Editing supercomputers/hector

{{sc title|HECToR}}
{{supercomputer
|name=HECToR
|logo=hector phase3.jpg
|sponsor=EPSRC
|sponsor 2=NERC
|sponsor 3=BBSRC
|designer=Cray
|operator=EPCC
|introduction=October 2007
|retired=March 24 2014
|peak dpflops=829 teraFLOPS
|price=$85,000,000
|successor=ARCHER
|successor link=supercomputers/archer
}}
'''HECToR''' ('''High-End Computing Terascale Resource''') was [[terascale]] [[x86]] [[supercomputer]] serving as the UK's primary academic research supercomputer. HECToR has been superseded by {{\\|ARCHER}}.

== Overview ==
HECToR was a UK academic national supercomputer service funded by NERC, EPSRC, and BBSRC for the academic community. HECTor was deployed in a number of phases with each phase upgrading it to over double the performance.

== Phases ==
HECToR underwent 5 major upgrades.

{| class="wikitable"
|-
! Cores || Peak Compute
|-
| 11,328 || 63.44 teraFLOPS
|-
| 12,288 || 113.05 teraFLOPS
|-
| 22,656 || 208.44 teraFLOPS
|-
| 44,544 || 366.74 teraFLOPS
|-
| 90,112 || 829.01 teraFLOPS
|}

=== Phase 1 ===
Phase 1 was the initial system was based on the Cray XT4 (Rainier) system. The system comprised of 1416 compute blades across 60 cabinets. There are four sockets per blade. The processors are dual-core AMD Opteron operating at 2.8 GHz. There is 6 GB of memory per socket for a total system memory of 33.2 TiB. Rainier had a peak compute performance of 63.44 teraFLOPS.

{| class="wikitable"
|-
! Type !! Count
|-
| Nodes || 1,416
|-
| CPUs || 5,664
|-
| Cores || 11,328
|}

In addition to the computational nodes, there were 24 additional service blades with dual-socket dual-core processors that were used for facilitating the supercomputer services such as login, control, and network management.

The system relies on Cray SeaStar2 interconnect. There's one SeaStar2 chip for every two node which comes with 6 network links. The network implements a 3D-torus. The MPI point-to-point bandwidth is was 2.17 GB/s, with a minimum bi-section bandwidth of 4.1 TB/s, and with a node-to-node latency of around 6 μs.

==== Phase 1b ====
The system was upgraded in August 2008 with a small vector system. 28 Cray X2 (Black Widow) vector compute nodes were added. Each node had 4 Cray vector processor. There are 32 GiB of memory for node for a total of 896 GiB of memory.

{| class="wikitable"
|-
! Type !! Count
|-
| Nodes || 28
|-
| CPUs || 112
|}

With each vector processor capable of 25.6 gigaFLOPS, the new addition adds 2.87 teraFLOPS of compute to the existing system. The MPI point-to-point bandwidth is was 16 GB/s, with a minimum bi-section bandwidth of 254 GB/s, and with a node-to-node latency of around 4.6 μs.

=== Phase 2 ===
[[File:hector-2a-xt4.jpg|thumb|right|Phase 2a]][[File:hector-2a-cabinets.jpg|right|thumb|Phase 2a]]
==== Phase 2a ====
HECToR phase two is a Cray XT4 system with a vector X2 system. The system comprises 1416 compute blades across 60 cabinets. Each blade incorporates four compute nodes. Within a compute node are quad-core AMD Opteron CPUs operating at 2.3 GHz along with 8 GB of memory for a total system memory of 44.25 TiB.

{| class="wikitable"
|-
! Type !! Count
|-
| Nodes || 1,416
|-
| CPUs || 5,664
|-
| Cores || 22,656
|}

With a total of 22,656 cores, the system has a peak compute performance of 208.4 teraFLOPS. In addition to the computational nodes, there are 24 additional blades with two dual-core processors per node which is used for various services such as login, control, and network management.

The system relies on Cray SeaStar2 interconnect. There's one SeaStar2 chip for every four nodes which comes with 6 network links. The network implements a 3D-torus. The MPI point-to-point bandwidth is was 2.17 GB/s, with a minimum bi-section bandwidth of 4.1 TB/s, and with a node-to-node latency of around 6 μs.

This system had an Cray X2 vector compute cluster with 28 nodes. There were 4 vector processors per node for a total of 112 processors. With each vector processor capable of 25.6 gigaFLOPS, the new addition adds 2.87 teraFLOPS of compute to the existing system. The MPI point-to-point bandwidth is was 16 GB/s, with a minimum bi-section bandwidth of 254 GB/s, and with a node-to-node latency of around 4.6 μs.

Storage systems

* high-performance RAID: 934 TiB
* NAS (Network Attached Storage): 70 TiB
* MAID (Massive Array of Idle Disks): 112 TiB

==== Phase 2b ====
Phase 2b upgraded the system to a Cray XE6 comprising 464 blades across 20 cabinets. There are four nodes per blade. Each node is a dual-socket 12-core AMD Opteron Magny Cours CPUs operating at 2.1 GHz. 

{| class="wikitable"
|-
! Type !! Count
|-
| Nodes || 464
|-
| CPUs || 3,712
|-
| Cores || 44,544
|}

There are 16 GiB of memory per socket for a total system memory of 58 TiB. With 44,544 cores, the system had a peak compute performance of 374.2 teraFLOPS. In addition to the computational nodes, there are 16 additional blades with two dual-core processors per node which is used for various services such as login, control, and network management.

The system relies on Cray Gemini interconnect. There's one Gemini chip for every two nodes which comes with 10 network links. The network implements a 3D-torus. The MPI point-to-point bandwidth is >= 5 GB/s with a node-to-node latency of around 1-1.5μs.

Storage:

* high-performance RAID disks: 596 TiB
* NAS (Network Attached Storage): 70 TiB
* MAID (Massive Array of Idle Disks): 168 TiB

=== Phase 3 ===
[[File:hector phase3 cabs2.jpg|right|thumb|Phase 3]]
HECToR Phase 3 is a Cray XE6 supercomputer. The system comprises 704 compute blades across 30 cabinets. Each blade incorporates four compute nodes. Within a compute node are two 16-core [[AMD]] {{amd|Opteron}} 2.3GHz {{amd|Interlagos|l=core}}-based processors.

{| class="wikitable"
|-
! Type !! Count
|-
| Nodes || 2,816
|-
| CPUs || 5,632
|-
| Cores || 90,112
|}

There are 16 GiB of memory per socket for a total system memory of 88 TiB. With 90,112 cores, the system had a peak compute performance of 829 teraFLOPS. In addition to the computational nodes, there are 16 additional blades with two dual-core processors per node which is used for various services such as login, control, and network management.

The system relies on Cray Gemini interconnect. There's one Gemini chip for every two node which comes with 10 network links. The network implements a 3D-torus. The MPI point-to-point bandwidth is >= 5 GB/s with a node-to-node latency of around 1-1.5μs.

* high-performance RAID disks: >1 PiB
* Network Attached Storage: 70 TiB
* MAID (Massive Array of Idle Disks): 168 TiB

== TOP500 ==
{| class="wikitable"
|-
! List !! Phase !! Rank !! Rmax !! Rpeak
|-
| 11/2007 || rowspan="4" | Phase 1 || 17 || rowspan="4" | 54.6 || rowspan="4" | 63.4
|-
| 06/2008 || 30
|-
| 11/2008 || 47
|-
| 06/2009 || 68
|-
| 11/2010 || rowspan="2" | Phase 2a || 78 || rowspan="2" | 95.1 || rowspan="2" | 113.0
|-
| 06/2011 || 94
|-
| 11/2009 || rowspan="2" | Phase 2b || 20 || rowspan="2" | 174.1 || rowspan="2" | 208.4
|-
| 06/2010 || 26
|-
| 06/2010 || rowspan="2" | Phase 2b || 16 || rowspan="2" | 274.7 || rowspan="2" | 366.7 
|-
| 11/2010 || 25
|-
| 06/2011 || rowspan="9" | Phase 3 || 24 || rowspan="9" | 660.2 || rowspan="9" | 829.0 
|-
| 11/2011 || 19
|-
| 06/2012 || 32 
|-
| 11/2012 || 35
|-
| 06/2013 || 41
|-
| 11/2013 || 50
|-
| 06/2014 || 59
|-
| 11/2014 || 79
|-
| 06/2015 || 105
|}
@@ Line 55: / Line 55: @@
 The system relies on Cray SeaStar2 interconnect. There's one SeaStar2 chip for every two node which comes with 6 network links. The network implements a 3D-torus. The MPI point-to-point bandwidth is was 2.17 GB/s, with a minimum bi-section bandwidth of 4.1 TB/s, and with a node-to-node latency of around 6 μs.
-The memory subsystem:
-{| class="wikitable"
-|-
-! Type !! Capacity
-|-
-| Direct attached storage || 576 TiB
-|-
-| NAS storage || 40 TiB
-|-
-| MAID storage || 56 TiB
-|-
-| Tape drives || 3 TiB
-|}
 ==== Phase 1b ====
@@ Line 84: / Line 69: @@
 With each vector processor capable of 25.6 gigaFLOPS, the new addition adds 2.87 teraFLOPS of compute to the existing system. The MPI point-to-point bandwidth is was 16 GB/s, with a minimum bi-section bandwidth of 254 GB/s, and with a node-to-node latency of around 4.6 μs.
-The memory has also been upgraded.
-{| class="wikitable"
-|-
-! Type !! Capacity
-|-
-| Direct attached storage || 934 TiB
-|-
-| NAS storage || 70 TiB
-|-
-| MAID storage || 112 TiB
-|-
-| Tape drives || 6 TiB
-|}
 === Phase 2 ===
designer	Cray +
discontinuation date	March 24 2014 +
introductory date	October 2007 +
logo	+
name	HECToR +
operator	EPCC +
peak flops (double-precision)	829,000,000,000,000 FLOPS (829,000,000,000 KFLOPS, 829,000,000 MFLOPS, 829,000 GFLOPS, 829 TFLOPS, 0.829 PFLOPS, 8.29e-4 EFLOPS, 8.29e-7 ZFLOPS) +
release price	$ 85,000,000.00 (€ 76,500,000.00, £ 68,850,000.00, ¥ 8,783,050,000.00) +
sponsor	EPSRC +, NERC + and BBSRC +