Date: Fri, 5 Mar 2010 14:13:59 +0000 (UTC) From: Luigi Rizzo <luigi@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r204758 - head/sbin/ipfw Message-ID: <201003051413.o25EDxwM019870@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: luigi Date: Fri Mar 5 14:13:58 2010 New Revision: 204758 URL: http://svn.freebsd.org/changeset/base/204758 Log: more documentation on new dummynet features. Modified: head/sbin/ipfw/ipfw.8 Modified: head/sbin/ipfw/ipfw.8 ============================================================================== --- head/sbin/ipfw/ipfw.8 Fri Mar 5 13:29:05 2010 (r204757) +++ head/sbin/ipfw/ipfw.8 Fri Mar 5 14:13:58 2010 (r204758) @@ -1404,7 +1404,7 @@ If not found, the match fails. Otherwise, the match succeeds and .Cm tablearg is set to the value extracted from the table. -.Br +.Pp This option can be useful to quickly dispatch traffic based on certain packet fields. See the @@ -1847,7 +1847,7 @@ is also the user interface for the .Nm dummynet traffic shaper, packet scheduler and network emulator, a subsystem that can artificially queue, delay or drop packets -emulator the behaviour of certain network links +emulating the behaviour of certain network links or queueing systems. .Pp .Nm dummynet @@ -1859,26 +1859,33 @@ Matching packets are then passed to eith different objects, which implement the traffic regulation: .Bl -hang -offset XXXX .It Em pipe -A pipe emulates a link with given bandwidth, propagation delay, +A +.Em pipe +emulates a +.Em link +with given bandwidth and propagation delay, +driven by a FIFO scheduler and a single queue with programmable queue size and packet loss rate. -Packets are queued in front of the pipe as they come out from the classifier, -and then transferred to the pipe according to the pipe's parameters. +Packets are appended to the queue as they come out from +.Nm ipfw , +and then transferred in FIFO order to the link at the desired rate. .It Em queue -A queue +A +.Em queue is an abstraction used to implement packet scheduling using one of several packet scheduling algorithms. -.Pp -The queue associates a -.Em weight -and a reference scheduler to each flow (a flow is a set of packets -with the same addresses and ports after masking). -A scheduler in turn is connected to a pipe, and arbitrates -the pipe's bandwidth among backlogged flows according to +Packets sent to a +.Em queue +are first grouped into flows according to a mask on the 5-tuple. +Flows are then passed to the scheduler associated to the +.Em queue , +and each flow uses scheduling parameters (weight and others) +as configured in the +.Em queue +itself. +A scheduler in turn is connected to an emulated link, +and arbitrates the link's bandwidth among backlogged flows according to weights and to the features of the scheduling algorithm in use. -.Pp -Note that weights are not priorities; a flow with a lower weight -is still guaranteed to get its fraction of the bandwidth even if a -flow with a higher weight is permanently backlogged. .El .Pp In practice, @@ -1887,6 +1894,52 @@ can be used to set hard limits to the ba .Em queues can be used to determine how different flows share the available bandwidth. .Pp +A graphical representation of the binding of queues, +flows, schedulers and links is below. +.Bd -literal -offset indent + (flow_mask|sched_mask) sched_mask + +---------+ weight Wx +-------------+ + | |->-[flow]-->--| |-+ + -->--| QUEUE x | ... | | | + | |->-[flow]-->--| SCHEDuler N | | + +---------+ | | | + ... | +--[LINK N]-->-- + +---------+ weight Wy | | +--[LINK N]-->-- + | |->-[flow]-->--| | | + -->--| QUEUE y | ... | | | + | |->-[flow]-->--| | | + +---------+ +-------------+ | + +-------------+ +.Ed +It is important to understand the role of the SCHED_MASK +and FLOW_MASK, which are configured through the commands +.Dl "ipfw sched N config mask SCHED_MASK ..." +and +.Dl "ipfw queue X config mask FLOW_MASK ..." . +.Pp +The SCHED_MASK is used to assign flows to one or more +scheduler instances, one for each +value of the packet's 5-fuple after applying SCHED_MASK. +As an example, using ``src-ip 0xffffff00'' creates one instance +for each /24 destination subnet. +.Pp +The FLOW_MASK, together with the SCHED_MASK, is used to split +packets into flows. As an example, using +``src-ip 0x000000ff'' +together with the previous SCHED_MASK makes a flow for +each individual source address. In turn, flows for each /24 +subnet will be sent to the same scheduler instance. +.Pp +The above diagram holds even for the +.Em pipe +case, with the only restriction that a +.Em pipe +only supports a SCHED_MASK, and forces the use of a FIFO +scheduler (these are for backward compatibility reasons; +in fact, internally, a +.Nm dummynet's +pipe is implemented exactly as above). +.Pp There are two modes of .Nm dummynet operation: @@ -2087,9 +2140,36 @@ The following parameters can be configur .Pp .Bl -tag -width indent -compact .It Cm type Ar {fifo | wf2qp | rr | qfq} +specifies the scheduling algorithm to use. +.Bl -tag -width indent -compact +.It cm fifo +is just a FIFO scheduler (which means that all packets +are stored in the same queue as they arrive to the scheduler). +FIFO has O(1) per-packet time complexity, with very low +constants (estimate 60-80ns on a 2Ghz desktop machine) +but gives no service guarantees. +.It Cm wf2qp +implements the WF2Q+ algorithm, which is a Weighted Fair Queueing +algorithm which permits flows to share bandwidth according to +their weights. Note that weights are not priorities; even a flow +with a minuscule weight will never starve. +WF2Q+ has O(log N) per-packet processing cost, where N is the number +of flows, and is the default algorithm used by previous versions +dummynet's queues. +.It Cm rr +implements the Deficit Round Robin algorithm, which has O(1) processing +costs (roughly, 100-150ns per packet) +and permits bandwidth allocation according to weights, but +with poor service guarantees. +.It Cm qfq +implements the QFQ algorithm, which is a very fast variant of +WF2Q+, with similar service guarantees and O(1) processing +costs (roughly, 200-250ns per packet). +.El .El .Pp -plus all the parameters allowed for a pipe. +In addition to the type, all parameters allowed for a pipe can also +be specified for a scheduler. .Pp Finally, the following parameters can be configured for both pipes and queues:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201003051413.o25EDxwM019870>