Tài liệu Queuing Mechanisms doc

Queuing Mechanisms

Overview

This module describes the queuing mechanisms that can be used on output

interfaces.

It includes the following topics:

n Queuing Overview

n FIFO Queuing

n Priority Queuing

n Custom Queuing

n Weighted Fair Queuing

n Distributed Weighted Fair Queuing

n Modified Deficit Round-robin

n IP RTP Prioritization

Objectives

Upon completion of this module, you will be able to perform the following tasks:

n Describe and configure FIFO Queuing (FQ)

n Describe and configure Priority Queuing (PQ)

n Describe and configure Custom Queuing (CQ)

n Describe and configure basic Weighted Fair Queuing (WFQ), distributed WFQ,

ToS-based distributed WFQ and QoS-group-based distributed WFQ

n Describe and configure Modified Weighted Round-robin (MDRR) queuing

n Describe and configure IP RTP Prioritization

3-2 Queuing Mechanisms Copyright ” 2001, Cisco Systems, Inc.

Queuing Overview

Objectives

Upon completion of this lesson, you will be able to perform the following tasks:

n Understand how queuing works on Cisco routers

n List the most used queuing mechanisms

Queuing in Cisco IOS

• Cisco routers running Cisco IOS have a number of

different queuing mechanisms

• This module focuses on the following:

– First In First Out (FIFO)

– Priority Queuing (PQ)

– Custom Queuing (CQ)

– Weighted Fair Queuing (WFQ) with the different

distributed versions

– Modified Deficit Round Robin (MDRR)

– IP RTP Prioritization

• These mechnisms are implemented as software

queues

The lesson discusses how output queuing mechanisms are implemented on Cisco

routers running Cisco IOS. It discusses most of the queuing mechanisms in detail,

except Class-based Weighted Fair Queuing and Class-based Low-latency

Queuing, which are discussed in the “IP QoS – Modular QoS CLI (Chapter 2)”

module.

3-4 Queuing Mechanisms Copyright ” 2001, Cisco Systems, Inc.

Output Interface Queue Structure

• Each Interface has its hardware and software queuing

system

• The hardware queuing system always uses FIFO queuing

(Transmission queue or TxQ)

• The software queuing system can be selected and

configured depending on the platform and Cisco IOS

version

Hardware

Queue

(TxQ)

Software

Queuing

System

Output

Interface Forwarder

Any supported Always FIFO

queuing mechanism

Queuing on routers is necessary to accommodate bursts when the arrival rate of

packets is greater than the departure rate due to one of the following two reasons:

n Input interface is faster than the output interface

n Output interface is receiving packets coming in from multiple other interfaces

Initial implementations of queuing used a single FIFO (first-in first-out or first-come

first-serve queuing) strategy. More complex queuing mechanisms were introduced

when special requirements need routers to differentiate between packets of

different importance.

Queuing was split into two parts:

n The hardware queue that still uses FIFO strategy, which is necessary for the

interface drivers to transmit packets one by one. The hardware queue is

sometimes referred to as the transmit queue or TxQ.

n The software queue that schedules packets into the hardware queue based on

the QoS requirements

Listed on the previous graphic are some of the available software queuing

strategies with their goals:

n FIFO: no differentiation of packets (true fairness but no guarantees)

n Priority Queuing (PQ): strict prioritizing of packets

n Custom Queuing (CQ): service (bandwidth) guaranteed to up to 16 classes

n Weighted Fair Queuing (WFQ) and Distributed WFQ: service (bandwidth)

guarantee to individual flows

n Distributed ToS-based WFQ: service (bandwidth) guaranteed to up to 4 classes

n Distributed QoS-group-based WFQ: service (bandwidth) guaranteed to up to

100 classes

n Modified Deficit Round-robin (MDRR): service (bandwidth) guaranteed to up

to 8 classes; low-delay guarantee if Strict or Alternate Priority is used

n IP RTP Prioritization: low-delay guarantee

Most queuing mechanisms depend on the availability on different platforms and

Cisco IOS versions. For example:

n MDRR is only available on Cisco 12000 series routers (GSR)

n Distributed ToS-based and QoS-group-based WFQ are only available on Cisco

7x00 series routers with Versatile Interface Processors (VIP)

3-6 Queuing Mechanisms Copyright ” 2001, Cisco Systems, Inc.

Bypassing the Software Queue

• When a packet is being forwarded the router will bypass the

software queue if:

– the software queue is empty and

– the hardware queue is not full

Software Queue

Empty?

Hardware Queue

Full?

Hardware

Queue

(TxQ)

Yes No

Software

Queuing

System

No Yes

The implementation of software queuing was optimized for periods when the

interface is not congested. The software queuing system is bypassed whenever

there is no packet in the software queue and there is room in the hardware queue.

The software queue is, therefore, only used when data must wait to be placed into

the hardware queue.

Hardware Queue (TxQ) Size

• Routers determine the length of the hardware

queue based on the configured bandwidth of

the interface

• Long TxQ may result in poor performance of

the software queue

• Short TxQ may result in a large number of

interrupts which causes high CPU utilization

and low link utilization

The double queuing strategy (software and hardware queue) has its impacts on the

result of overall queuing:

n Software queue is used for a certain reason. If the hardware queue is too long

it will contain a large number of packets scheduled in the FIFO fashion. This is

probably against the QoS design that required a certain complex software

queuing system (for example, Custom Queuing).

So why use the hardware queue at all? Or why not just set its length to one? That

would force all packets to go through the software queue and be scheduled one by

one to the interface for transmission. This approach has the following drawbacks:

n Each time a packet is transmitted, the interface driver interrupts the CPU and

requests more packets to be delivered into its hardware queue. Some queuing

mechanisms have complex scheduling that takes time to deliver more packets.

The interface does not send anything during that time (link utilization is

decreased) if the hardware queue is empty because its maximum size is one.

n The CPU schedules packets one by one instead of many at the same time (in

the same interrupt interval). This increases the CPU utilization.

Choosing the appropriate length of the hardware queue is very important. The

default TxQ size is determined by the IOS based on the bandwidth of the media

and should be fine for most queuing implementations. Faster interfaces have longer

hardware queues because they produce less delay. Slower interfaces have shorter

hardware queues to prevent too much delay in the worst-case scenario where the

entire hardware queue is full of MTU-sized packets.

3-8 Queuing Mechanisms Copyright ” 2001, Cisco Systems, Inc.

Software Queuing System

Hardware

Queuing System

Queuing Components

• Each queuing mechanism has three main components that define it:

– Classification (selecting the class)

– Insertion policy (determining whether a packet can be enqueued)

– Service policy (scheduling packets to be put into the hardware queue)

Class 1?

Class 2?

Class n?

Queue 1

Queue 2

Queue n

Scheduler Interface

Forwarded Packets

Hardware Q

Add/Drop

The figure illustrates the actions that have to be taken before a packet can be

transmitted:

n Most queuing mechanisms include classification of packets.

n Once a packet is classified, a router has to determine whether it can put the

packet into the queue or it has to drop the packet. Most queuing mechanisms

will drop a packet only if the corresponding queue is full (tail-drop). Some

mechanisms use a more intelligent dropping scheme (Weighted Fair Queuing)

or a random dropping scheme (Weighted Random Early Detection).

n If the packet is allowed to be enqueued it will be put into the FIFO queue for

that particular class.

n Packets are then taken from the individual per-class queues and put into the

hardware queue.

Queuing systems differ in the following ways:

n Classification options: some mechanisms classify packets automatically (for

example, WFQ), while other mechanisms require manual configuration of

classification (for example, PQ or CQ).

n Insertion policy: most queuing mechanisms use the tail-dropping scheme.

n Scheduling policy: this is the most important part of every queuing mechanism

because it determines the order in which the packets will leave the router.

Summary

After completing this lesson, you should be able to perform the following tasks:

n Understand how queuing works on Cisco routers

n List the most used queuing mechanisms

Review Questions

Answer the following questions:

n Which queuing mechanisms do Cisco routers support?

n When is a software queuing mechanisms not used?

n How does TxQ length affect the software queuing system?

3-10 Queuing Mechanisms Copyright ” 2001, Cisco Systems, Inc.

FIFO Queuing

Objectives

Upon completion of this lesson, you will be able to perform the following tasks:

n Describe FIFO queuing

n Describe the drawbacks of FIFO queuing

n Configure FIFO queuing on Cisco routers

n Monitor and troubleshoot FIFO queuing

FIFO Queuing

• Software FIFO queue is basically an extension to the

hardware FIFO queue

FIFO Queuing System Hardware

Queuing System

All in one

queue

Queue 1 FIFO

Scheduler Interface

Forwarded Packets

Tail-drop Hardware Q

Routers serve packets in the

first-come first-serve fashion

FIFO uses one single queue

Newly arriving packets are

dropped if the queue is full All packets are

classified into one class

FIFO queuing has no classification because all packets belong to the same class.

Packets are dropped when the output queue is full (tail-drop). The scheduler

services packets in the order they arrived.

Software FIFO queue is basically an extension of the hardware FIFO queue.

3-12 Queuing Mechanisms Copyright ” 2001, Cisco Systems, Inc.

Benefits and Drawbacks of FIFO

Queuing

+ Benefits

• Simple and fast (one single queue with a simple

scheduling mechanism)

• Supported on all platforms

• Supported in all switching paths

• Supported in all IOS versions

– Drawbacks

• Unfair allocation of bandwidth among multiple flows

• Causes starvation (aggressive flows can monopolize

links)

• Causes jitter (bursts or packet trains temporarily fill

the queue)

FIFO queuing might be regarded as the fairest queuing mechanism but it has a long

list of drawbacks:

n FIFO does not fairly allocate bandwidth among multiple flows. Some flows

receive more bandwidth because they use larger packets or send more packets.

n FIFO is extremely unfair when an aggressive flow is contesting with a fragile

flow. Aggressive flows send a large number of packets, many of which are

dropped. Fragile flows send a modest amount of packets and most of them are

dropped because the queue is always full due to the aggressive flow. This type

of behavior is called starvation.

n Short or long bursts cause a FIFO queue to fill. Packets entering an almost full

queue have to wait a long time before they can be transmitted. Another time,

the queue might be empty causing packets of the same flow to experience

almost no delay. Variation in delay is called jitter.

In spite of all the drawbacks FIFO is still the most used queuing mechanism

because of the following benefits:

n It is simple and fast. Most high-end routers with fast interfaces are not really

challenged by the drawbacks mentioned earlier. Furthermore, routers are not

capable of complex classification and scheduling when they have to process a

large number of packets per second. FIFO is, therefore, the most suitable

queuing mechanisms on these platforms.

n It is supported on all platforms.

n It is supported in all IOS versions.

Configuring FIFO Queuing

Router(config-if)#

• FIFO queuing is enabled by default on all interfaces

that have a default bandwidth of more than 2Mbsp

• Weighted Fair Queuing is enabled if the bandwidth is

less than 2Mbps

• Disable WFQ to enable FIFO on interfaces that have

less than 2Mbps of bandwidth

no fair-queue

Cisco routers have two default queuing mechanisms:

n All interfaces with the default bandwidth above 2Mbps use FIFO queuing. No

configuration is necessary on such interfaces.

n All interfaces with the default bandwidth below 2Mbps use Weighted Fair

Queuing (WFQ). The no fair-queue command must be used to disable WFQ

and enable FIFO.

There is no special command that specifically enables FIFO.

3-14 Queuing Mechanisms Copyright ” 2001, Cisco Systems, Inc.

Configuring FIFO Queuing

Router(config-if)#

• FIFO queuing allows a maximum of 40 packets to be

stored in the output queue

• This command can be used to increase or decrease

the maximum number of buffered packets

• A large value can be set to support longer bursts

(less drops, more buffer usage)

• A small value can be set to prevent bursts (more

drops)

hold-queue <buffers> out

One of the considerations when using FIFO queuing is the maximum burst size.

Routers allow (by default) up to 40 packets to be in the output queue. Shortening

the queue causes more drops, especially for bursty sessions. Lengthening the

queue allows more packets to be enqueued. A long queue should be used to allow

bursts without drops.

The hold-queue command is used to set the maximum number of packets in the

output queue.

Thư viện tri thức trực tuyến

Tài liệu Queuing Mechanisms doc

Nội dung xem thử

Mô tả chi tiết

Tài liệu tương tự (6)

Tài liệu QUEUEING THEORY WITH APPLICATIONS TO PACKET TELECOMMUNICATION pptx

Tài liệu Queueing mạng lưới và chuỗi Markov P11 pptx

Tài liệu Queueing mạng lưới và chuỗi Markov P12 ppt

Tài liệu Queueing mạng lưới và chuỗi Markov P13 docx

Tài liệu Queueing mạng lưới và chuỗi Markov P10 pdf

Tài liệu