Network resources are always limited. As long as there is a situation of robbing network resources, there will be requirements for service quality. The quality of service is relative to the network business. While ensuring the service quality of a certain business, it may damage the service quality of other businesses. For example, when the total bandwidth of the network is fixed, the more bandwidth a service occupies, the less bandwidth other services can use, which may affect the use of other services. Therefore, network managers need to plan and allocate network resources reasonably according to the characteristics of various services, so as to make efficient use of network resources.
Function:
After the packet is identified, it should be marked so that other network devices can easily identify the data. Because classification can be very complicated, it is best to do it only once. After an application is identified, its packets must be marked to ensure that the switch or router on the network can give priority to the application. By using two industry standards to mark data, namely IEEE 802. 1p or Differentiated Service Coding Point (DSCP), it can be guaranteed that multi-vendor network devices can use this service first.
When choosing a product such as a switch or router, make sure that it can recognize two labeling schemes. Although DSCP can replace IEEE802. 1p, which is the dominant labeling scheme in LAN environment, compared with IEEE 802. 1p, the implementation of DSCP has certain limitations. In a certain period of time, the compatibility with IEEE 802. 1p equipment will be very important. As a transition mechanism, you should choose a switch that can switch from one scheme to another.
Generally, QoS provides the following three service models:
L best effort service (best effort service mode)
L integrated services (integrated services)
Differentiated service
1. Best effort service model is a single service model and the simplest one. For the best-effort service model, the network sends as many messages as possible. However, it does not provide any performance guarantees, such as time delay and reliability.
The best-effort service model is the default service model of the network, which is implemented by FIFO (first-in, first-out) queue. It is suitable for most network applications, such as FTP and email.
2.Int-serv service model Int-Serv is a comprehensive service model, which can meet various QoS requirements. This model uses Resource Reservation Protocol (RSVP), which runs on every device from source to destination, and can monitor each stream to prevent it from consuming too many resources. The system can clearly distinguish and ensure the quality of service of each service flow, and provide the most fine-grained service quality distinction for the network.
However, Inter-Serv mode requires a large number of devices, and when the number of data streams in the network is large, the storage and processing capacity of devices will encounter great pressure. The inter-server model has poor scalability and is difficult to implement in the Internet core network.
3.Diff-serv service model Diff-Serv is a multi-service model, which can meet different QoS requirements. Unlike Int-Serv, it does not need to inform the network to reserve resources for each service. Differentiated service is simple to realize and has good expansibility.
Supplement:
Today's routers generally support QoS.
QoS on the router can be obtained in the following ways:
1. Obtained through large bandwidth.
Don't do any extra work on the router except increasing the interface bandwidth to ensure QoS. Because there is no corresponding recognized mathematical model to ensure data communication, this method can only be roughly estimated by empirical values. It is generally believed that when the bandwidth utilization rate reaches 50%, it should be expanded to ensure that the interface bandwidth utilization rate is less than 50%.
2. Through end-to-end bandwidth reservation.
This method reserves end-to-end bandwidth between nodes communicating in the whole network by using RSVP or similar protocols. This method can guarantee QoS, but the cost is too high, and it usually only runs on the enterprise network or private network, which can not be realized on the large public network.
3. Obtained through access control, congestion control and differentiated services.
This method can't guarantee QoS completely. This can be combined with increasing the interface bandwidth to provide relative QoS to some extent.
4. Obtained through MPLS traffic engineering.