rfc1254.txt

中、英文RFC文档大全打包下载完全版 .

   little traffic.  The selection of packets to be dropped is completely
   uniform.  Therefore, a user who generates traffic of an amount below
   the "fair share" (as defined in Section 2.3) may also experience a
   small amount of packet loss at a congested gateway. This character of
   uniformity is in fact a primary goal that Random Drop attempts to
   achieve.

   The other primary goal that Random Drop attempts to achieve is a
   theoretical overhead which is scaled to the number of shared
   resources in the gateway rather than the number of its users.  If a
   gateway congestion algorithm has more computation the more users
   there are, this can lead to processing demands that are higher as
   congestion increases.  Also the low-overhead goal of Random Drop
   addresses concerns about the scale of gateway processing that will be
   required in the mid-term Internet as gateways with fast processors
   and links are shared by very large active sets of users.

3.2.1  For Congestion Recovery

   Random Drop has been proposed as an improvement to packet dropping at
   the operating point where the gateway's packet buffers overflow.
   This is using Random Drop strictly as a congestion recovery
   mechanism.

   In Random Drop congestion recovery, instead of dropping the last
   packet to arrive at the queue, a packet is selected randomly from the
   queue.  Measurements of an implementation of Random Drop Congestion
   Recovery [Man90] showed that a user with a low demand, due to a
   longer round-trip time path than other users of the gateway, had a
   higher drop rate with RDCR than without.  The throughput accorded to
   users with the same round-trip time paths was nearly equal with RDCR
   as compared to without it.  These results suggest that RDCR should be
   avoided unless it is used within a scheme that groups traffic more or
   less by round-trip time.



Performance and Congestion Control Working Group               [Page 10]

RFC 1254           Gateway Congestion Control Survey         August 1991


3.2.2  For Congestion Avoidance

   Random Drop is also proposed as a congestion avoidance policy
   [Jac89].  The intent is to initiate dropping packets when the gateway
   is anticipated to become congested and remain so unless there is some
   control exercised.  This  implies  selection  of  incoming packets to
   be randomly dropped at a rate derived from identifying the level of
   congestion at the gateway.  The rate is the number of arrivals
   allowed between drops. It depends on the current operating point and
   the prediction of congestion.

   A part of the policy is to determine that congestion will soon occur
   and that the gateway is beginning to operate beyond the knee of the
   power curve.  With a suitably chosen interval (Section 2.1), the
   number of packets from each individual user in a sample over that
   interval is proportional to each user's demand on the gateway.  Then,
   dropping one or more random packets indicates to some user(s) the
   need to reduce the level of demand that is driving the gateway beyond
   the desired operating point.  This is the goal that a policy of
   Random Drop for congestion avoidance attempts to achieve.

   There are several parameters to be determined for a Random Drop
   congestion avoidance policy. The first is an interval, in terms of
   number of packet arrivals, over which packets are dropped with
   uniform probability.  For instance, in a sample implementation, if
   this interval spanned 2000 packet arrivals, and a suitable
   probability of drop was 0.001, then two random variables would be
   drawn in a uniform distribution in the range of 1 to 2,000.  The
   values drawn would be used by counting to select the packets dropped
   at arrival.  The second parameter is the value for the probability of
   drop.  This parameter would be a function of an estimate of the
   number of users, their appropriate control intervals, and possibly
   the length of time that congestion has persisted.  [Jac89] has
   suggested successively increasing the probability of drop when
   congestion persists over multiple control intervals.  The motivation
   for increasing the packet drop probability is that the implicit
   estimate of the number of users and random selection of their packets
   to drop does not guarantee that all users have received enough
   signals to decrease demand.  Increasing the probability of drop
   increases the probability that enough feedback is provided.
   Congestion detection is also needed in Random Drop congestion
   avoidance, and could be implemented in a variety of ways.  The
   simplest is a static threshold, but dynamically averaged measures of
   demand or utilization are suggested.

   The packets dropped in Random Drop congestion avoidance would not be
   from the initial inputs to the gateway.  We suggest that they would
   be selected only from packets destined for the resource which is



Performance and Congestion Control Working Group               [Page 11]

RFC 1254           Gateway Congestion Control Survey         August 1991


   predicted to be approaching congestion.  For example, in the case of
   a gateway with multiple outbound links, access to each individual
   link is treated as a separate resource, the Random Drop policy is
   applied at each link independently.  Random Drop congestion avoidance
   would provide uniform treatment of all cooperating transport users,
   even over individual patterns of traffic multiplexed within one
   user's stream.  There is no aggregation of users.

   Simulation studies [Zha89], [Has90] have presented evidence that
   Random Drop is not fair across cooperating and non-cooperating
   transport users.  A transport user whose sending policies included
   Go-Back-N retransmissions and did not include Slow-start received an
   excessive share of bandwidth from a simple implementation of Random
   Drop.  The simultaneously active Slow-start users received unfairly
   low shares.  Considering this, it can be seen that when users do not
   respond to control, over a prolonged period, the Random Drop
   congestion avoidance mechanism would have an increased probability of
   penalizing users with lower demand.  Their packets dropped, these
   users exert the controls leading to their giving up bandwidth.

   Another problem can be seen to arise in Random Drop [She89] across
   users whose communication paths are of different lengths.  If the
   path spans congested resources at multiple gateways, then the user's
   probability of receiving an unfair drop and subsequent control (if
   cooperating) is exponentially increased.  This is a significant
   scaling problem.

   Unequal paths cause problems for other congestion avoidance policies
   as well.  Selective Feedback Congestion Indication was devised to
   enhance Congestion Indication specifically because of the problem of
   unequal paths.  In Fair Queueing by source-destination pairs, each
   path gets its own queue in all the gateways.

3.3  Congestion Indication

   The Congestion Indication policy is often referred to as the DEC Bit
   policy. It was developed at DEC [JRC87], originally for the Digital
   Network Architecture (DNA).  It has also been specified for the
   congestion avoidance of the ISO protocols TP4 and CLNP [NIST88].
   Like Source Quench, it uses explicit communications from the
   congested gateway to the user.  However, to use the lowest possible
   network resources for indicating congestion, the information is
   communicated in a single bit, the Congestion Experienced Bit, set in
   the network header of the packets already being forwarded by a
   gateway.  Based on the condition of this bit, the end-system user
   makes an adjustment to the sending window. In the NSP transport
   protocol of DECNET, the source makes an adjustment to its window; in
   the ISO transport protocol, TP4, the destination makes this



Performance and Congestion Control Working Group               [Page 12]

RFC 1254           Gateway Congestion Control Survey         August 1991


   adjustment in the window offered to the sender.

   This policy attempts to avoid congestion by setting the bit whenever
   the average queue length over the previous queue regeneration cycle
   plus part of the current cycle is one or more.  The feedback is
   determined independently at each resource.

3.4  Selective Feedback Congestion Indication

   The simple Congestion Indication policy works based upon the total
   demand on the gateway.  The total number of users or the fact that
   only a few of the users might be causing congestion is not
   considered.  For fairness, only those users who are sending more than
   their fair share should be asked to reduce their load, while others
   could attempt to increase where possible.  In Selective Feedback
   Congestion Indication, the Congestion Experienced Bit is used to
   carry out this goal.

   Selective Feedback works by keeping a count of the number of packets
   sent by different users since the beginning of the queue averaging
   interval.  This is equivalent to monitoring their throughputs. Based
   on the total throughput, a fair share for each user is determined and
   the congestion bit is set, when congestion approaches, for the users
   whose demand is higher than their fair share.  If the gateway is
   operating below the throughput-delay knee, congestion indications are
   not set.

   A min-max algorithm used to determine the fair share of capacity and
   other details of this policy are described in [RJC87].  One parameter
   to be computed is the capacity of each resource to be divided among
   the users.  This metric depends on the distribution of inter-arrival
   times and packet sizes of the users.  Attempting to determine these
   in real time in the gateway is unacceptable.  The capacity is instead
   estimated from on the throughput seen when the gateway is operating
   in congestion, as indicated by the average queue length.  In
   congestion (above the knee), the service rate of the gateway limits
   its throughput.  Multiplying the throughput obtained at this
   operating point by a capacity factor (between 0.5 and 0.9) to adjust
   for the distributions yields an acceptable capacity estimate in
   simulations.

   Selective Feedback Congestion Indication takes as input a vector of
   the number of packets sent by each source-destination pair of end-
   systems.  Other alternatives include 1) destination address, 2)
   input/output link, and 3) transport connection (source/destination
   addresses and ports).

   These alternatives give different granularities for fairness.  In the



Performance and Congestion Control Working Group               [Page 13]

RFC 1254           Gateway Congestion Control Survey         August 1991


   case where paths are the same or round-trip times of users are close
   together, throughputs are equalized similarly by basing the selective
   feedback on source or destination address.  In fact, if the RTTs are
   close enough, the simple congestion indication policy would result in
   a fair allocation.  Counts based on source/destination pairs ensure
   that paths with different lengths and network conditions get a fair
   throughput at the individual gateways.  Counting packets based on
   link pairs has a low overhead, but may result in unfairness to users
   whose demand is below the fair share and are using a long path.
   Counts based on transport layer connection identifiers, if this
   information was available to Internet gateways, would make good
   distinctions, since the differences of demand of different
   applications and instances of applications would be separately
   monitored.

   Problems with Selective Feedback Congestion Indication include that
   the gateway has significant processing to do.  With the feasible
   choice of aggregation at the gateway, unfairness across users within
   the group is likely.  For example, an interactive connection
   aggregated with one or more bulk transfer connections will receive
   congestion indications though its own use of the gateway resources is
   very low.

3.5  Fair Queueing

   Fair Queueing is the policy of maintaining separate gateway output
   queues for individual end-systems by source-destination pair.  In the
   policy as proposed by [Nag85], the gateway's processing and link
   resources are distributed to the end-systems on a round-robin basis.
   On congestion, packets are dropped from the longest queue.  This
   policy leads to equal allocations of resources to each source-
   destination pair.  A source-destination pair that demands more than a
   fair share simply increases its own queueing delay and congestion
   drops.

3.5.1  Bit-Round Fair Queueing

   An enhancement of Nagle Fair Queueing, the Bit-Round Fair Queuing
   algorithm described and simulated by [DKS89] addresses several
   shortcomings of Nagle's scheme. It computes the order of service to
   packets using their lengths, with a technique that emulates a bit-
   by-bit round-robin discipline, so that long packets do not get an
   advantage over short ones.  Otherwise the round-robin would be
   unfair, for example, giving more bandwidth to hosts whose traffic is
   mainly long packets than to hosts sourcing short packets.

   The aggregation of users of a source-destination pair by Fair
   Queueing has the property of grouping the users whose round-trips are



Performance and Congestion Control Working Group               [Page 14]