From Molecular Modeling Wiki

(Difference between revisions)

Revision as of 11:51, 17 March 2009

Overview

Description

Depending on the hardware equipment and the age of a cluster there is always a set of typical applications that fit the cluster capabilities. The idea is to run any calculation on a cluster that just fits the job requirements; running it on less capable cluster would slow down the calculation or prevent running it at all, starting a low-demanding job on highly equippped cluster would waste the resources and prevent running jobs that require such equipment.

Cluster survey

Name	Year of construction	Memory size [GB] / core	Disk (scratch) size [GB]	Usage	Access
cobalt	(?) 2003	1	120	simple calculations, student tests, teaching	public
argon	2005	2	160	single-processor calculations with lower requirements	public
krypton	2005	1 - 2	160	single-processor calculations with lower requirements	public
radon	2005	2 - 4	320	single-processor calculations with average requirements	public
palladium	2004 - 2005	3 - 6	240 - 600	older cluster for memory and/or disk space demanding calculations	public
iridium	2006 - 2008	2	320 - 1000	general use	restricted
helium	2006 - 2008	2	250 - 320	parallel molecular dynamics calculations	restricted
francium	2007	4	640	parallel molecular dynamics calculations with infiniband	restricted
lithium	2006 - 2008	4 - 16	1600 - 6000	highly demanding parallel and/or single-processor calculations with large memory and disk space requirements	restricted
uranium	2008	2	1500	general use	public
barium	2007 - 2008	1 - 2	320 - 1500	mostly parallel molecular dynamics calculations	restricted
thallium	2007 - 2008	1 - 2	1500	general use	restricted

Clusters

The following text provides detailed information about clusters and nodes. Tables show resources available for calculations, which are usually smaller than total resources mentioned in the general overview above; some of the total node memory (RAM) must always remain reserved for operating system tasks.

cobalt

Total nodes: 40
Total cores: 40
Parallel calculations: no
Queuing system: DQS

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
c01 - c40	40	c01_a - c40_a	Intel Pentium 4 2.80GHz	1	800 MB	120 GB

argon

Total nodes: 44
Total cores: 44
Parallel calculations: no
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
a01 - a44	44	aq	Intel Pentium 4 3.40 GHz	1	1.8 GB	160 GB

krypton

Total nodes: 44
Total cores: 44
Parallel calculations: no
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
k01 - k16	16	kq2	Intel Pentium 4 3.40 GHz	1	1.8 GB	160 GB
k17 - k44	28	kq1	Intel Pentium 4 3.40 GHz	1	0.8 GB	160 GB

radon

Total nodes: 46
Total cores: 46
Parallel calculations: no
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
r01 - r24	24	rq4	AMD Athlon 64 3500+	1	3.6 GB	320 GB
r25 - r46	22	rq2	AMD Athlon 64 3500+	1	1.6 GB	320 GB

palladium

Total nodes: 14
Total cores: 10
Parallel calculations: yes, within a single node (in pqXp queues, max 2 processor per job)
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
p01 - p04	4	pqos	AMD Opteron 244	2	2.6 GB	240 GB
p05 - p14	10	pqop	AMD Opteron 244	2	2.6 GB	240 GB
p15 - p18	4	pqns	AMD Opteron 250	2	5.6 GB	600 GB
p19 - p24	6	pqnp	AMD Opteron 250	2	5.6 GB	600 GB

iridium

Total nodes: 50
Total cores: 120
Parallel calculations: yes, special setup
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
i01 - i34	34	<unknown>	Intel Core2 6600 2.40GHz	2	1.6 GB	320 GB
i35 - i40	6	<unknown>	Intel Core2 Duo E6750 2.66GHz	2	1.6 GB	320 GB
i41 - i42	2	<unknown>	Intel Core2 Quad Q6600 2.40GHz	4	1.8 GB	1000 GB
i43 - i50	8	<unknown>	Intel Core2 Quad Q9550 2.83GHz	4	1.8 GB	1000 GB

helium

Total nodes: 16
Total cores: 64
Parallel calculations: yes, accross nodes, MPI
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
h01 - h06	6	hq	AMD Opteron Dual Core 275	4	1.8 GB	320 GB
h07 - h16	10	hq	AMD Opteron Dual Core 2216	4	1.8 GB	250 GB

francium

Total nodes: 24
Total cores: 96
Parallel calculations: yes, accross nodes, MPI, infiniband
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
f01 - f24	24	fq	AMD Opteron Dual Core 2214	4	3.8 GB	640 GB

lithium

Total nodes: 16
Total cores: 82
Parallel calculations: yes, within a single node, MPI
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
l01 - l04	4	lq-4-8	AMD Opteron Dual Core 285	4	7.8 GB	3000 GB
l05 - l08	4	lq-2-16	AMD Opteron 256	2	15.4 GB	3000 GB
l09	1	lq-2-16	AMD Opteron 854	2	15.4 GB	1600 GB
l10 - l16	7	lq-8-4	Intel Xeon E5430 2.66GHz	8	3.8 GB	6000 GB

uranium

Total nodes: 24
Total cores: 192
Parallel calculations: yes, within a single node, MPI
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
u01 - u14	14	columbusq	Intel Xeon E5430 2.66GHz	8	1.8 GB	1500 GB
u15 - u24	10	uq	Intel Xeon E5430 2.66GHz	8	1.8 GB	1500 GB

barium

Total nodes: 19
Total cores: 132
Parallel calculations: yes, within a single node, MPI
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
b01 - b02	2	bq-4-1	Intel Core2 Quad 2.40GHz	4	0.8 GB	320 GB
b03 - b05	3	bq-4-2	Intel Core2 Quad 2.40GHz	4	1.8 GB	320 GB
b06 - b14	9	bq-8-1	Intel Xeon E5345 2.33GHz	8	0.8 GB	1320 GB
b15 - b19	5	bq-8-1	Intel Xeon E5430 2.66GHz	8	0.8 GB	1500 GB

thallium

Total nodes: 19
Total cores: 124
Parallel calculations: yes, accross multiple nodes, MPI
Queuing system: SGE

Nodes	# of nodes	Queue	CPU	Cores/node	Available RAM/core	Scratch/node
t01 - t07	7	tq-4-2	Intel Core2 Quad 2.40GHz	4	1.8 GB	1500 GB
t08 - t14	7	tq-8-1	Intel Xeon E5345 2.33GHz	8	0.8 GB	1500 GB
t15 - t19	5	tq-4-2	Intel Xeon E5430 2.66GHz	8	0.8 GB	1500 GB

Totals

Total # of nodes	Total # of cores	Total RAM	Total disk space
366	1032	2.5 TB	240 TB

@@ Line 3: / Line 3: @@
 == Description ==
-Depending on the hardware equipment and the age of a cluster there is always a set of typical applications that fit the cluster capabilities. The idea is to run any calculation on a cluster that just fits the job requirements; running it on less capable cluster would slow down the calculation or prevent running it at all, starting a low-demanding job on highly equippped cluster would waste the resources and prevent running jobs that require such equipment. [[what is this]]
+Depending on the hardware equipment and the age of a cluster there is always a set of typical applications that fit the cluster capabilities. The idea is to run any calculation on a cluster that just fits the job requirements; running it on less capable cluster would slow down the calculation or prevent running it at all, starting a low-demanding job on highly equippped cluster would waste the resources and prevent running jobs that require such equipment.
 == Cluster survey ==

Cluster list