QUADIS Remote Control: Test Plan

---

1. Baseline performance measurement
   1. estimation of RTT, maximum loss (UDP traffic), maximum throughput (TCP traffic), packet loss (packet loss should be null)
   2. estimation of RTT and packet loss as a function of the load on the line
   3. estimation of RTT and packet loss with line satuaration (TCP traffic is used, this measurement is important to estimate the gain introduced by traffic differentiation)

2. Synchronization of end-system clocks (NTP) to verify if one-way delay can be estimated with a reasonable resolution

3. RTT, one-way delay (?) and packet loss measurement with traffic differentiation:
   Notes:
   • traffic is differentiated into 8 classes, traffic to/from the VME controller has higher priority (precedence 7), background traffic is injected to feed classes identified by precedence 0-6
   • background traffic is sourced by:
     - CNAF: workstation sunlab1
     - FNAL: workstation sgi
     - initial tests can be conduced with just one-way traffic for semplicity
   • Functionality CAR is enabled on the input interface of the two routers at CNAF and FNAL
     - Traffic between hosts: 192.168.190.9 and 192.190.218.1 is marked with precedence 7 (verify if we can use DSCPs instead of the precedence, depends on the IOS image) - we call this traffic "priority traffic"
     - background traffic is classified according to the source/ destination address (192.168.190.11 and 192.190.218.2) and to the port number
   1. Evaluation of RTT performance with different queuing algorithms:
      - queueing is enabled on the output (ATM) interface on both the test router at CNAF amd the test router at FNAL
      • Priority queuing: 1 priority queue for priority traffic and 7 WFQ queues for background traffic
      • WFQ: all the queues are served according to the WFQ policy. The service rate of the priority queue is used as test parameter and tuned as a function of the input rate

   2. Evaluation of the application performance with priority traffic multiplexing:
      we need to verify the performance when several CMR controller clients are run in parallel. If multiple clients cannot be run in parallel, additional priority traffic will be injected through traffic generators

4. Evaluation of the suitability of the RTT monitoring facility supported by the applcation

5. Addition of a second class of pririty traffic:
   1. ROOT, for the remote display of analysis results ?
      PHB: Assured Forwarding (see proposal)
   2. video application for additional monitoring of the control room (the video application can be run in parallel, any video source file can be chosen for test purposes)

6. Estimation of application performance in the production environment:
   • RTT estimantion
   • packet loss estimation on production routers (through MIB access)

---

Last modified: Aug 17, 2000