QoS Technology in network

QoS summary

Whether it's voice or video , stay IP Online IP Data packet transmission , The required bandwidth can be fully satisfied by the broadband multi service network . But in order to ensure that under the interference of other data , Or during peak congestion , At many to one traffic convergence points , Meet the bandwidth of key services 、 Time delay 、 Requirements for jitter and packet loss , You need the network to provide the necessary QoS（ Service quality assurance ） technology .

Different data types , For example, voice. 、 Images , Successful transmission in the network has different requirements .IP Network QoS Give the device an intelligence , Enable them to prioritize certain data based on network policies .QoS It allows network managers to control network bandwidth 、 Delay 、 Jitter and data loss .QoS Not a function on the device , It is not only a function of the data link layer ,QoS It is an end-to-end system architecture , A powerful QoS The solution includes a wide range of technologies , And provide good scalability and services that do not depend on any media in the whole network , And it has the ability of system function detection .

QoS The framework of technology can be roughly divided into ： classification 、 Inflow control 、 line up 、 Flow control .

Classification technology enables us to select important information from a large number of information packets / Is not important 、 Security / Unsafe and other traffic , And treat them differently . It can be used for QoS guarantee , It can also be used for security filtering . Typical classification methods are based on Ip Address 、 be based on TCP/UCP Protocol type and port number, etc . Advanced classification can be based on time （ For example, a special application that starts at night ）、 Based on packet size 、 E-commerce applications, etc , It can also be classified according to the content , For example, it contains reactionary 、 Pornographic content Web Page etc. . Classification techniques compare losses CPU resources , Therefore, it is generally completed at the edge of the network , Put a unified label after classification （ Such as IP Precedence or DSCP etc. ）, So that other devices on the network do not have to be reclassified , And directly according to the label QoS. Therefore, the number range of the label will determine how many traffic categories the whole network can distinguish .

For public security high-speed network , It is recommended not to distinguish too many categories , Because too many categories mean complex queue management 、 Complex bandwidth control and additional switching processing . No more categories , But the intelligent control of a certain type of traffic , Such as bandwidth range 、 Packet loss strategy 、 Auditing and other functions are very important . Common classification techniques include CAR、ACL etc. .

Flow adjustment is required after flow classification , That is, set the allowable transmission bandwidth range according to its category , Including input bandwidth and output bandwidth control . Common technologies include CBWFQ、CAR、WRED etc. .

The queuing function ensures that important traffic categories are transmitted first , No packet loss during congestion , Fair treatment of different data of the same category . Common technologies are WFQ、CBWFQ、MDRR etc. .

Investigate QoS Technology must also consider its impact on network performance , Mature network devices can activate a variety of QoS High forwarding rate is maintained after the technology .

To make the network QoS Deploy correctly and get the most out of it , All parts need to coordinate . because QoS A lot of computation is required , These jobs are allocated to edge and core devices , They may be switches or routers , This method is like a low-speed high contact edge device and high-speed 、 Low touch core devices to optimize network efficiency and scalability . Edge switches and routers do a lot of processing work , Identify data flow and classify data packets based on user policies , Edge devices provide bandwidth management , Core device acceleration has QoS Packet forwarding .

Qos Technology and QoS Solution

QoS The aim is to provide better 、 More predictive Web Services , It provides dedicated bandwidth 、 Controllable jitter and delay , And can improve the lost characteristics .QoS The way to achieve these goals is to provide a set of tools to manage network congestion . Shaping network traffic 、 More efficient use of expensive wide area links , And set the traffic policy in the network .

QoS The tool has three main functions

• Congestion management （ Queuing and scheduling ）、
• Congestion avoidance 、
• Traffic shaping and policy making .

These tools are used within a single network element . Generally speaking , These tools can be launched at an interface provider to provide correct... For special network applications QoS characteristic . in addition , The network device software also provides a link efficiency mechanism integrated with the other three functions , Different tools can work together , Can be improved QoS service .

Congestion management tools

One of the ways for network components to deal with the overflow of incoming traffic is to use a queuing algorithm to classify the traffic , Then select a method to prioritize it , Input output link . The following queuing algorithms can be used to achieve queuing ：

• fifo （FIFO） line up
• Priority queuing （PQ）
• Custom queuing （CQ）
• Weighted fair queuing （WFQ）

Each queuing algorithm is designed to solve specific network traffic problems , It has different effects on network performance , The details are as follows ：

1．FIFO It has basic store and forward functions

FIFO Is the simplest way to queue up . When the network is congested , It can store packets , And when the congestion disappears, it will be forwarded out according to its arrival order . In some cases ,FIFO Is the default queuing algorithm , Therefore, no configuration is required . But it has several disadvantages , most important of all FIFO Queuing does not consider the priority of packets , The order in which packets arrive will determine the bandwidth they use 、 Processing speed and buffer allocation . It does not prevent the application （ Source ） Malicious acts of . Groups of information sources will have a large delay in transmitting time sensitive application traffic , It will potentially affect the transmission of network control and signaling information . In controlling network traffic ,FIFO Queuing is only the first necessary step , Today's intelligent networks need more mature algorithms .

2．PQ Prioritize traffic

PQ Ensure that important traffic can be handled as quickly as possible where it is used . It is designed to provide strict prioritization for critical traffic . The priority queuing algorithm can be based on the network protocol （ Such as IP、IPX or AppleTalk）、 Input interface 、 Packet size 、 Source / The destination address can flexibly prioritize the traffic .

stay PQ In the algorithm, , According to the assigned priority , Each packet is placed in one of the four queues ： high 、 in 、 General or low-level queues . Packets without priority list classification will enter the general queue . In transmission , The algorithm will provide absolute priority for higher priority queues . This is a simple and intuitive method , However, this will randomly transfer the delay that higher priority traffic might have experienced to lower priority traffic , Thus, the jitter of lower priority traffic is increased . To solve this problem , It can limit the rate of higher priority traffic .

PQ It can play a great role in ensuring that mission critical traffic over various WAN links is prioritized .PQ Currently using static configuration , Therefore, it cannot be adjusted automatically , To adapt to changing network requirements .

3．CQ Ensure the use of bandwidth

Custom queuing （CQ） The design of the algorithm allows various applications or institutions to share the network with applications with specified minimum bandwidth or delay requirements . In this environment , Bandwidth must be allocated proportionally between applications and users . User available CQ Features provide bandwidth guarantee at potential congestion points , Ensure that the specified traffic gets a fixed percentage of the available bandwidth , The remaining bandwidth is used by other traffic . Custom queuing processes traffic by providing a specific amount of queue space for each type of packet , Then service each queue in a round robin fashion .

This queuing algorithm can put information into 17 One of the queues （ queue 0 The information system stores , If it remains active 、 Signaling, etc ）, And vacate according to the weighted priority . The router routes queues in a round robin fashion 1 To 16 Serve in turn , In each cycle, data is fetched from each queue according to the configured bytes . This feature can ensure that when the line load is heavy , Any application （ Or specified application group ） Can not use more than a predetermined proportion of capacity . And PQ equally ,CQ It is also statically configured , It cannot be adjusted automatically , To adapt to changing network requirements .

4．WFQ： It is suitable for today's network intelligent queuing tools

In some cases , It is necessary to provide consistent response time for network users with more and less traffic , Without increasing bandwidth , The solution is WFQ.WFQ Is a major queuing technology , It is a flow based queuing algorithm , But two things at the same time ： Scheduling interactive traffic to the front of the queue has less response time , And make each high bandwidth stream share the remaining capacity fairly .

WFQ The design minimizes configuration work , And it can be automatically adjusted according to the changing network traffic . in fact ,WFQ It provides excellent service for most applications , Has become the most configured to E1 Or below E1 The default queuing mode on the serial interface for rate operation （2.048M bps）.

WFQ Very efficient , In the absence of higher priority flows , It can use any available bandwidth to send traffic from lower priority streams . This is similar to time division multiplexing （TDM） Is different , because TDM Simply divide the bandwidth , If there is no specific traffic type in the network , Bandwidth will be wasted .WFQ And IP Priority and RSVP The common use , Can provide differentiation QoS And security QoS service .

WFQ The algorithm also solves the problem of cyclic delay variability . If multiple high-volume sessions are active , Then its transmission rate and interval time are easier to predict .WFQ Some algorithms have been greatly improved , Such as SNA Logical link control （ILC）、 Transmission control protocol （TCP） Congestion and slow start characteristics . The result is a more predictable throughput and response time for each active flow .

WFQ and QoS Signaling technology sharing , It is the most effective queuing technology at present .

WFQ Be familiar with IP Priority , in other words , It can detect that IP High priority packets marked as priority by the forwarder , And provide faster scheduling and excellent response time for these traffic .IP The value of priority is 0（ The default value ） To 7. With this priority worth increasing , The algorithm will allocate more bandwidth to it , To ensure that congestion occurs , It can get faster service .WFQ Assign a weight to each flow , This weight determines the transmission order of queued packets . In this way , The information flow with low value is served first .Ip The priority value will be the divisor of the weight . for instance ,IP The priority field value is 7 Of traffic , Its weight is relatively IP The priority field value is 3 The flow rate should be low , So it takes precedence in the transmission order .

for example ： Suppose there is an information flow for each priority on an interface , Each gets the priority value of the total ＋1 Partial link , namely ：1＋2＋3＋4＋5＋6＋7＋8＝36 That is to say, each stream will obtain 8/36、7/36、6/36、5/36 wait . However , If there is 18 The priority values are 2 The flow of , Then the above formula will become ：1＋18×2＋3＋4＋5＋6＋7＋8＝70

Then, each stream will get a link capacity of 8/70、7/70、……1/70, in addition 18 Each stream will get about 2/70 Link capacity .

in addition , In a frame relay network , When congestion occurs, it can be notified clearly from the front （FECN） And backward congestion notification （BECN） Bits to identify . When the traffic is exchanged by the frame relay switching module , Frame relay discard qualification （DE）、FECN and BECN Bit will affect WFQ A weight . Once congestion is identified , The values in the algorithm will change , This will reduce the transmission frequency of the congested session .

Congestion avoidance tools

Congestion avoidance technology by monitoring network traffic load , It can predict and avoid the congestion at the bottleneck of public network . This is different from the congestion management technology that manages congestion when it occurs . The main tool to avoid congestion is weighted random detection （WRED）, It will be described below .

1．WRED To avoid congestion

Random early detection （RED） The algorithm can solve the network congestion problem before , To avoid congestion .RED Monitor traffic load at various points in the network , If congestion starts to increase , Random packet discarding measures will be taken . The result of discarding is that the information source will find that there is traffic loss , Thus reducing its transmission rate .RED Mainly in the IP In the Internet environment TCP Agreement sharing .

2．WRED And QoS Signaling technology sharing

WRED Combined with the IP Priority and RED The function of algorithm . This combination can provide priority traffic processing for higher priority packets . When the interface begins to jam , It will selectively discard lower priority traffic , And provide different performance characteristics for different service levels .

Traffic shaping and policy tools

QoS The software solution includes two traffic shaping tools － Universal flow shaping （GTS） And frame relay traffic shaping （FRTS） Tools , Manage traffic and congestion in the network .

1．GTS control 、 Output information flow

Universal flow shaping （GTS） Provides a mechanism to control the flow of information on a specific interface . By limiting the rate of specified traffic （ Also known as token bucket method ）, It reduces output , Thus, congestion is avoided , At the same time, the burst of specific traffic is queued . such , Traffic that complies with specific standards can be shaped to meet the requirements of downstream flow , Eliminate the network topology bottleneck caused by data rate mismatch .

GTS Can be performed on each interface , The access list can be used to select the traffic to be shaped , And can be combined with various second layer technologies , Such as frame relay 、ATM、 Switched multi megabit data service （SMDS） And Ethernet .

In a frame, on the relay sub interface , adopt BECN Signal integration ,GTS Can be set , Dynamically adapt to changes in available bandwidth , Or directly set the shaping to a predetermined rate .GTS Also available at ATM/AIP Configure on the interface card , For static configuration ATM Permanent virtual circuit （RVC） Upper RSVP Respond to the signal .

2．FRTS Manage frame relay traffic

Frame relay traffic shaping （FRTS） The parameters provided can be used to manage network traffic congestion . Its functions include ： Commit information rate （CIR）、 Forward and backward explicit congestion notification （FECN/BECN） And disqualification （DE） position . It can improve the scalability and performance of frame relay network , Increase virtual circuit density , Other features to speed up response time ,FRTS Features can be built on this frame relay support .

for example , The user can force the rate （ Configured peak rate , The output flow can be limited ） configure CIR Or other defined values , For example, excessive information rate is configured on each virtual circuit （EIR）.

The user can also be in VC Or sub interface level to define priority and customize queuing algorithm . This will make traffic prioritization and queuing more refined , For each VC Give more control to the flow on . If you will CQ With each VC Queuing and rate enforcement on , Frame relay VC You can transfer multiple traffic types , Such as IP、 System network architecture （SNA） And internet packet exchange （IPZ）, Each traffic type will be guaranteed bandwidth .

FRTS A high-speed connection at the center and a low-speed connection at the branch can be used to eliminate the bottleneck in the frame relay network . The user can limit the central location through the configuration of rate enforcement VC Data sending rate on . Users can also force the rate to the existing DLCI Combine optimization features , Further performance .

FRTS Only used for frame relay permanent virtual connection （PVC） And switched virtual connections （SVC）.

According to the received from the network BECN Mark the information in the packet ,FRTS It can also reduce the flow dynamically . On the basis of BECN Weakening of , Packets can be placed in the router buffer , This reduces the data flow from the router into the frame relay network . This weakening occurs in every VC On , Transmission rate based on received BECN Mark the number of packets to adjust .

FRTS It also provides the ability to make multiple VC Mechanisms for sharing media . The rate forcing function allows the use of standards other than line speed （ Such as CIR or EIR） To control the transmission rate used by the router . The rate forcing feature can also be for each VC Preallocate bandwidth , Create a virtual time division multiplexing network .

Link efficiency mechanism

at present , There are two link efficiency mechanisms － Real time protocol header compression （RTP-HC） And link partition and insertion （LFI）. Use them with queuing and traffic shaping functions , Availability efficiency and predictability of application service level .

1．LFI Yes IP Traffic is divided and inserted

When the network processes large packets （ For example, through WAN Of LAN between FTP delivery ） when , Especially when they are queued on a lower speed chain , Interactive traffic （Telnet、IP Voice etc. ） The delay and jitter will increase . Link splitting and insertion （LFI） The feature obtains smaller packets by splitting large datagrams and inserting low latency traffic packets , Thus, the delay and jitter on the low-speed link are reduced .

LFI It is especially designed for low-speed links with large serial delay .LFI It is required to configure multiple links on the interface PPP, It also has the function of inserting . Related to this IETF The draft is called multi link PPP Multilevel extension of （MCML）, The function realized is related to LFI Basically the same .

2．RTP Header compression can increase the efficiency of real-time traffic

Real time transmission protocol （RTP） Is a host to host protocol , Applicable to newer multimedia application traffic , Include IP Packet based audio and video transmission over the network .RTP The end-to-end network transmission function provided is mainly aimed at real-time applications , Such as the audio of unicast or multicast network services 、 Video or analog data, etc .RTP The header compression function greatly improves the performance of many low-speed links RTP The relatively new IP Efficiency of voice or multimedia applications .

For net load compressed audio applications ,RTP The information package has 40 The header sum of bytes is generally 20 To 150 Net load of bytes . in consideration of IP/UDP/RTP The size of the header combination , Transmitting an uncompressed header will make it less efficient . By means of RTP/UDP/IP The header starts from 40 Bytes are compressed to 2 To 5 byte ,RTP Travel efficiency , Especially on the low-speed link, the operation efficiency is greatly improved . This is true for low-speed links （385K bps And the following ） Smaller packets on （ Such as IP Voice traffic ） Especially the key , This is because RTP Header compression can greatly reduce system overhead and transmission delay .

RTP Header compression reduces multimedia RTP The line overhead of traffic and its corresponding delay , This is especially true for traffic that uses packets and headers that are shorter .

Use frame relay 、HDLC or PPP Encapsulated serial lines also support RTP Header compression .ISDN This function is also supported on the interface . Related to this IETF The draft is called compression RTP（CRTP）, The functions defined therein are essentially the same .

QoS Signaling

Can be QoS Signaling is regarded as a way of network communication . It provides a method for a terminal or network element to make a specific request to an adjacent element . for example ,IP Network usable part IP Packet header to request special processing for priority traffic or time sensitive traffic .QoS The configuration aspect of services also plays a key role .

True end-to-end QoS Each element in the network path is required （ Switch 、 Router 、 A firewall 、 host 、 Client, etc ） All provide their own parts of QoS, And it has to go through QoS Signaling for collaboration . However , Looking for a powerful... That can be served in a diverse network infrastructure QoS Signaling solutions have become a huge challenge . Although there are many feasible QoS Signaling solutions can be provided in certain parts of the infrastructure QoS, But there are often limits in the network .

Here we focus on IP Priority and RSVP, This is because both methods make use of Ip End to end properties of the protocol . Whereas most applications use IP As its main networking protocol ,IP Priority and RSVP by QoS Signaling provides a method for QoS Powerful combination ,IP Priority signaling is used for differentiation QoS,RSVP To protect QoS.

1．IP The priority is differentiation QoS Provide signaling

IP priority of use Ipv4 Headlines TOS In domain 3 Priority bits to specify the service level for each packet . Use IP Priority can divide the flow into at most 6 Service levels （ The other two are used internally by the network ）. The queuing technology of the whole network can use this kind of signaling to provide corresponding fast processing .

Policy based routing and CAR And other features can be used to set the priority according to the classification of the extended access list . This will bring considerable flexibility to the allocation of priorities , In this way, according to the application or user 、 Destination and source subnets . Generally speaking , This function is as close to the edge of the network as possible （ Or administrative domain ） External implementation , such , Then each network element can provide services according to the set policy .

IP Priority can also be set in the host or network client , This kind of signaling is optional ; However , This signaling can be covered by policies in the network .

IP The existing network queuing mechanism can be used preferentially （ Such as WFQ or WRED） To achieve service levels , Users do not need to change existing applications or complex network requirements . Please note that , Use its priority , The same method can be easily extended to Ipv6.

2．RSVP Guarantee QoS

RSVP It's a kind of IETE Internet standard （RFC2205） agreement , Allows applications to dynamically reserve network bandwidth . Applications are available RSVP Request specific for data flow QoS.

Host and router use RSVP, Propose to the router along the data flow path QoS request , And maintain the router and host state to provide the required services （ Generally bandwidth and delay ）.RSVP According to the average data rate 、 The maximum and minimum amount of data that the router maintains in the queue QoS To determine bandwidth reservation .

WFQ or WRED yes RSVP Main force , Set up the required packet classification and scheduling for the reserved flow . Use WFQ,RSVP It can provide comprehensive business guarantee services . Use WRED, It provides controlled load Services .WFQ By accelerating the processing of interactive traffic , Share the remaining bandwidth fairly between high bandwidth streams , Continue to provide excellent handling for non reserved traffic .WRED For the wrong RSVP Traffic provides equal advantages .

Label switching enables flexible traffic engineering

The tag exchange function includes mechanisms that can be used with RSVP and IP Priority signaling for interoperability , And benefit from it . The label switching header contains a three bit domain that can be used as traffic priority signaling , It can also be used to map specific flows and traffic levels along a designed label switching path , In order to obtain the network tag switching part QoS.

Public Security IP Voice network and IP What should be adopted in the video network QoS Safeguards

IP Voice and IP Video as a real-time service , To provide a high quality of service , The bearer network is required to support voice 、 Video data provides low latency 、 Transmission performance with low delay jitter . Public Security IP Voice network 、IP Video network to public security IP The computer network is a unified bearing platform , Facing the increasingly abundant computer business , The bandwidth competition among various services will continue to intensify . To ensure the quality of voice and other real-time services , On the premise that the data transmission performance of the public security backbone network is less affected , Use the plan to cover the whole network QoS Strategy becomes inevitable .

1． Edge access network

Access the router at the edge 、 The access side of the switch uses CAR Yes, voice 、 High priority settings for video packets , At the same time, in order to protect voice 、 Basic bandwidth of video service , For other data services , Limit bandwidth ; The output side shall utilize WFQ、PQ Waiting for queuing technology , Ensure the priority transmission of high priority messages .

2. backbone network

Use the priority setting of the edge access network , use WFQ As a queuing technique , Efficiently complete message forwarding ; use WRED And other congestion avoidance technologies , Reasonably discard low priority messages , Avoid congestion .

3．RTP Application of header compression technology

use RTP Header compression technology , Lower voice 、 Video etc. RTP Bandwidth overhead of the message , Make full use of existing bandwidth .

4． Packet length adjustment technology under the condition of low bandwidth

stay 384K bps Under the following low bandwidth conditions , utilize LFI Wait for packet length adjustment technology , Adjust the package length reasonably ,, Avoid the delay of voice message caused by long data packet .