Network Quality of Service (QoS) Technology

1. Introduction

Branches of the national financial industry connect to the headquarters data center by renting WAN dedicated lines, carrying a wide variety of services such as online transaction production systems, smart tellers, customer service voice, collaborative office, instant messaging, file transfer, etc., and different services have different requirements for network performance. The default network forwarding rule is first-in-first-out and best-effort. If all services are treated equally without distinction, it is easy to cause network congestion during business peak hours, and important production services cannot obtain end-to-end network performance guarantees, which has a negative impact on user experience. Therefore, it is very necessary to deploy QoS technology for differentiated services in the WAN.

QoS is the abbreviation of Quality of Service. Its purpose is to provide end-to-end service quality assurance for various business requirements.

2. Basic Concepts of QoS

QoS uses the following parameters (network characteristics) as metrics to provide service quality assurance for key businesses so that they can obtain predictable service levels.

• Bandwidth (throughput)
• Latency
• Jitter (Delay Variation)
• Packet loss rate

2.1 Bandwidth

Bandwidth, also known as throughput, refers to the maximum number of data bits transmitted from one end of the network to the other in a unit of time (1s). It can also be understood as the average rate of a specific data flow between two nodes in the network. The unit of bandwidth is bit/second (bit/s, abbreviated as bps).

Bandwidth can be compared to a city’s water supply network to help understand its meaning: the diameter of the water supply pipe is like the bandwidth, and water is like the data transmitted by the network. The larger the diameter of the pipe, the greater its water supply and drainage capacity, that is, the greater the bandwidth, the greater its ability to transmit data.

Figure 1

With the rapid development of the financial industry, business traffic has also surged, and the requirements for network transmission capacity have become increasingly higher. The best way to adapt to business development is of course to increase the bandwidth of network lines. However, due to the high cost of long-distance WAN bandwidth across regions, it is impossible to expand network bandwidth indefinitely from the perspective of operation and maintenance costs. Therefore, this requires network maintenance personnel to make full use of limited WAN bandwidth line resources to ensure that all business traffic can be forwarded normally and efficiently on the network.

2.2 Latency

Latency, also known as delay, refers to the time it takes for a message or packet to be sent from one end of the network to the other. Taking voice transmission as an example, latency refers to the time from when the speaker starts speaking to when the other party hears what is said. If the latency is too large, the sound will be incoherent or fragmented.

Usually, each node on the network will cause delays. Some delays are very small, so we can ignore them in the actual environment; some delays are relatively large and need to be paid attention to. There are mainly the following types of delays in the network:

Forwarding/processing delay: refers to the processing time from when a packet is fully received by the router (or switch) to when it is placed in the output queue. The internal processing power of different models of equipment varies, but in the calculation of the overall delay budget, forwarding/processing delay is usually a small enough part to be ignored.

Queue delay: Queue delay refers to the time that occurs inside the router device, when the data packet is waiting for the device to send other data packets in the queue first. Taking the example of nucleic acid testing, in fact, the nucleic acid throat swab test time is relatively short, but if there are too many people being tested, there will be a queue. Therefore, queue delay can be understood as the long waiting time in the queue due to the large number of people when doing nucleic acid testing. Then the waiting time from the start of queuing to the start of nucleic acid testing can be understood as queue delay. From the perspective of the router, usually the queue delay is only the output queue of the router, because the input queue in the router is usually negligible. However, the queuing time can also reach hundreds of milliseconds or longer.

Serialization delay: This is the time it takes for a device's physical interface to encode the bits in a packet. If the link speed is faster, the bits are encoded onto the link faster; if the link speed is slower, it takes more time to encode the bits onto the link. Likewise, it takes less time to encode the bits in a smaller packet onto the link than it does for a larger packet. The serialization delay of a packet can be calculated using the formula: bits sent / link speed.

Transmission delay: refers to the time it takes for a data packet to reach the other end of a cable from one end. The propagation speed of electrical and optical interfaces is close to the speed of light, and the only factor that affects transmission delay is the link length. Therefore, in practice, the transmission delay of common LAN links can be ignored, while the transmission delay of WANs that reach hundreds or even thousands of kilometers becomes an important part of the overall delay.

Figure 2

2.3 Jitter

Jitter, or delay variation, refers to the different delays experienced by different packets in the same service flow. In packet networks, delays come in many different forms, so jitter always exists, but the question is whether the jitter is severe enough to degrade the service quality of the application.

Generally speaking, data applications will experience some jitter, but it will not reduce the service quality. However, some types of services, especially real-time services such as audio and video, have a very low tolerance for jitter, and the difference in packet arrival time will cause discontinuity in voice.

2.4 Packet Loss Rate

Packet loss refers to the loss of data packets. There are many factors that cause packet loss, such as device hardware failure, line failure, network congestion, frame check sequence (FCS) failure, etc.

A small amount of packet loss has little impact on the service. For example, in voice transmission, if a bit or packet of information is lost, the two parties may not notice it or it may not affect the understanding of the meaning. Using the Transmission Control Protocol (TCP) to transmit data can also handle a small amount of packet loss because TCP allows lost information to be retransmitted. However, a large amount of packet loss will affect transmission efficiency. Therefore, QoS needs to pay attention to the statistical data of packet loss - packet loss rate.

The packet loss rate refers to the percentage of lost packets to transmitted packets during network transmission.

3.QoS application deployment

Since different end-user (service) traffic has different requirements for network performance, and the financial industry has its own level of concern for different services, the following QoS deployment can be performed in the WAN with reference to the above two aspects:

First, all business traffic is classified into customer service voice (audio), key production (critical), general production (product), video (video), office (office), other business (other), etc.

Then, each type of business traffic is processed accordingly with QoS. ​

3.1 Customer Service Voice Service

Since voice services have a low tolerance for latency and jitter, that is, they require consistent low latency; and voice customer service is at the highest level in the importance ranking sequence of financial services, the queue scheduling priority of customer service voice service traffic (audio) is set to the highest priority. This ensures that customer service voice traffic is always forwarded first, minimizes the impact of bandwidth, and keeps latency and jitter at the lowest level within a controllable range.

Although the customer service voice service is very important and its queue scheduling priority is set to the highest priority, if the traffic of this service is very large under special circumstances and occupies all bandwidth resources, other services will be collectively interrupted due to lack of bandwidth resources, which is obviously unreasonable. Therefore, it is necessary to set a certain speed limit for the customer service voice service traffic.

Therefore, the following QoS processing can be performed on the customer service voice service: PQ (Priority Queue) is used for queue scheduling, and the maximum bandwidth speed is limited according to the normal required rate.

3.2 Key production and general production operations

Online transaction production is a critical production business, which has high requirements for bandwidth and packet loss rate. Therefore, the critical production business traffic (critical) is given priority QoS bandwidth guarantee to ensure that the critical production business traffic always has enough bandwidth for forwarding, so as not to cause a large number of packet loss due to network congestion. In actual deployment, we can also subdivide the critical production business into several specific important businesses.

General production business is second only to key production business in importance, so it also requires certain QoS bandwidth guarantees, and the guaranteed bandwidth ratio is slightly lower than that of key production.

3.3 Video Services

Video services mainly refer to video conferencing services. Video conferencing requires consistent low latency and high throughput, so certain QoS bandwidth guarantees are also required for video service traffic (video). At the same time, with the help of routing selection strategies, the main WAN dedicated line for video service traffic is set to the third line (the backup line for the production main line and the office main line), which is separated from the production and office service traffic, and can fully utilize the bandwidth resources of multiple WAN dedicated lines.

3.4 Office Business

There are many types of office services, and a certain QoS bandwidth guarantee is allocated overall. At the same time, with the help of routing selection strategy, the main WAN dedicated line of office business traffic (office) is set to the second line (the backup line of the production main line and the video main line), which is separated from the production and video business traffic, and can fully utilize the bandwidth resources of multiple WAN dedicated lines.

3.5 Other Business

Other services refer to the remaining service traffic (other) that is not clearly classified. The service importance level is relatively low. To ensure that all types of service traffic have the opportunity to be forwarded, a relatively low proportion of QoS bandwidth is still guaranteed for other service traffic (other).

3.6 QoS Deployment Summary

To summarize, we can deploy the following QoS applications for financial services.

Table 1

​

Figure 3

4. Summary

As the business develops, QoS policies also need to be dynamically optimized. Traditional networks are based on distributed protocol implementation, and the configuration process adopts the form of hop-by-hop configuration, which inevitably leads to complex QoS policy deployment and optimization and adjustment operations, and is prone to errors. The wide area network based on SDN technology can make QoS policy deployment more flexible and agile with the characteristics of SDN automated network configuration; it can collect more accurate global network traffic status information, such as transmission delay, bandwidth, and packet loss rate, to make QoS network traffic scheduling more accurate; it can centrally control all network devices and network traffic from a global network view to ensure the consistency of QoS policies across the network. In short, in the wide area network of the financial industry, by configuring QoS policies, network traffic is regulated, network congestion is avoided and managed, the message loss rate is reduced, and differential services are provided for different services (voice, video, data, etc.), ensuring the stable operation of important production services.

The network is not born this way. With the development of technology and changes in application scenarios, the network architecture is also evolving iteratively, and ultimately forms a reliable and robust foundation for the digital economy network infrastructure. In this, we analyze various network scenarios to show you how the network is. ​