1. QoS Overview
Realizing the communication of any media information at any time, anywhere and anyone is an eternal need of human beings in the field of communication. Before the maturity of IP technology, all networks were a single service network. For example, PSTN could only operate telephone services, cable TV networks could only carry television services, and X.25 network could only carry data services, etc. The separation of the network causes the separation of services and reduces the efficiency of communication.
Due to the popularity of the Internet, IP applications are becoming increasingly widespread, and IP networks have penetrated into various traditional communication ranges, making it possible to build a multi-service network based on IP. However, different services have different requirements for the network. How to implement multiple real-time and non-real-time services in a packetized IP network has become an important topic. People have proposed the concept of QoS (Quality of Service).
IP QoS refers to an ability of an IP network, that is, to provide the services they need for specific services on IP networks spanning multiple underlying network technologies (FR, ATM, Ethernet, SDH, etc.). QoS includes multiple aspects, such as bandwidth, delay, delay jitter, etc. Each service has specific requirements for QoS. Some may have higher requirements for some of the indicators, while others may have higher requirements for other indicators. The technical indicators for measuring IP QoS include the following.
(1) Available bandwidth: refers to the average rate of specific application service flows between two nodes of the network. It mainly measures the user's ability to obtain service data from the network. All real-time services have certain requirements for bandwidth. For example, for video services, when the available bandwidth is lower than the encoding rate of the video source, the image quality cannot be guaranteed.
(2) Delay: refers to the average round-trip time of data packets being transmitted between two nodes of the network. All real-time services have certain requirements for delays. For example, VoIP services generally require the network delay to be less than 200ms. When the network delay is greater than 400ms, the call will become unbearable.
(3) Packet loss rate: refers to the percentage of lost packets during network transmission, which is used to measure the network's ability to correctly forward user data. Different services have different sensitivity to packet loss. In multimedia services, packet loss is the most fundamental reason for deterioration in image quality. A small amount of packet loss may cause mosaics in the image.
(4) Delay jitter: refers to changes in delay. Some businesses, such as streaming media services, can reduce the impact of delay jitter on the business through appropriate cache; while some businesses are very sensitive to delay jitter. For example, voice services, a slight delay jitter will lead to a rapid decline in voice quality.
(5) Packet error rate: refers to the percentage of errors in the message during network transmission. The bit error rate has a particularly great impact on some encryption data services.
In addition, QoS may also include some other metrics, such as network availability. QoS indicators are actually technical descriptions of business quality. For different businesses, when QoS lacks guarantee, the business appearance presented is different.
Currently, there are two main solutions for QoS: IntServ and DiffServ.
IntServ is an end-to-end stream-based QoS technology. Using IntServ, to some extent, adopts the connection-oriented idea in a circuit-switched network. Before communication between the two ends of the service, QoS requirements are required to be submitted to the network according to the service type. The network determines whether to accept the request of the service according to a certain acceptance policy control. If the network has sufficient resources to meet the requirements of the service, it will accept the service flow and must be responsible for ensuring the resources applied for by the service. End-to-end communication path is established through out-of-band RSVP signaling. Every network device along the way needs to record the status information of each service flow - "soft status", and provide corresponding resource reservations to ensure the service quality of the service.
From a technical point of view, RSVP is an effective QoS guarantee method on the current network. However, due to the characteristics of the IP network traffic model and business model, the Internet backbone network instantly needs to provide services to thousands of service flows. Therefore, the solution idea of reserved paths for individual flows cannot be expanded on the Internet backbone network, which seriously restricts the application of IntServ in actual networks. Of course, there are other factors that restrict IntServ applications, including the large-scale deployment of RSVP signaling, interoperability between equipment of different manufacturers, and service-based management (authentication, billing), etc. In addition, IntServ requires end-to-end network support. Due to the complexity of RSVP technology, it can be said to be one of the most complex IP technologies at present. It is unrealistic to upgrade all network hardware devices to support RSVP.
At the same time, the RSVP model actually IP-based translation of the concept of circuit switching services, which largely subverts the concept of open IP interconnection and jump-by-hop forwarding. There are many unknown problems in the integration with other IP services and technologies. Overall, IntServ is currently very practical.
DiffServ is a DSCP-based QoS solution launched by the IETF organization in 1998. This is a class-based QoS technology that is mainly used in backbone networks. Use DiffServ to classify services and control traffic at the network entrance according to service requirements, and set up the DSCP domain of the packets at the same time; in the network, each type of communication (based on the DSCP value of the packet) is distinguished in accordance with the implemented QoS mechanism, and serve it (including resource allocation, queue scheduling, packet discarding policy, etc., collectively referred to as PHB). All nodes in the DiffServ domain will comply with PHB based on the DSCP field of the packet. DiffServ can solve the scalability problem well by defining business as a limited class. At the same time, DiffServ follows the technical concept of IP itself well. Relatively speaking, it is easy to implement in existing IP networks and products. Therefore, the QoS implementation in commercial networks is generally based on the DiffServ model.
At present, some people propose to use IntServ and DiffServ in combination. In addition, some other QoS technologies have emerged, such as MPLS QoS combined with MPLS technology, Traffic Engineering (TE, Traffic Engineering), etc.
Throughput, transmission delay, delay jitter and packet error rate are commonly used QoS parameters. Different multimedia applications have different requirements for network performance. At the beginning of communication, the QoS parameters submitted by users to the network actually describe the application's needs for network resources. The network can serve as the basis for managing internal shared resources (such as bandwidth, processing capacity, cache space, etc.). If the network resources cannot meet the user's QoS requirements, or if a new call is accepted, it will violate the resources reserved for the lines in which the communication is being communicated, thereby reducing the QoS of these communications, the network will not accept the new call. This mechanism is often called Connection Admission Control (CAC).
Once the network accepts a user's call, it is responsible for ensuring the user's QoS requirements throughout the session. Therefore, the network must reserve resources for this call, monitor performance during the communication process, dynamically adjust resource allocation, and notify the relevant users until the relevant communication is terminated. The above functions constitute the QoS guarantee mechanism of the network, and currently only a few networks have implemented or partially implemented these functions. 2. QoS solution in collective communication
In the past, businesses were isolated, with different businesses, and the standards, systems, terminals, and even basic networks were used differently. Different systems could not communicate or interoperate, which eventually led to multiple isolated business environments, and each type of business could only achieve limited communication. The maturity of IP network technology and the maturity of media technology have promoted the development of collective communications. Collective communication is based on a unified network, controlled by a unified service support system, adopts common communication components, and uses intelligent multimedia terminals to realize the organic combination of various services and realize integrated communication means such as voice, images, and data.
Collective communication realizes the sharing of network resources by multiple services, greatly improving the efficiency of network applications. From another perspective, it can also be said that resource competition among various services is introduced into the IP network. How to coordinate these different services is exactly the problem that QoS needs to solve.
Services are carried by the network. Without high-quality IP basic networks, QoS technology cannot be realized, so ensuring the service quality of multiple services becomes a silhouette. According to the "IP Network Technical Requirements - Network Performance Parameters and Indicators" standard of the People's *: multimedia transmission (video service) must be carried out, and the network performance requirements must reach level 1 or above. The network performance level parameters specified in the communications industry standard of the People's * "IP Network Technical Requirements - Network Performance Parameters and Indicators" are shown in Table 1.
Since the network capacity is limited, the traffic flow that meets the above indicators is also limited. To this end, in terms of network operability, only by giving the ultimate service capacity (equivalent maximum number of concurrent users) that meets the above indicators can the user's service quality be truly guaranteed.
Only by implementing a reasonable QoS strategy on the basis of high-quality IP networks can we truly ensure the QoS of all services in the collective communications, see Figure 1 "China Multimedia Video" No. 12. 1. Network Edge
At the main network edge, the most important task is to identify and classify and mark traffic. Flow classification is often used in conjunction with acceptance control strategies, traffic monitoring, etc. The traffic classification maps service messages to a certain type of service, and the Admission Policy Control determines whether the QoS request of the service can/should be met. Traffic monitoring monitors each service flow to ensure that it does not abuse network resources.
In actual operation, whether the IP packet comes from a specific service terminal (such as a video conferencing terminal, MCU) can be simply used on the edge router to specify priority marking of its service packets. However, for a collective communication environment, service flows with completely different characteristics share the same network resource, and simple stream classification measures are difficult to meet the requirements, especially many services may be based on port 80. In this way, the basic network requires strong service perception capabilities, which can be roughly divided into the following aspects.
In-depth message analysis: Can fully parse IP messages at any level and field;
In-depth behavioral analysis: Able to analyze business connections and status; In-depth flow analysis: Able to conduct in-depth analysis of business flow content.
Through in-depth analysis of service messages, combined with business behavior, etc., the service flow can be dynamically and intelligently identified, thus laying the foundation for subsequent QoS scheduling operations. 2. QoS Policy Center
After identifying and classifying the business, the next job is to take corresponding actions based on the booking strategy. In fact, the acceptance control strategy is part of the QoS strategy. For some services, it is legal and requires priority protection, such as voice and video services, such services should be accepted, and the services should be given high priority, and even appropriate resource reservations should be made; for some services, such as BT services, if they may be rejected according to the policy, the network equipment will be directly instructed to discard their messages; for other services, such as Internet services, they will be given passes, but certain restrictions should be imposed based on the existing network resource status.
QoS policy control is not only limited to acceptance control at the edge of the network, but also guides the scheduling and processing strategies at the core level, and dynamically adjusts the strategies according to changes in network conditions, and instructs the business system to take corresponding measures. 3. Network core
The network core performs differential scheduling processing based on the QoS tags in the service message. Generally, the scheduling operations of packets by the core network are mainly divided into two categories: congestion management and congestion avoidance.
When the speed at which the message reaches the network device interface is greater than the interface's transmission capability, congestion will occur; when congestion occurs, queue scheduling technology is generally used to solve it. Each queue scheduling technology is used to solve specific problems and will have a specific impact on network performance. The queue scheduling technologies currently provided by VRP include FIFO, PQ, CQ, WFQ, RTP real-time queue, and CBWFQ/LLQ.
Congestion avoidance is used to monitor network load, foresee and avoid the occurrence of congestion, and congestion avoidance is generally achieved through packet loss technology. Generally, the core network provides a variety of congestion avoidance mechanisms to meet different applications, including tail discarding, RED, and WRED.
Congestion avoidance and congestion management mechanisms are closely linked. For each queue scheduling technology, a corresponding packet loss mechanism can be used to cooperate with it; congestion management and avoidance are PHBs that all routers must provide.
The service quality of the service not only depends on the transmission of the network itself, but also on the functions and performance that service system equipment (such as video terminals, MCUs, etc.) can provide. For example, some video terminals can automatically adjust their transmission bandwidth according to the network bandwidth situation. When it is found that the network bandwidth is insufficient, they will automatically select a encoding method with smaller bandwidth requirements. In collective communication, the basic network not only needs to have the ability to have deep service perception, but also needs to be able to form linkage with the service system. When the status of a device in the network changes and causes resource tightness, this information is reported to the Policy Center. The Policy Center determines whether these changes will affect the current business and whether these effects can be digested through the regulation and digestion of the network itself. If there is an impact that the network itself cannot handle, the corresponding service system needs to be notified in order to make corresponding adjustments, such as reducing the transmission bandwidth, applying for more cache resources, etc.
It can be seen that for collective communication, relying solely on the service system itself or the basic network cannot truly guarantee high-quality services. Only by achieving the organic integration of the basic network and the business system can network resources be effectively and rationally used to ensure the quality of various services. III. Conclusion
A true QoS solution should be a system project, involving not only IP network equipment and business systems, but also the following aspects.
Line quality: Line quality will directly affect packet loss rate and packet error rate. In multimedia services, these problems cause problems such as image mosaic, image jitter, and intermittent sounds; in addition, the two may cause retransmission of messages, further deteriorating network quality.
Network planning: For example, reasonably plan network topology, improve network dynamic regulation capabilities, shorten the number of hops of the service flow from the source to the destination, etc.; reasonably plan IP addresses, carry out routing convergence, and set up routing direct access for certain high-level services.
Purification of network applications: Some network applications may irresponsibly abuse network resources, such as some malicious websites or online games; in addition, destructive applications such as network viruses and worms have a great impact on the network and seriously affect the normal application of other businesses.
In addition, the service quality of various services in the network also depends on non-technical factors such as business operation model and regulatory strategies. Article entry: csh Editor in charge: csh
Realizing the communication of any media information at any time, anywhere and anyone is an eternal need of human beings in the field of communication. Before the maturity of IP technology, all networks were a single service network. For example, PSTN could only operate telephone services, cable TV networks could only carry television services, and X.25 network could only carry data services, etc. The separation of the network causes the separation of services and reduces the efficiency of communication.
Due to the popularity of the Internet, IP applications are becoming increasingly widespread, and IP networks have penetrated into various traditional communication ranges, making it possible to build a multi-service network based on IP. However, different services have different requirements for the network. How to implement multiple real-time and non-real-time services in a packetized IP network has become an important topic. People have proposed the concept of QoS (Quality of Service).
IP QoS refers to an ability of an IP network, that is, to provide the services they need for specific services on IP networks spanning multiple underlying network technologies (FR, ATM, Ethernet, SDH, etc.). QoS includes multiple aspects, such as bandwidth, delay, delay jitter, etc. Each service has specific requirements for QoS. Some may have higher requirements for some of the indicators, while others may have higher requirements for other indicators. The technical indicators for measuring IP QoS include the following.
(1) Available bandwidth: refers to the average rate of specific application service flows between two nodes of the network. It mainly measures the user's ability to obtain service data from the network. All real-time services have certain requirements for bandwidth. For example, for video services, when the available bandwidth is lower than the encoding rate of the video source, the image quality cannot be guaranteed.
(2) Delay: refers to the average round-trip time of data packets being transmitted between two nodes of the network. All real-time services have certain requirements for delays. For example, VoIP services generally require the network delay to be less than 200ms. When the network delay is greater than 400ms, the call will become unbearable.
(3) Packet loss rate: refers to the percentage of lost packets during network transmission, which is used to measure the network's ability to correctly forward user data. Different services have different sensitivity to packet loss. In multimedia services, packet loss is the most fundamental reason for deterioration in image quality. A small amount of packet loss may cause mosaics in the image.
(4) Delay jitter: refers to changes in delay. Some businesses, such as streaming media services, can reduce the impact of delay jitter on the business through appropriate cache; while some businesses are very sensitive to delay jitter. For example, voice services, a slight delay jitter will lead to a rapid decline in voice quality.
(5) Packet error rate: refers to the percentage of errors in the message during network transmission. The bit error rate has a particularly great impact on some encryption data services.
In addition, QoS may also include some other metrics, such as network availability. QoS indicators are actually technical descriptions of business quality. For different businesses, when QoS lacks guarantee, the business appearance presented is different.
Currently, there are two main solutions for QoS: IntServ and DiffServ.
IntServ is an end-to-end stream-based QoS technology. Using IntServ, to some extent, adopts the connection-oriented idea in a circuit-switched network. Before communication between the two ends of the service, QoS requirements are required to be submitted to the network according to the service type. The network determines whether to accept the request of the service according to a certain acceptance policy control. If the network has sufficient resources to meet the requirements of the service, it will accept the service flow and must be responsible for ensuring the resources applied for by the service. End-to-end communication path is established through out-of-band RSVP signaling. Every network device along the way needs to record the status information of each service flow - "soft status", and provide corresponding resource reservations to ensure the service quality of the service.
From a technical point of view, RSVP is an effective QoS guarantee method on the current network. However, due to the characteristics of the IP network traffic model and business model, the Internet backbone network instantly needs to provide services to thousands of service flows. Therefore, the solution idea of reserved paths for individual flows cannot be expanded on the Internet backbone network, which seriously restricts the application of IntServ in actual networks. Of course, there are other factors that restrict IntServ applications, including the large-scale deployment of RSVP signaling, interoperability between equipment of different manufacturers, and service-based management (authentication, billing), etc. In addition, IntServ requires end-to-end network support. Due to the complexity of RSVP technology, it can be said to be one of the most complex IP technologies at present. It is unrealistic to upgrade all network hardware devices to support RSVP.
At the same time, the RSVP model actually IP-based translation of the concept of circuit switching services, which largely subverts the concept of open IP interconnection and jump-by-hop forwarding. There are many unknown problems in the integration with other IP services and technologies. Overall, IntServ is currently very practical.
DiffServ is a DSCP-based QoS solution launched by the IETF organization in 1998. This is a class-based QoS technology that is mainly used in backbone networks. Use DiffServ to classify services and control traffic at the network entrance according to service requirements, and set up the DSCP domain of the packets at the same time; in the network, each type of communication (based on the DSCP value of the packet) is distinguished in accordance with the implemented QoS mechanism, and serve it (including resource allocation, queue scheduling, packet discarding policy, etc., collectively referred to as PHB). All nodes in the DiffServ domain will comply with PHB based on the DSCP field of the packet. DiffServ can solve the scalability problem well by defining business as a limited class. At the same time, DiffServ follows the technical concept of IP itself well. Relatively speaking, it is easy to implement in existing IP networks and products. Therefore, the QoS implementation in commercial networks is generally based on the DiffServ model.
At present, some people propose to use IntServ and DiffServ in combination. In addition, some other QoS technologies have emerged, such as MPLS QoS combined with MPLS technology, Traffic Engineering (TE, Traffic Engineering), etc.
Throughput, transmission delay, delay jitter and packet error rate are commonly used QoS parameters. Different multimedia applications have different requirements for network performance. At the beginning of communication, the QoS parameters submitted by users to the network actually describe the application's needs for network resources. The network can serve as the basis for managing internal shared resources (such as bandwidth, processing capacity, cache space, etc.). If the network resources cannot meet the user's QoS requirements, or if a new call is accepted, it will violate the resources reserved for the lines in which the communication is being communicated, thereby reducing the QoS of these communications, the network will not accept the new call. This mechanism is often called Connection Admission Control (CAC).
Once the network accepts a user's call, it is responsible for ensuring the user's QoS requirements throughout the session. Therefore, the network must reserve resources for this call, monitor performance during the communication process, dynamically adjust resource allocation, and notify the relevant users until the relevant communication is terminated. The above functions constitute the QoS guarantee mechanism of the network, and currently only a few networks have implemented or partially implemented these functions. 2. QoS solution in collective communication
In the past, businesses were isolated, with different businesses, and the standards, systems, terminals, and even basic networks were used differently. Different systems could not communicate or interoperate, which eventually led to multiple isolated business environments, and each type of business could only achieve limited communication. The maturity of IP network technology and the maturity of media technology have promoted the development of collective communications. Collective communication is based on a unified network, controlled by a unified service support system, adopts common communication components, and uses intelligent multimedia terminals to realize the organic combination of various services and realize integrated communication means such as voice, images, and data.
Collective communication realizes the sharing of network resources by multiple services, greatly improving the efficiency of network applications. From another perspective, it can also be said that resource competition among various services is introduced into the IP network. How to coordinate these different services is exactly the problem that QoS needs to solve.
Services are carried by the network. Without high-quality IP basic networks, QoS technology cannot be realized, so ensuring the service quality of multiple services becomes a silhouette. According to the "IP Network Technical Requirements - Network Performance Parameters and Indicators" standard of the People's *: multimedia transmission (video service) must be carried out, and the network performance requirements must reach level 1 or above. The network performance level parameters specified in the communications industry standard of the People's * "IP Network Technical Requirements - Network Performance Parameters and Indicators" are shown in Table 1.
Since the network capacity is limited, the traffic flow that meets the above indicators is also limited. To this end, in terms of network operability, only by giving the ultimate service capacity (equivalent maximum number of concurrent users) that meets the above indicators can the user's service quality be truly guaranteed.
Only by implementing a reasonable QoS strategy on the basis of high-quality IP networks can we truly ensure the QoS of all services in the collective communications, see Figure 1 "China Multimedia Video" No. 12. 1. Network Edge
At the main network edge, the most important task is to identify and classify and mark traffic. Flow classification is often used in conjunction with acceptance control strategies, traffic monitoring, etc. The traffic classification maps service messages to a certain type of service, and the Admission Policy Control determines whether the QoS request of the service can/should be met. Traffic monitoring monitors each service flow to ensure that it does not abuse network resources.
In actual operation, whether the IP packet comes from a specific service terminal (such as a video conferencing terminal, MCU) can be simply used on the edge router to specify priority marking of its service packets. However, for a collective communication environment, service flows with completely different characteristics share the same network resource, and simple stream classification measures are difficult to meet the requirements, especially many services may be based on port 80. In this way, the basic network requires strong service perception capabilities, which can be roughly divided into the following aspects.
In-depth message analysis: Can fully parse IP messages at any level and field;
In-depth behavioral analysis: Able to analyze business connections and status; In-depth flow analysis: Able to conduct in-depth analysis of business flow content.
Through in-depth analysis of service messages, combined with business behavior, etc., the service flow can be dynamically and intelligently identified, thus laying the foundation for subsequent QoS scheduling operations. 2. QoS Policy Center
After identifying and classifying the business, the next job is to take corresponding actions based on the booking strategy. In fact, the acceptance control strategy is part of the QoS strategy. For some services, it is legal and requires priority protection, such as voice and video services, such services should be accepted, and the services should be given high priority, and even appropriate resource reservations should be made; for some services, such as BT services, if they may be rejected according to the policy, the network equipment will be directly instructed to discard their messages; for other services, such as Internet services, they will be given passes, but certain restrictions should be imposed based on the existing network resource status.
QoS policy control is not only limited to acceptance control at the edge of the network, but also guides the scheduling and processing strategies at the core level, and dynamically adjusts the strategies according to changes in network conditions, and instructs the business system to take corresponding measures. 3. Network core
The network core performs differential scheduling processing based on the QoS tags in the service message. Generally, the scheduling operations of packets by the core network are mainly divided into two categories: congestion management and congestion avoidance.
When the speed at which the message reaches the network device interface is greater than the interface's transmission capability, congestion will occur; when congestion occurs, queue scheduling technology is generally used to solve it. Each queue scheduling technology is used to solve specific problems and will have a specific impact on network performance. The queue scheduling technologies currently provided by VRP include FIFO, PQ, CQ, WFQ, RTP real-time queue, and CBWFQ/LLQ.
Congestion avoidance is used to monitor network load, foresee and avoid the occurrence of congestion, and congestion avoidance is generally achieved through packet loss technology. Generally, the core network provides a variety of congestion avoidance mechanisms to meet different applications, including tail discarding, RED, and WRED.
Congestion avoidance and congestion management mechanisms are closely linked. For each queue scheduling technology, a corresponding packet loss mechanism can be used to cooperate with it; congestion management and avoidance are PHBs that all routers must provide.
The service quality of the service not only depends on the transmission of the network itself, but also on the functions and performance that service system equipment (such as video terminals, MCUs, etc.) can provide. For example, some video terminals can automatically adjust their transmission bandwidth according to the network bandwidth situation. When it is found that the network bandwidth is insufficient, they will automatically select a encoding method with smaller bandwidth requirements. In collective communication, the basic network not only needs to have the ability to have deep service perception, but also needs to be able to form linkage with the service system. When the status of a device in the network changes and causes resource tightness, this information is reported to the Policy Center. The Policy Center determines whether these changes will affect the current business and whether these effects can be digested through the regulation and digestion of the network itself. If there is an impact that the network itself cannot handle, the corresponding service system needs to be notified in order to make corresponding adjustments, such as reducing the transmission bandwidth, applying for more cache resources, etc.
It can be seen that for collective communication, relying solely on the service system itself or the basic network cannot truly guarantee high-quality services. Only by achieving the organic integration of the basic network and the business system can network resources be effectively and rationally used to ensure the quality of various services. III. Conclusion
A true QoS solution should be a system project, involving not only IP network equipment and business systems, but also the following aspects.
Line quality: Line quality will directly affect packet loss rate and packet error rate. In multimedia services, these problems cause problems such as image mosaic, image jitter, and intermittent sounds; in addition, the two may cause retransmission of messages, further deteriorating network quality.
Network planning: For example, reasonably plan network topology, improve network dynamic regulation capabilities, shorten the number of hops of the service flow from the source to the destination, etc.; reasonably plan IP addresses, carry out routing convergence, and set up routing direct access for certain high-level services.
Purification of network applications: Some network applications may irresponsibly abuse network resources, such as some malicious websites or online games; in addition, destructive applications such as network viruses and worms have a great impact on the network and seriously affect the normal application of other businesses.
In addition, the service quality of various services in the network also depends on non-technical factors such as business operation model and regulatory strategies. Article entry: csh Editor in charge: csh