FQ-PIE queue disc¶
This chapter describes the FQ-PIE ([Ram19]) queue disc implementation in ns-3.
The FQ-PIE queue disc combines the Proportional Integral Controller Enhanced (PIE) AQM algorithm with the FlowQueue scheduler that is part of FQ-CoDel (also available in ns-3). FQ-PIE was introduced to Linux kernel version 5.6.
Model Description¶
The source code for the FqPieQueueDisc is located in the directory
src/traffic-control/model and consists of 2 files fq-pie-queue-disc.h
and fq-pie-queue-disc.cc defining a FqPieQueueDisc class and a helper
FqPieFlow class. The code was ported to ns-3 based on Linux kernel code
implemented by Mohit P. Tahiliani.
This model calculates drop probability independently in each flow queue. One difficulty, as pointed out by [CableLabs14], is that PIE calculates drop probability based on the departure rate of a (flow) queue, which may be more highly variable than the aggregate queue. An alternative, which CableLabs has called SFQ-PIE, is to calculate an overall drop probability for the entire queue structure, and then scale this drop probability based on the ratio of the queue depth of each flow queue compared with the depth of the current largest queue. This ns-3 model does not implement the SFQ-PIE variant described by CableLabs.
References¶
Ramakrishnan, M. Bhasi, V. Saicharan, L. Monis, S. D. Patil and M. P. Tahiliani, “FQ-PIE Queue Discipline in the Linux Kernel: Design, Implementation and Challenges,” 2019 IEEE 44th LCN Symposium on Emerging Topics in Networking (LCN Symposium), Osnabrueck, Germany, 2019, pp. 117-124,
White, Active Queue Management in DOCSIS 3.X Cable Modems, CableLabs white paper, May 2014.
Attributes¶
The key attributes that the FqPieQueue class holds include the following. First, there are PIE-specific attributes that are copied into the individual PIE flow queues:
UseEcn:Whether to use ECN markingMarkEcnThreshold:ECN marking threshold (RFC 8033 suggests 0.1 (i.e., 10%) default).UseL4s:Whether to use L4S (only mark ECT1 packets at CE threshold)MeanPktSize:Constant used to roughly convert bytes to packetsA:Alpha value in PIE algorithm drop probability calculationB:Beta value in PIE algorithm drop probability calculationTupdate:Time period to calculate drop probabilitySupdate:Start time of the update timerDequeueThreshold:Minimum queue size in bytes before dequeue rate is measuredQueueDelayReference:AQM latency targetMaxBurstAllowance:AQM max burst allowance before random dropUseDequeueRateEstimator:Enable/Disable usage of Dequeue Rate EstimatorUseCapDropAdjustment:Enable/Disable Cap Drop Adjustment feature mentioned in RFC 8033UseDerandomization:Enable/Disable Derandomization feature mentioned in RFC 8033
Second, there are QueueDisc level, or FQ-specific attributes::
* MaxSize: Maximum number of packets in the queue disc
* Flows: Maximum number of flow queues
* DropBatchSize: Maximum number of packets dropped from the fat flow
* Perturbation: Salt value used as hash input when classifying flows
* EnableSetAssociativeHash: Enable or disable set associative hash
* SetWays: Size of a set of queues in set associative hash
Examples¶
A typical usage pattern is to create a traffic control helper and to configure the type and attributes of queue disc and filters from the helper. For example, FqPIE can be configured as follows:
TrafficControlHelper tch;
tch.SetRootQueueDisc ("ns3::FqPieQueueDisc", "DropBatchSize", UintegerValue (1)
"Perturbation", UintegerValue (256));
QueueDiscContainer qdiscs = tch.Install (devices);
Validation¶
The FqPie model is tested using FqPieQueueDiscTestSuite class defined in src/test/ns3tc/fq-pie-queue-test-suite.cc.
The tests are similar to the ones for FqCoDel queue disc mentioned in first section of this document. The test suite can be run using the following commands:
$ ./ns3 configure --enable-examples --enable-tests
$ ./ns3 build
$ ./test.py -s fq-pie-queue-disc
or:
$ NS_LOG="FqPieQueueDisc" ./ns3 run "test-runner --suite=fq-pie-queue-disc"