"
Accelerators DB
2. JINR Library Reference DB system
3. DB for IP addresses of JINR
4. Russian WWW-Servers DB
5. Central services DB for administrative economic
activities
Publishing Department DB
STAFF
Topical Plan
Etc.
GENERAL WWW SUPPORT
"Physics Information Servers and Data Bases"
Physics Servers and Services around the World give access to Physics
Encyclopedia, Educational Resources, News in Physics, Data and Tables on
Physical Constants et cetera.
Physics Institutions around the World.
Physics Conferences, Workshops and Summer Schools.
Publishing Offices home pages which give access to book and journal catalogs,
their tables of contents and abstracts of articles, etc
High Energy Physics(HEP) section includes current news, educational materials,
information about HEP centers, programs and service packages on data analysis
and modelling, detector simulation, event generators etc.
Low and Medium Energy Physics
CMS-RDMS Web site
Main JINR Web site
General INFO
Main documents
Journal "JINR Rapid Communications"
Newspaper "DUBNA"
JINR Information Bulletin
JINR Scientific Programme for the Years 1997 - 1999
Etc.
LCTA main Web site
RDMS CMS WWW-SERVER
CMS informational system is heavily based on the world-wide web (WWW)
Web-server http://sunct2.jinr.dubna.su was designed at LCTA and contains
information on RDMS CMS collaboration activities
This web-server was adopted as an official web-server of RDMS CMS
collaboration by the RDMS CMS Collaboration Board in June, 1997
Now there references on RDMS CMS web-server http://sunct2.jinr.dubna.su from
CERN CMS web-servers CMSDOC and CMSINFO
LCTA is responsible for the further development and support of RDMS CMS
web-server
PARTICIPATION IN RUSSIAN INTERDEPARTMENTAL
PROGRAMMES
1."Creation of National Network of Telecommunications for Science and
Higher Schools"
- development of the network RBNet and earth channels (section of
the Programme 3.3.2. jointly with ROSNIIROS)
- development of the network RUHep - "Dubna-Moscow" link
creation (jointly with the Research Institute for Nuclear Physics
MSU)
- head organization within the project BAPHYS -creation of a data
base network for nuclear physics.
2. Working out an interdepartmental programme "Creation of highperformance computer centres in Russia" (funds have been allocated for
organization of HPCC at JINR).
3. A project "A computing farm for solving the problems of modeling and
mass processing of homogeneous physical information" has been worked out
for participation in the programme "Advanced Information Technologies".
PLAN FOR THE YEAR 1998
 To finish the ATM-project on creation of the ATM-based backbones in all JINR
laboratories.
 To provide a reliable operation of the external communication links by means of a
tender selection of the service-provider.
 The most reasonable strategy for development of external computer communications
for JINR, in view of the limited funds, involves aid in solving the question on creation of a
node for a high-speed Europe Backbone (project TEN34) in Moscow, as well as a high-speed
communication link to the USA in frames of projects on the development of the national
network of computer telecommunications for science and higher schools.
 To start up the HPCC at JINR at its first stage -SPP200+DLT robot+mass memory
system.
 To realize a first step in creation of the computing farm.
 To work out a project on organizing an electronic complex for message exchange
within JINR by using a unified data base of all registered JINR users.
 To install an electronic libraries server, based on HPC servers (full-text data bases,
photo-archive data bases, physical data bases etc.)
 To design a centralized Backup system for general JINR servers by using the HPC
facilities.
 To pursue a policy of software standardization and licensing for creation of workplaces
for users of the JINR networking informational-computational infrastructure(NICE95/NT).
 Further development of algorithms and methods for researches under way at JINR in
the field of nuclear physics.
Development Strategy
Link to Europe/USA
from 2 Mbps to 155 Mbps
JINR LAN
Full transfer to ATM
Mass Storage
From Gbytes to Petabytes
Tele-videoconferences
At least on room in each Laboratory
PC-Farms for large experiments
ATLAS, CMS, STAR etc.
High-performance Computing
(nonbudgetary sources)
OO, CASE-tools
Standard-Software, algorithms and methods
LCTA expenses for the development of the JINR informational-computational infrastructure
in 1997 (thous. dollars)
Equipment ordered or purchased services provided.
Price of the ordered
equipment
Expenses
attributed to
LCTA
Modernization (internal and external memory upgrade) of the
database server AlphaServer 2100. Installation of DLT 200 GB
external magnetic tape memory
Purchase of a workstation SUN of UltraSPARC 1 model
Order for telephone-communication gadgets ATM of the
NewBridge company
Purchase of managing program products for ATM-technology
Purchase of workstation Hewlett-Packard Open View for JINR
LAN management
Payments for the network providers' services
Expansion of cluster CONVEX computers. Introduction into
service C3840 machine.
Order of a high-performance system of class S from HewlettPackard
Installation of networking equipment for the 2Mb/sec computer
communication channel
22,86
22,86
6,08
116,4
6,08
30,0
Fee of Germany
Fee of Germany
56,4
27,2
5,64
27,2
Fee of Germany
Fee of Germany
73,5
200,0
8,8
0
250,0
0
50,0
0
0
Construction on a 2MB/s fibre optic channel JINR – University –
SCC2
23,0
0
1
Rent fee for the 2 Mb/sec computer channel
4,9 per month
0
830,34
100,58
Total
Means of financing at LCTA
Fee from Germany
Topic –1019
Transferred for the free use according to a
protocol with Rossendorf
Grant of Ministry of Science
Transferred free of charge according to a
protocol with
RosNIIROS.
At the expense of fund coming from
Ministry of Science within the
interdepartmental programme
Funding is provided at the expense of
Ministry of Science in frames of protocol
JINR – RosNIIROS.
In according to Memorandum of Understanding of collaboration between the
Laboratory of Computing Techniques and Automation of the Joint Institute for
Nuclear Research, Dubna, Russia and the SPS/LEP Division of the European
Organization for Nuclear Research, Geneva, Switzerland during 1997 year has
been done:
1. TDM Monitoring System
2. MOPOS timing diagnostics system
3. LabVIEW on the WEB
4. TCP/IP based Message Handler (LV\_mhm)
5. Synchronisation of LabView application
with an accelerator's cycle
6. An Object-Oriented technology in
LabVIEW programming
http://hpslz24.cern.ch:8080/cgi-bin/tdm\_protect.vi
http://hpslz24.cern.ch:8080/examples/ex\_mopos.html
http://hpslz24.cern.ch:8080/
http://hpslz24.cern.ch:8080/others/rem\_exec.html
http://hpslz24.cern.ch:8080/sync/LV\_sync.html
http://hpslz24.cern.ch:8080/others/LV\_oop.html
(report ICALEPS'97 Beijing 1-7.11.1997)
Future developments and projects on 1998 year (a few points has been
done now)
1. BEAM LOSS MONITORING System http://hpslz24.cern.ch:8080/examples/BML.html
2. COMMON LABVIEW ENVIRONMENT
http://hpslz24.cern.ch:8080/others/LV\_slops.html
for the operational machines
3. Evaluation of ACE
http://hpslz24.cern.ch:8080/others/LV\_ace.html