Hierarchical Rate Limits Overview - Juniper JUNOSE SOFTWARE FOR E SERIES 11.3.X - POLICY MANAGEMENT CONFIGURATION GUIDE 2010-10-04 Configuration Manual

Hierarchical Rate Limits Overview

Copyright © 2010, Juniper Networks, Inc.
Traffic passes through the rate limiter causing a draining of tokens. The drain rate is
dependent on how large the packets are and how much time elapses between packets.
At any given instant the level of tokens in each bucket is a function of the fill rate, size of
packets, and elapsed time between packets.
When packets are received on an interface with a rate limiter applied, the level of tokens
in each bucket dynamically changes in both of the following ways:
Tokens are added every 100-ms sample period
Tokens are removed based on the size and rate of incoming packets
In another type of rate limiting, rate-limit hierarchies enable lower priority traffic to access
unused bandwidth allocated for real-time traffic, such as voice or video, during times
when no real-time traffic is flowing. IP subscribers receive multiple services, such as Web,
video, and file transfer, that have a maximum bandwidth. A rate-limit hierarchy can apply
a common rate limit to several classified flows, enabling them to share bandwidth
according to the preferences set in the hierarchical rate limits.
You can also use rate-limit hierarchies in a layer 2 (ATM) access network for DSL where
many routing gateways lead into one Broadband Access Server. The Broadband Access
Server uses rate-limit hierarchies to allocate shareable bandwidth to each routing
gateway, which enables unused bandwidth from one routing gateway to be used by
others. The hierarchy in the rate limit represents the hierarchy in the access network.
Rate-limit hierarchies enable you to share unused bandwidth dynamically, taking unused
preferred bandwidth. They also enable real-time traffic to use all guaranteed bandwidth
at any time without violating the configured limit on the total interface bandwidth. While
preferred traffic fluctuates, the interface rate limit adjusts, dropping non-preferred packets
to keep the total flow through the interface under a configured maximum rate, because
preferred packets cannot be dropped by the shared rate limits, only by their individual
rate limits.
Shared rate limits in the hierarchy keep the combined traffic below a configured maximum
without dropping preferred packets. Preferred packets always reduce tokens on these
rate limits, making their token counts negative, if necessary. Later non-preferred packets
are then dropped in greater volume, bringing the total traffic through the shared rate limit
below its configured maximum.
Every packet passing through a rate limit hierarchy has an owner, which is the last rate
limit that can modify the packet; for example, by changing its color or dropping it. Preferred
packets are owned by their individual preferred rate limits, which do not transfer ownership
of the packet while the packet traverses the hierarchy. Ownership of non-preferred
packets is transferred while they move from one rate-limit to the next in the hierarchy,
so shared rate limits can change the packet color or drop them.
Chapter 5: Creating Rate-Limit Profiles
59