Child pages
  • HPRC Infrastructure History
Skip to end of metadata
Go to start of metadata

Network Interconnect

IDTimelineSwitch(es) InformationSSH (Gb/s)MPI/NFS (Gb/s)
012005 – 2011Ethernet:  Nortel 5510 (x5)11
022011 – 2016

Ethernet:  SMC 8848M (x2)

1


2011 – 2016Infiniband:  Voltaire 4036 (x2)
40
2013 – 2017Ethernet:  Edge-corE 4610-50T (x1)1
2014 – 2017Infiniband:  Mellanox SX6036 (x1)
40/56
2016 – 2017Infiniband:  Mellanox SX6036 (x2)5640/56
032017 – presentEthernet:  Ubiquiti EL48s (x3)1
2017 – 2019Ethernet:  HPE FF5950 (x1)
25/100
2018 – presentEthernet:  Juniper EX4650 (x1)
25/100
2019 – presentEthernet:  Juniper EX4300 (x2)1
2019 – presentEthernet:  Juniper QFX5100 (x1)
10/40

Storage Interconnect

IDTimelineSwitch(es) InformationPort speeds (Gb/s)
112000 – 2007FC (shared)1Gb/s
122005 – 2013Brocade FC (shared)up to 8Gb/s
132013 – presentTwo (2) Brocade 6510 FCup to 16Gb/s

Storage Platforms (dates related to HPC use only)

IDTimelineDescriptionRaw sizeSpindle count and speed(s)Other notes
212000 – 2007SGI TP91002.5TB

54 (10K RPM)


222005 – 2007

SUN FLX380^

6.2TB

28 (10K & 7200 RPM) - shared


232008 – 2011

Compellent*

8.4TB

16 (15K & 7200 RPM)


2009 – 2011

Compellent (expansion)*

32.8TB

48 (15K & 7200 RPM) - shared


242010-2012

IBM XIV Storage Arrays

??

?? (7200 RPM with very large cache) - shared.For research VMs
252011-2016

SGI IS5000 (scratch)

48TB

24 spindles (7200 RPM)

4x8Gb/s FC
2011-2016SGI IS5000+120TB60 spindles (7200 RPM)4x8Gb/s FC
262013 – presentCompellent??

?? (15K & 7200 RPM ) - shared

For research VMs
272013 – 2018SGI IS5500540TB180 spindles (7200 RPM)8x8Gb/s FC
282014 – 2020DDN SFA12K402240TB12 SSDs, 560 spindles (7200 RPM)4x56Gb/s IB
292016 – 2021DELL SC4020 Storage1070TB14 SSDs, 160 spindles (7200 RPM)
302019 – presentDDN SFA7990E768TB12 SSDs, 64 spindles (7200 RPM)

^  Controllers weren't strictly compatible with HPC NAS/HSM server(s) at the time.

*  Controllers weren't strictly compatible with HPC NAS/HSM server(s) at the time.  Long-term performance problems - found to be due to dirty fibres.  Thin provisioning, safety features, and funding issues led to outages.

+ Platform was upgraded by the addition of 20x3TB disks purchased by the VECNet project.

Tape Libraries (near-line storage for HSM)

IDTimelineDescriptionExtra Information
412000 – 2007StorageTek L1804 DLT4 tape drives (until 2005)
4 LTO2 tape drives (2005 – 2007)
422007 – 2011Sun SL5004 LTO4 tape drives, up to189 LTO4 cartridges were installed
432011 – 2016SpectraLogic T9508 LTO5 tape drives, 300 LTO5 cartriges initially, increased to 600 over time

Servers

 Cray (1992 – 1994)

Supercomputing came to JCU with the arrival of a CRAY EL-92 system.
The details of this system (long since decommissioned) aren't known.

 IBM (1994 – 1997)

Handwritten notes appear to indicate that an IBM RS6000 system (grunter) was installed in 1994.
The details of this system (long since decommissioned) aren't known.

 SGI (1997 – 2002)

An SGI PowerChallenge (octopus) system was installed in 1997.  The details of this system (long since decommissioned) aren't fully known. The information below comes from an archive of an old HPC web page. Given the name of this system, it is likely that it was purchased with only 8 CPUs and was later upgraded to 16 CPUs.

Component

Available Details

CPUs

16x(MIPS R10000 195MHz)

ECC memory

4GB

SCSI disks

6x(4.3GB), 10x(9GB), 3x(18GB)

Networking

100Mb/s (FDDI and ethernet)

 SGI (2000 – 2005)

An SGI Origin 3200 (hydra) system was installed in 2000.  An SGI 1200 server was also installed for administration of the Origin 3200 (reused, later, for Altix 350).
The details of the SGI Origin system (long since decommissioned) aren't fully known. The information below comes from archives of old HPC web pages and some from human memory. Several upgrades were made to FC storage capacity and the TP9100 also used for ITR mail and web services storage. Near the end of it's life hydra likely presented out closer to 2TB of FC storage.


Initial

2002 Upgrade

CPUs

16x(MIPS R14000 500MHz)

16x(MIPS R14000 500MHz)

CPUs

4x(MIPS R12000 400MHz)

68x(MIPS R12000 400MHz)

ECC memory

10GB

42GB

SCSI storage

2x(18GB), RAID-1

2x(18GB), RAID-1

FC storage

1.1TB from TP9100

1.1TB from TP9100

Networking

2x(1Gb/s)

2x(1Gb/s)

FC HBAs

2x(1Gb/s)

2x(1Gb/s)

 SGI (2005 – 2008)

An SGI Altix 350 (dmf) was installed in 2005 to serve as a NAS+HSM server for the HPC compute cluster.  The details of this system aren't fully known. The information below comes from packing slips and human memory. I believe this system was upgraded at some stage in its life, but haven't found any documentation associated with an upgrade yet.

CPUs

2x(Itanium2 1.5GHz)

ECC memory

8GB

SCSI storage

2x(146GB 10000 RPM), RAID-1

FC storage

4.5TB from FLX380

Networking

4x(1Gb/s)

FC HBAs

2x(2Gb/s)

 SGI (2005 – 2011)

The first cluster configured for JCU researcher use consisted of 75 SUN V20z servers, 3 SUN V40z servers, and 1 DELL PowerEdge 2850 server.  The latter was used to host APAC National Grid VMs.  A reconfiguration of V20z and V40z servers was performed in 2010 to double the memory per core ratio.

Original Server Builds


V20z(s)V40z(s)DELL 2850
CPUs2x(2.6GHz AMD Opeteron 252)2x(2.4GHz AMD Opteron 852)2x(2.8GHz Intel Xeon)
RAM4GB (400MHz)4GB (333MHz)4GB (400MHz)
Disks2x(73GB, 10K RPM), RAID-12x(73GB, 10K RPM), RAID-12x(146GB, 10K RPM), RAID-1
Networking2x1Gb/s4x1Gb/s4x1Gb/s
FC HBAs
2x1Gb/s
 SGI (2008 -- 2011)

An SGI Altix 450 (dmf1) was installed as a NAS+HSM server for the new HPC cluster, which consisted of 33 SUN X2200M2 servers (C1), 2 SUN X2200M2 servers (C2), and 3 SUN X4600M2 servers.  A second Altix 450 (dmf2) was purchased in 2009 so that a highly available NAS+HSM service could be configured (by SGI professional services staff).  In 2010, the memory in X2200M2 and X4600M2 servers saw a doubling of installed RAM.

Original Server Builds


X2200M2 (C1)X2200M2 (C2)X4600M2Altix 450
CPUs2x(2.3GHz quad-core AMD Opteron 2356)8x(2.8GHz dual-core AMD Opteron 8220)4x(1.6GHz dual-core Itanium 9140M)
RAM8GB (667MHz)32GB (667MHz)32GB (667MHz)
Disks1x(250GB, 7200RPM)2x(146GB, 15K RPM)2x(146GB, 15K RPM)2x(300GB 15000RPM), RAID-1
Networking4x1Gb/s4x1Gb/s10x1Gb/s
FC HBAs


4x4Gb/s
 SGI (2011--2018)

32xC2112-RP4 (CN)

2xC2112 (BM)

4xC2108 (VECNET)

1xC2018 (admin)

2xC3106 (nashsm?)

 SGI (2013 – 2019)


 SGI (2014 – present)


 DELL (2018 – present)


Workstations

In the timeframe 2000-2005, the following workstation equipment was received and used by HPC staff and select JCU researchers.

Qty

Description

Processor(s)

Memory

Storage

Network

6

SGI O2

1x(MIPS)




2

SGI Octane

1x(195MHz MIPS R10000 - 1MB cache)




2

SGI 750

2x(733MHz Intel Itanium - 2MB L2 cache)

1GB

1x(18GB Ultra-3 SCSI)

10Mb/s

1

SUN Netra N1

1x(500MHz UltraSPARC IIe - 512KB L2 cache)

1GB

2x(40GB IDE)

100Mb/s

8

SUN Netra N1

1x(500MHz UltraSPARC IIe - 512KB L2 cache)

1GB

1x(40GB IDE)

100Mb/s

1

SUNblade 2000

2x(900MHz UltraSPARC III - 2MB L2 cache)

4GB



20

SGI 330L

2x(933MHz Intel Pentium III - 256KB L2 cache)


1x(18GB Ultra160 SCSI)


One SUN Netra N1 provided the QPSF website service. Another SUN Netra N1 provided the Tarantella service (X terminal access from a web browser and more).

  • No labels