Other Research Interests

Page Sections

Applied Grid Computing

Being a CFD researcher I am interested in High Performance Computing (HPC) particularly low cost grid and distributed computing. The largest grid computer that I have created was based on the computers used in several of the Faculty's undergraduated student computing laboratories - I borrow the computers when the labs are closed for maintanence over break periods. The specifications of the compute nodes are listed below in Table 1.

Table 1 - Compute Node Specifications
Paramter Value

Processor Manufacturer

Intel

Processor Speed

2.8GHz

Level 1 Cache

128k bytes

Level 2 Cache

1024k bytes

Network Interconnect

Copper Gigabit

Hard Disk Capacity

80G byte ATA

Operating System

RedHat Enterprise Linux Workstation Version 3 update 5

The grid computer was split across three faculty labs and used the Faculty network for laboratory interconnect. The individual laboratory switches used were Cisco Catalyst 3000 series gigabit switches with multimode fiber channel connections to the core routers. The core routers were Cisco boxes. Two of the laboratories were connected to the same switch which meant that internal traffic between two thirds of the compute nodes never left the switch back plane. Table 2 shows the overall grid statistics. The maximum acheived processing power was based on the HPL Benchmark which is the benchmark used by the Top500 supercomputer list.

Table 2 - Grid Computer Statistics

Parameter

Value

Compute nodes

74

Theoretical Processing Power

402Gflops

Maximum Acchived Processing Power

 

Total Scratch Space for jobs

2.2TB

Total Storage Space

1.5TB

 

To provide central servicing and management of the grid I was loaned a Dell PowerEdge 1650 server from the Faculty's Computer Support Program - this was a server that had been purchased but not yet introduced into production service. The server provided the primary Network File System (NFS) mount point, collation and distribution of monitoring and remote secure shell access. A diagram of the network layout is shown below in Figure 1.

Network Diagram

Central monitoring of the cluster was undertaken using the Ganglia Toolkit. This is an open source toolkit designed to monitor very large clusters. Particularly it can scale up to clusters of thousands of computers. A sample image of the monitoring provided by Ganglia is shown below in Figure 2. If you are within the Faculty of Engineering fire wall while the project is running live server statistics are available.

Potential Capstone Topics

As I am a current PhD student and not a full-time member of the academic staff I can not directly supervise capstone projects. However in conjunction with my supervisors I "oversee" several capstone projects related to most aspects of CFD modelling and theory. If you are interested in a capstone project please contact me or Dr Matthew Gaston (my supervisor). Alternatively some of the topics that I am currently interested in or researching are listed below. Also check out my research page to view my personal research.