Bandwidth estimation: metrics, measurement techniques, and tools Presenter: Yuhang Wang

Content Three Bandwidth-related Metrics Different Techniques for accurate bandwidth estimation Taxonomy of Bandwidth Estimation Tools

Bandwidth-related Metric: Capacity Individual link Capacity  Depending on the file size and the capacity of link layer End-to-End Path Capacity  the maximum IP layer rate that the path can transfer from source to sink  The minimum link capacity in the path determines the end-to-end capacity C L2 : capacity of link layer H L2 : Link layer header L L3 : file size of network layer

Bandwidth-related Metric: available bandwidth The available bandwidth of a link relates to the unused, or “spare”, capacity of the link during a certain time period. The average available bandwidth A i of hop i is given by the unutilized fraction of capacity A i =(1-u i )C i.  where, u i is the average of instantaneous utilization over the time interval of interest.

Bandwidth-related Metric: Bulk- Transfer- Capacity (BTC) Bulk-Transfer- Capacity (BTC) is maximum throughput obtainable by a single TCP connection. Affecting factors:  transfer size  type of cross traffic (UDP or TCP)  number of competing TCP connections  TCP socket buffer sizes at both sender and receiver sides  congestion along reverse (ACK) path  size of router buffers and capacity  load of each link

Techniques: Variable Packet Size (VPS) probing Aim: the capacity of each hop along a path Idea: measuring the RTT from the source to each hop of the path as a function of the probing packet size. Methodology: Forcing probing packets to expire at a particular hop by manipulating Time-To- Live (TTL) field in IP header. Corresponding router discards the probing packets, returning ICMP “Time- exceeded” error messages to the source. The source uses the received ICMP packets to measure the RTT to that hop

Techniques: Variable Packet Size (VPS) probing Delay in the way:  serialization delays –transmission delay  propagation delays –time for a packet traverse the link  queueing delays-time waiting in buffers of routers or switches Minimum RTT:  VPS sends multiple probing packets of a given size  VPS assumes that at least one will not encounter any queueing delays  Minimum RTT:

Techniques: Variable Packet Size (VPS) probing

Techniques: Packet Pair/Train Dispersion (PPTD) probing Packet pair probing is used to measure the end-to- end capacity of a path.

Techniques: Packet Pair/Train Dispersion (PPTD) probing

Packet train probing extends packet pair probing by using multiple back-to-back packets Dispersion rate: Path capacity : C=E[D], with a large number of packets in the chain.

Techniques: Self-Loading Periodic Streams (SLoPS)

Techniques: Trains of Packet Pairs (TOPP) Trains of Packet Pairs to estimate the available bandwidth of a network path. Methodology:  sending many packet pairs at gradually increasing rates.  Comparing the link rate at every hop.  Measured rate at receiver: Where R C =C-A, C is the capacity of end-to-end path, A is the available band-width.

Techniques: Other bandwidth estimation methodologies Methodology estimates the utilization of a single bottleneck Methodology estimates available bandwidth using different packet stream patterns from those of SLoPS and TOPP Methodology estimates the capacity of targeted subpaths in a path.

Taxonomy of Bandwidth Estimation Tools

Discussion Do you know some new techniques to measure bandwidth-related metric? Is there any drawbacks in them?

Thank you for your attention!