SURFnet5 UDPmon Large Results and SNMP

• Description
• SURFnet5 Test Results
• More Result Links:
  • Other UDPmon Large Results
    • More SARA & EVL UDPmon Large Results with SNMP via Netherlight
  • Compact Overview UDPmon Results Diverse Configurations
  • Compact Overview UDPmon Results Back-to-back
  • TCP / UDP Measurements

In this document some UDP experiments are described that were performed with a modified version of the UDPmon package written by R.E. Hughes-Jones. The tool has been modified in the sense that from 5000 packets the sequence number and the arrival time in microsec. are specified. The optional wait time between two successive packets was chosen here to be zero. In the setup described here the tests were executed between the hosts located at EVL, Chicago IL, and those located at SARA, Amsterdam, that are connected with SURFnet5. However, more general UDPmon Large results are available in this document. Other results, also with SNMP, between EVL and SARA via Netherlight are described in this document.

Setup

In this section tests were performed between host gwgsara1, Amsterdam and host prusin, Chicago IL, using SURFnet5. The SNMP counters from the SSR located at SARA, Amsterdam has been read before and after the tests to check for packets lost at that device.

UDPmon Large Results

These tests were executed with packet sizes of 500, 1000, 1200, and 1460 bytes. The time differences with the first packet as function of the packet sequence number are displayed for the direction gwgsara1 => prusin in the . The data for the reverse direction are presented in the . In the bandwidths, calculated with , have been specified for both directions as a function of the packet size. In that table also the # packets lost have been given. Please note that Last_Recv_Seq_Nr in  stands for the sequence number from the last received packet without lost.

(Last_Recv_Seq_Nr * Packet_Size * 8) / Rel_Time

.I.

The relative receiving time as function of the packet sequence number for the UDP stream gwgsara1 => prusin using SURFnet5. The packet size was 500 bytes.

.II.

The relative receiving time as function of the packet sequence number for the UDP stream gwgsara1 => prusin using SURFnet5. The packet size was 1000 bytes.

.III.

The relative receiving time as function of the packet sequence number for the UDP stream gwgsara1 => prusin using SURFnet5. The packet size was 1200 bytes.

.IV.

The relative receiving time as function of the packet sequence number for the UDP stream gwgsara1 => prusin using SURFnet5. The packet size was 1460 bytes.

.I.

The relative receiving time as function of the packet sequence number for the UDP stream prusin => gwgsara1 using SURFnet5. The packet size was 500 bytes.

.II.

The relative receiving time as function of the packet sequence number for the UDP stream prusin => gwgsara1 using SURFnet5. The packet size was 1000 bytes.

.III.

The relative receiving time as function of the packet sequence number for the UDP stream prusin => gwgsara1 using SURFnet5. The packet size was 1200 bytes.

.IV.

The relative receiving time as function of the packet sequence number for the UDP stream prusin => gwgsara1 using SURFnet5. The packet size was 1460 bytes.

SNMP Counters

For these tests the only available counters were the incoming and outgoing octet counters at the SSR from the interfaces connected with the hosts and SURFnet5. In the packages lost have been listed that were found at the SSR for the corresponding path and packet sizes. The table header "SNMP" -> "Packets lost" is the loosely conversion of the SNMP octets lost to SNMP packets lost by division by the packet size. Only non-zero SNMP and / or UDPmon packages lost are specified.

Conclusions

From the UDPmon Large data and the SNMP counters the following conclusions can be drawn:

• Package lost was only found for packet sizes <= 1000 bytes and for the largest part in the direction gwgsara1 => prusin. That is probably due to the less performance of the latter host.
• SNMP packets lost at the SSR were only found in the direction prusin => gwgsara1 for a packet size of 500 bytes. Packets lost for the other packet sizes are probably caused by host effects.
• When the UDPmon data, presented in "More SARA & EVL UDPmon Large Results and SNMP", section "SSR + LSD 6509 Test Results" by the figures 4.I ,..., 4.IV, 6.I ,..., 6.IV, and the tables 3 and 4, that the bandwidths across the Lambda and SURFnet5 are in general comparably host determined. However, the packets lost at the Lambda are considerable larger. A possible explanation might be that the larger amount of router buffer storage in SURFnet5 could limit the losses.