Hardware
More than 3,500 CPU cores and 3.5 PB of storage space will be installed till 2013. Information on purchased hardware will be added to this page. The centre hasn't started from scratch, though, SCB operated over 550 CPU cores and over 150 TB of disk space.
Hardware Resources Financed by the Project
Hardware resources are obtained according to the following time schedule:
Computation Resources
The resources are operated by the project activity
Purchase and operation of the computation cloud.
Currently, a SMP cluster Zewura is available to the users.
The Zewura Cluster
Nodes zewura1.cerit-sc.cz -- zewura8.cerit-sc.cz, 640 CPU nodes in total.
The cluster has been purchased and put into operation in December 2011. The cluster contains 8 servers HP ProLiant DL980 G7, each of them in configuration
- 8 Intel Xeon E7-2860 processors (10 CPU cores, 2.26 GHz),
- 256 GB RAM (will be upgraded to 512 GB shortly),
- 20x 900GB hard drives to store temporary data (
/scratch), configured in RAID-10, thus having 8 TB capacity.
The performance of each node (as measured during acceptation tests of the cluster) reaches 1250 points in the SPECfp2006 base rate benchmark.
The nodes are interconnected with:
- InfiniBand: two 4xQDR interfaces per node,
- 10 Gbit/s Ethernet: primary access network,
- 1 Gbit/s Ethernet: four interfaces per node, used mainly for experimental configuration of virtual networks.
The space for permanent user data storage (/home) should have been provided by a disk storage system (obtained concurrently with the cluster). The supply didn't satisfy specified requirements both in terms of performance and functionality, it was therefore returned to the supplier in December 2011. Currently, a new tender is being worked on. The disk storage should be available in the second quarter of 2012.
The /home space is therefore taken from a part of the /scratch area of the zewura8 node. Being formatted in RAID-50, it provides approx. 7.9 TB of capacity (traded for a small perfomance loss owing to the configuration).
Data Storage
See the Building and Operation of Data Storage activity.
Legacy Resources Operated by SCB
In the beginning of 2011, SCB operated over 550 CPU cores owned by the Masaryk University. They were mainly the quark cluster (Faculty of Informatics, 62 CPU cores), orca and perian clusters (National Centre for Biomolecular Research, approx. 500 CPU cores), cluster loslab (Loschmidt Laboratories, 24 cores), and cluster manwe (SCB, 48 cores intergrated in the national grid). According to up-to-date classification, all of them are considered HD clusters. The computers are located in ICS computer rooms and in the University Campus in Bohunice. We also host 112 more CPU cores of cluster manwe and 212 cores of cluster skirit, both owned by the CESNET association.
In the data storage area, we operate three storage servers distributed in three university buildings, having 144 TB of capacity in total, and a set of older disk arrays of 34 TB. These resources serve mainly for scientific data of university employees. In our computers rooms, disk arrays serving as home directories of MetaCentrum are also hosted.
On Cluster Names
The Hitch Hiker's Guide to the Galaxy says that zewura is a primitive organism of the planet Cefeus Theta Three. They loosely resemble Polychaeta on Earth. In order to sustain the climatic conditions of the most recent ice age on the planet, their body mass grew up to several stones, and the performance of their metabolism evolved to monstrous scale. Consequently, in the following warmer climatic period they bred excessively, consuming virtually all the biomass of the planet shortly, and causing a total collapse of the whole ecosystem. Therefore they can serve as a textbook example of the evolution carrying too far sometimes.
