Specifications

Chapter 6

Measurements and Results

The tester was put into operation in the CERN network and some measurements were

performed at the IP level ([7]). The system was composed of 2 PCs located in diﬀerent

buildings inside CERN. We measured the average latency at diﬀerent loads. For the clock

synchronization we used the GPS global clock system. The network topology between the

two ends is shown in Figure 6.1.

Primergy

SWITCH

GPS

PCATB56

PCATBGPS01

100Mbps

SWITCH

CERN

1Gbps

100Mbps

R = Router

r513−an6

b513−gb8

r40−gb35

b40−bb10

b513−bb1

Network

r513−gb8

b513−bb1

b513−bb10

b40−gb35

r40−an4

Figure 6.1: The CERN network between the two buildings

The packets pass through 5 routers and take a diﬀerent route on each direction. The

packet size was 1518 bytes and the load was set to 20%. The Type of Service ﬁeld in the

IP packets was also changed but no signiﬁcant variations were observed. In Figure 6.2 we

show a latency histogram for the traﬃc between the two buildings. The sharp peak on the

left side in the distribution indicates that the load between the two building was rather

low, and packets were traversing the route without waiting in the queues in various routers

and switches. Knowing the exact conﬁguration of the route we were able to calculate, that

the packets were spending 500 µs on wires and the remaining 540 µs inside routers and

switches.

Another set of measurements was performed between CERN and Cracow ([8]). These

measurements are part of a feasibility study for moving part of the Atlas event processing

machines to oﬀ-site institutes. The ﬁrst tests are using the existing network infra-structure

— the traﬃc passes through the CERN local network to Cracow via the GEANT backbone