Connection settings

Between (almost) each pair of the DAS cluster interfaces an ATM PVC (Permanent Virtual Connection) is available. This implies that the net performance measurements (throughput and roundtrip, see page  in section 2) has to be executed for the connections between all interface pairs across the ATM route, and when available also across the ``normal'' Internet route, in the sequel shortly denoted as the Internet route.

The available ATM bandwidth can be calculated as follows. The total bandwidth of the connection is 155 Mbit/s. There are also 75 idle cells for each 3 cells. This implies that the bandwidth for the interfaces B_I is

$$B_I = 155 \, \frac{3}{3 + 75} \, \mbox{Mbit/s} = 6.0 \, \mbox{Mbit/s}$$

Because the ATM PVC connections are symmetric, the tests are performed one-directional. This implies that for the interfaces, ordered as $I_1, \ldots, I_4$, the connection I_i => I_j is tested only when i < j. For the Internet route this is in general less true, but the main goal of this package, used in this setup, is to watch the ATM route. Another complication is that not all cluster interfaces can be reached from the Internet, so we ordered the gateways such that as many as possible routes across the Internet could be tested.

The only Internet connection which could not be tested is the connection from the VU cluster to the UVA cluster.

Please note that the RTPL package can be configured to:

1.

Do the measurements across a Quality-of-Service route and across the Internet, or use only one route between a site pair.

2.

Do the measurement between a site pair one-directional or two-directional.

Both items can be combined in all four possibilities.