One-way delay and instantaneous one-way delay variation measurement
for Expedited Forwarding (EF) traffic

---

1. Goal:
   study of time properties of Expedited Forwarding traffic in different test scenarios (LAN, MAN, WAN) with different platforms and different router configurations for the optimum implementation of the Virtual Leased Line service.

2. Test methodology

3. IBM testing (LAN):
   • IBM network layout
   • Results:
     • Average one-way delay
     • Average ipdv
     • One-way delay in time
     • ipdv over time

4. CISCO testing (MAN):
   • CISCO network layout
   • Addressing
     • TEST 1: impact of EF buffer size (queue-limit) on performance
       Parameters:
       queue-limit, packet size, with/without best-effort
       tx-ring-limit is constant, set to its default value
       • Router configuration
       • Results:
         • 1 EF stream:
           • Average one-way delay and average ipdv
           • One-way delay and ipdv over time
         • 1 EF stream + 1 best-effort stream (only for PBS = 1 pack):
           • Average one-way delay and average ipdv
           • One-way delay and ipdv over time
     • TEST 2: Impact of tx ring (tx queue) depth on performance
       Parameters:
       tx-ring-limit = 5 or 40 (default value) (see test 3 for more tx-ring-limit values)
       packet size, with/without best-effort
       • Router configuration
       • Results
     • TEST 3: Impact of tx ring (tx queue) depth on performance (similar to test 2 but with more tx-ring-limit values)
       Parameters:
       tx-ring-limit: [5, 500], constant EF pack size = 1024 by
       queue-limit = 10
       • Results
     • TEST 4: Impact of EF service rate on one-way delay and ipdv with WFQ
       Parameters
       packet size, EF service rate (bandwidth)
       tx-ring-limit = 5, queue-limit = 10
       • Router configuration
       • Results
     • TEST 5: Comparison between WFQ and PQ when applied to an EF queue
       Parameters
       EF packet size, scheduling algorithm (WFQ and PQ)
       tx-ring-limit = 5, queue-limit = 10, WFQ EF service rate = 400 Kbps
       • Router configuration
       • Results
     • TEST 5(b):
       1. Comparison between WFQ and PQ (2 queues) with multiple BE streams (improved version of TEST 5)
          Parameters
          EF packet size, scheduling algorithm (WFQ and PQ)
          tx-ring-limit = 5, queue-limit = 10, WFQ EF service rate = 300 Kbps
          • Results
       2. Comparison between WFQ and PQ (8 queues)
          Parameters
          EF packet size, scheduling algorithm (WFQ and PQ)
          tx-ring-limit = 5, queue-limit = 10, EF load = 100 Kbps, WFQ EF service rate = 300 Kbps
          • Results
     • TEST 6: Comparison between WFQ and PQ with and without BE traffic
       Parameters
       with/without best-effort traffic
       tx-ring-limit = 5, queue-limit = 10, WFQ EF service rate = 400 Kbps
       EF packet size = 1024 by, BE packet size = 1000 by
       • Router configuration
       • Results
     • TEST 7: Priority Queuing performance with variable BE pack size)
       Parameters
       EF packet size, BE packet size
       tx-ring-limit = 5, queue-limit = 10, PQ EF service rate = 300 Kbps
       • Router configuration
       • Results
     • TEST 7(b>: Priority Queuing performance with multiple BE streams and different BE paket distributions (improved version of TEST 7)
       Parameters
       BE packet size distribution
       tx-ring-limit = 5, queue-limit = 10, PQ EF service rate = 300 Kbps
       • Results
     • TEST 8: Test of EF in the wide area with multiple congestion and aggregation points (VLL service)
       Parameters
       with/without best-effort traffic
       with/without EF aggregation
       tx-ring-limit = 5, queue-limit = 10, Priority Queuing
       variable EF frame size
       • Router configuration and results
     • TEST 9: BE in the wide area with congestion (baseline testing)
       Parameters
       BE packet size
       • Results
     • TEST 10: Aggregation and EF traffic load effect on packet loss, one-way delay and ipdv in the wide are, with congestion
       Parameters
       EF traffic load
       EF traffic aggregation degree
       EF test frame size
       tx-ring-limit = 5, queue-limit = 10, Priority Queuing
       • Router configuration and results
     • TEST 11: EF burstiness when departure rate < arrival rate with priority queuing
       • TEST 11 (a): EF burstiness as a function of the EF load
         Parameters
         EF load
         Priority queuing
         • Router configuration and results
       • TEST 11 (b): EF burstiness as a function of the aggregation degree
         Parameters
         number of EF streams (constant load)
         Priority queuing
         • Router configuration and results
       • TEST 11 (c): EF burstiness as a function of the EF packet size
         Parameters
         EF packet size, constant EF load, constant number of streams
         Priority queuing
         • Router configuration and results
     • TEST 12: EF burstiness when departure rate < arrival rate, comparison of priority queuing and WFQ
       Parameters
       Scheduling: PQ and WFQ
       EF packet size
       • Router configuration and results
     • TEST 13: Performance in an uncongested best-effort network
       Parameters
       No QoS features enabled
       Variable BE load
       • Router configuration and results
     • TEST 14: EF demo

---

Last modified: July 12, 2000