Project

Back to overview

Renewal of Central Computational Equipment

English title Renewal of compute cluster
Applicant Röthlisberger Ursula
Number 128752
Funding scheme R'EQUIP
Research institution Laboratoire de chimie et biochimie computationnelles EPFL - SB - ISIC - LCBC
Institution of higher education EPF Lausanne - EPFL
Main discipline Physical Chemistry
Start/End 01.12.2009 - 30.11.2010
Approved amount 80'000.00
Show all

Keywords (6)

hardware; PC cluster; hardware upgrade; computer cluster; parallel calculations; computational chemistry

Lay Summary (English)

Lead
Lay summary
The research activity of the Laboratory of Computational Chemistry and Biochemistry (LCBC) at the EPFL focuses on the development and the application of first-principles and mixed quantum mechanical/molecular mechanical (QM/MM) molecular dynamics (MD) methods based on density functional and time-dependent density functional theory. The group's central codes are CPU, memory and communication intensive. LCBC was formed in 2002 and typically consist of 15-20 members. As the group is involved in purely theoretical and computational research, sufficient local computational resources are absolutely essential for our work. The group's local computer clusters are now 3-4 years old and have to be renewed.Here, we propose to replace the old 4-node high memory machines with six 2x quad-core AMD Opteron 2384 with 32 GB memory and 2.7 GHz processors, and the 64-CPU-medium memory-Myrinet cluster with 21 twin nodes 4 x quad-core AMD Opteron 2384 with 16 GB memory, 2.7 GHz processors and Infiniband interconnect. The total cost for the renewal of the computational equipment amounts to 237k SFr. We request 50% (118.5k SFr) from the R'equip programme of the SNF and the other 50% from EPFL as a matching fund.Since LCBC is involved in many collaborations with experimental groups, e.g. within EPFL we are performing calculations for projects with the groups of Majed Chergui, Paul Dyson, Michael Graetzel, Vassily Hatzimanikatis, Hilal Lashuel, Kay Severin, Thomas Rizzo, and Horst Vogel, several groups at the EPFL will indirectly profit from this acquisition too.
Direct link to Lay Summary Last update: 21.02.2013

Responsible applicant and co-applicants

Associated projects

Number Title Start Funding scheme
125173 Rational design of organo-ruthenium anticancer compounds with novel modes of action 01.04.2009 Project funding (Div. I-III)
116294 Multiscale Modelling of Electronically Triggered Biological Signalling 01.05.2007 Project funding (Div. I-III)

Abstract

The research activity of the Laboratory of Computational Chemistry and Biochemistry (LCBC) focuses on the development and the application of first-principles and mixed quantum mechanical/molecular mechanical (QM/MM) molecular dynamics (MD) methods based on density functional and time-dependent density functional theory. The group's central codes are CPU, memory and communication intensive. LCBC was formed in 2002 and typically consist of 15-20 members. As the group is involved in purely theoretical and computational research, sufficient local computational resources are absolutely essential for our work. The basic computational equipment needed for our work are (1) personalized workstations with high floating point performance that serve as working place for each group member for code development, test calculations, data processing and as terminals to remote machines, and (2) purpose-specific small to medium size high performance compute clusters as shared resources of the group for dedicated number crunching, i.e. productions runs for medium sized projects. Currently, the group has three different platforms of this type: a 4 years-old 4-nodes dual Intel Xeon 5140 cluster with 32 GB memory per node and high performance Myrinet switch that is dedicated to quantum chemical (post Hartree Fock) memory intensive applications; a loosely coupled (Gigabit Ethernet switch) 3-years old 64 processors HP-Opteron-246 cluster with 2 GB memory per node specifically designed to perform classical MD simulations in combination with enhanced sampling techniques, in particular parallel tempering, and a 4-years old 64 processors dual AMD Opteron-250 cluster with Myrinet switch used for highly parallelized applications (mainly ab initio Car-Parrinello and QM/MM MD) with medium memory requirements (4 GB memory per node). The latter was bought immediately when the group was established in 2002 and renewed 4 years ago. The machine was covered by a 3-years warranty that is now over. All our computational resources are running at full load. As can be expected from intensely used 4-years old computers, hardware failures (defect ventilators, hard disks etc.) are becoming more and more frequent. Furthermore, the provided computational power is no longer competitive with today's standards. Due to the ever increasing maintenance costs and the loss of computational power, it has become necessary to renew all hardware older than 3 years, namely the high-memory-4-nodes quantum cluster and the medium-memory 64 CPU PC cluster. The original acquisition price of these two platforms was ca. 230k SFr. Here, we propose to replace the 4-node high memory machines with six 2x quad-core AMD Opteron 2384 with 32 GB memory and 2.7 GHz processors, and the 64-CPU-medium memory-Myrinet cluster with 21 twin nodes 4 x quad-core AMD Opteron 2384 with 16 GB memory, 2.7 GHz processors and Infiniband interconnect. The total cost for the renewal of the computational equipment amounts to 237k SFr. We request 50% (118.5k SFr) from the R'equip programme of the SNF and the other 50% from EPFL as a matching fund.Since LCBC is involved in many collaborations with experimental groups, e.g. within EPFL we are performing calculations for projects with the groups of Majed Chergui, Paul Dyson, Michael Graetzel, Vassily Hatzimanikatis, Hilal Lashuel, Kay Severin, Thomas Rizzo, and Horst Vogel, several groups at the EPFL will indirectly profit from this acquisition too.
-