Computing Resources

 

The Institute uses the computing resources of the business partner – OJSC «Antel neft».  The special-purpose geophysical computer system is developed by the company on the basis of purchased DELL equipment and is called a special-purpose computer system – 128 (SPCS-128). The basic parameters of SPCS-128 are: efficiency – 12 teraflops, memory – 4 terabytes.

 

The structure of the SPCS is shown in Figure 1.

 

 

Figure 1. The structure of the SPCS-128

 

The SPCS-128 is structured in the following basic functional blocks:

a) the computing field;

b) the controlling server;

c) the network sub-system;

d) the data storage system;

e) the system-wide software;

f) the computer workstations.

 

The structure of the SPCS-128 hardware package is developed in accordance with the industrial standards that makes it possible to reach compatibility with a wide range of controlling and application software.

 

The SPCS-128 consists of 128 computational nodes belonging to the single computing field that is assembled by the high-performance network. Dell PowerEdge M600 blade servers are used as computational nodes.

 

The controlling server is used to control cluster in general, its monitoring and distribution of its tasks. The controlling server relies upon Dell PowerEdge 1950 III.

 

The data storage system is based on the NFS industrial standard and consists of a NetApp disk array. The size of effective disk space is up to 100 terabytes.

 

Connection of a hard drive array of a data communications network is performed by four 10 Gb Ethernet ports.

 

Network Appliance FAS3040 GX is used as the foundation of the data storage system in embodiment of multi-way data storage cluster.

 

Different networking (hardware and software) that together form a network sub-system is used to provide communication between computer modules of the SPCS-128.

 

The network sub-system consists of three physically fixed networks corresponding to the following functionalities:

1) a computer network (or cluster interconnect);

2) a control network;

3) a data communications network.

 

The computer network (or the cluster interconnect) – is a system area network based on the Infiniband standard representing high-speed communication environment that provides performance of a computational efficiency task. The Infiniband standard-based network provides high speed data exchange between cluster nodes during operation of parallel applications.

 

The control network is based on the Gigabit Ethernet standard and allows to control, monitor and distribute tasks between the nodes in the cluster. It is used for files exchange, network booting of the node operating system and OS-level node management.

 

The control network is used to control power, monitor a temperature regime and other parameters of  cluster compute nodes operation.

 

The data communications network is based on the Gigabit Ethernet standard and allows for connection of cluster nodes to the data storage system. Each node of the cluster is connected to the network using the Gigabit Ethernet Adapter.

 

The SPCS-128 system-wide software includes the operating system, batch job management system and the software that provides support for data exchange between the compute nodes during Cluster Interconnect calculation. The Rocks Clusters system – is a software distribution with high level means for installation of software in the cluster and operation of the computing cluster.

 

The SPCS-128 runs on Linux which is a standard for high performance computing clusters.