Trace Server and Compute Cluster
The cluster is designed to store and efficiently process large network traffic traces. Currently around 100 traces comprising 140 gigabytes of high precision passive packet level Internet data are stored, served over gigabit ethernet to dedicated processors with a total of 1.5 GB of high speed memory. The facility supports fundamental research on the statistical nature of Internet traffic and the dynamics of Internet protocols, and the development of tools for the automatic detection of traffic anomalies such as Denial of Service attacks. The current administrator for the trace cluster is Nicolas Hohn.
Hardware
The cluster is composed of 3 Dell workstations called respectively cpu1, cpu2 and nfs1 with the following common characteristics:
| Machine name | cpu1 | cpu2 | nfs1 |
|---|---|---|---|
| CPU | Dual 900MHz Xeon |
Dual 900MHz Xeon | Dual 900MHz Xeon |
| RAM | 1 GB DDR |
512 kB DDR |
256 kB DDR |
| Disk | 10 GB SCSI |
10 GB SCSI | 10 GB SCSI + 150 GB RAID |
| 10/100 Ethernet | yes |
yes |
yes |
| Gigabit Ethernet | yes |
yes |
yes (2 cards) |
| Operating System | RedHat 7.3 |
RedHat 7.3 |
FreeBSD 4.5 |
Software
The programs used to process the trace are based on CoralReef and Matlab.
Data
140 GB of compressed traces are currently stored on nfs1 RAID. They include measurements from the following locations:
- University of Auckland
- Abilene network
- Bell Labs
- University of North Carolina at Chapel Hill
- Small Melbourne ISP