Traffic Shaping, Traffic Policing Peter Puschner, Institut fr - - PowerPoint PPT Presentation

Traffic Shaping, Traffic Policing

Peter Puschner, Institut für Technische Informatik

Peter Puschner, TU Wien

Traffic Shaping, Traffic Policing

• Enforce compliance of traffic to a given traffic

profile (e.g., rate limiting)

• By delaying or dropping certain packets, one can

(i) optimize or guarantee performance, (ii) improve latency, and/or (iii) increase or guarantee bandwidth for other packets

• Traffic shaping: delays non-conforming traffic
• Traffic policing: drops or marks non-conforming

traffic

2

Peter Puschner, TU Wien

Traffic Shaping

• Traffic metering to check compliance of packets

with traffic contract e.g., leaky bucket / token bucket algorithm

• Imposes limits on bandwidth and burstiness
• Buffering of packets that arrive early

– Buffer dimensioning (?)

• Strategy to deal with full buffer

– Tail drop (à policing) – Random Early Discard – Unshaped forwarding of overflow traffic

3

Peter Puschner, TU Wien

Traffic Shaping

• Self limiting sources
• Shaping by network switches
• Shaping traffic uniformly by rate
• More sophisticated characteristics (allow for

defined variability in traffic)

4

Peter Puschner, TU Wien

• Bucket capacity: C [tokens]
• Token arrival rate: r [tokens per second]
• When a packet of n bytes arrives, n tokens are removed

from the bucket and the packet is sent

• If fewer than n tokens available, no token is removed

and the packet is considered to be non-conformant

5

Token Bucket Algorithm

Peter Puschner, TU Wien 6

Leaky Bucket Algorithm

C n n C n r … leak rate

Peter Puschner, TU Wien

r … leak rate

7

Leaky Bucket Algorithm

C n n C n

• Bucket with capacity C leaks at fixed rate r
• The bucket must never overflow
• If the bucket is empty it stops leaking
• A packet is conformant, if the amount of water, n, can be

added to the bucket without causing an overflow; n is either constant or proportional to packet size

• For non-conformant packets, no water is added to the

bucket

Peter Puschner, TU Wien

Leaky Bucket Properties

Best average rate (over infinite time) r [bytes/s] r /n [messages/s], with message size n bytes Maximum burst size assume max. transmission rate M [bytes/s]

8

Tmax = C M – r if r < M ∞ otherwise Lmax = Tmax × M [bytes] Lmax = ( Tmax × M ) / n [messages]

msg

Peter Puschner, TU Wien

Leaky Bucket - Notes

C – n … characterizes max. jitter tolerance “how early can a packet be sent?” C = n … minimum message-send interval: n/r r / n … maximum rate at which packets can conform once the bucket is full The bucket stops leaking when it is empty à this limits the tolerance accrual in times when the incoming traffic rate is low C, r, n, queue size determine outgoing traffic

9

Peter Puschner, TU Wien

Leaky Bucket – Remark

Strategy can be used to measure rate of any stochastic process e.g., detect when an event rate increases above some acceptable threshold (bucket overflow) Different use of leaky-bucket counter on overflow condition

• Unbounded increment to penalize overload cond.
• No increment when bucket is full à allows to

detect when situation is back to normal

10

Peter Puschner, TU Wien 11

Summary

• Strategies to enforce compliance of traffic to a

given profile

– Traffic Shaping: delays non-conforming traffic – Traffic Policing: drops/marks non-conforming traffic

• Who does the shaping?

– Sender: to guarantee that network accepts traffic – Network switches: to enforce contract compliance

• What is controlled?

– Rate, jitter, burstiness