Implement traffic engineering using MPLS in large IP Networks

With the continuous and rapid development of the Internet, people began to rethink the quality, reliability and efficiency of their services. One of the key solutions is to use Multi-Protocol Label Switching MPLS) to comprehensively improve the performance of IP networks. This trend leads to the convergence of traditional routing protocols and ATM core technologies. One of the most important aspects is to allow MPLS to use the Traffic control mechanism of ATM to implement Traffic Engineering of MPLS ).
Traffic Engineering

What is traffic engineering? Traffic Engineering is actually a set of tools and methods. It can extract the best service from a given infrastructure, regardless of whether the network device or transmission line is normal or invalid. In other words, it needs to optimize installed resources. In fact, it is a supplement and Improvement Measure for network engineering or network planning. Traffic Engineering tries to make the actual network traffic exist in the physical network in an optimal way.
At present, the Internet is facing two major pressures. First, the current routing protocol is essentially connectionless, which leads to an unoptimal feature of the overall network capacity utilization. The current routing is only based on the destination IP address and shortest path, ignoring the requirements of the available link capacity of the network and the grouped stream. As shown in Case 1, super aggregation will lead to overload or congestion of some links, while other links are in the condition of insufficient utilization. Second, the current IP service essentially has no service quality, or Best Effort. However, IP technology is expected to support a wide range of businesses, including voice and video. This is powerless or powerless for the current IP technology.
So far, the provision of traffic engineering on the backbone network relies on the ATM technology. Of course, ATM is a connection-oriented exchange technology, from the first day of its planning and design, it promised to provide a traffic engineering mechanism to optimize the network. In fact, this technology has done the same. However, most of the current high-level applications are based on TCP/IP technology, and the traffic control mechanism of ATM technology indirectly controls IP services, therefore, IP technology considers adding traffic control to its own functions, or calling it a traffic engineering. Obviously, this is to provide a direct traffic control mechanism. In this context, MPLS has to adopt the idea of ATM to launch a "Peaceful technological evolution ".
MPLS Traffic Engineering
In the previous MPLS introduction, we have introduced the concept of MPLS, that is, an MPLS node is actually a device that uses tags to exchange the processing traffic. Both ATM and frame relay switches can provide the switching function here. When using ATM, the tags are directly mapped to the VPI/VCI of the ATM cells. In the IP network, the traffic engineering discusses the problem of solving the congestion and avoiding the conflict between the transitional utilization of some network resources due to the invalid ing between the business flow and resources, and the insufficient utilization of other resources; based on the network topology, the business flow is quickly, accurately, and effectively reassigned dynamically, especially in the case of network line or equipment failure.
The Traffic Engineering of advanced MPLS provides dynamic and real-time automatic resource optimization, and uses the so-called "smart connection" to achieve the IP network traffic engineering goal, that is, to solve the above two challenges.
Problems to be Solved by Traffic Engineering
Smart connection
To optimize network resources and reduce the burden on management personnel, MPLS nodes use the signaling protocol to establish an end-to-end marking switching channel LSR ). Intelligent connections can be determined by the user and can be implemented by path computing algorithms. The currently used link status database can be a OSPF-TEOSPF traffic engineering) or IS-IS-TEIS-IS traffic engineering) library, but the calculation algorithm is the same. It is important to note that the path Computing Algorithm without restrictions will undoubtedly be the minimum cost path. Without additional restrictions, the path algorithm can only significantly improve the management staff's configuration to clarify the routing workload, but it does not actually contribute to the improvement of network traffic distribution.
Bandwidth Estimation
Other parameters are required to improve the validity of the traffic allocated by the path calculation algorithm. Estimated bandwidth is one of them. In fact, each LSP requires an estimated bandwidth, and each link requires a link congestion coefficient, the coefficient is calculated based on the capacity of the link and the capacity and valuation of the occupied LSP. When a new path is added, the congestion coefficient of each related path must be re-calculated. The new connection selects a link with the lowest congestion factor.
Bandwidth Measurement
Although the estimated bandwidth has greatly improved the utilization of network resources, it is still an estimation. Obviously, the more accurate the estimation of the actual business volume, the more effective the path selection is. Because the preceding path calculation algorithm adopts a static method and does not consider the actual time-varying Link Utilization, the result may be that the business volume of a link is estimated too high or too low. Low traffic estimation may lead to overload of the actual link business volume, leading to congestion and packet loss that we want to avoid. Excessive traffic estimation may lead to insufficient utilization of the actual link, this may cause unnecessary potential congestion in other parts of the network.
Therefore, the best way to make the estimation bandwidth algorithm work is to use the measurement bandwidth. Regular measurement of link usage, usage fluctuations, buffer usage, and buffer usage fluctuations will be crucial. These new parameters can be published on a regular basis through the routing protocol or triggered when a certain time limit is exceeded. This allows you to dynamically calculate the effective bandwidth of the aggregated business volume. This method is more accurate than ordinary bandwidth estimation, which further improves the utilization of network resources.

Network Elasticity
At present, Internet services are crucial to the success of enterprises, so users need different service availability. MPLS control mechanisms should be able to provide failure recovery functions for lines or devices. The traditional method is timeout detection, which is a passive method. The new method should adopt an active method to detect and predict faults at an early stage. The underlying failure information should be directly related to the routing and signaling layers to trigger recovery measures earlier.
Connection priority
It is obvious that some connections have a higher priority than other connections. Therefore, LSR in MPLS must have this capability, including the priority of establishing connections and releasing connections. It is worth noting that priority processing under fault conditions is very important to shorten the recovery time of high-priority connections. In the case of a roundabout route, due to the dynamic nature of network faults, the optimal roundabout route can only be determined at the last moment. Therefore, the connection priority must be dynamically configured.
Network restructuring rules
In modern communication networks, failure of network lines or devices may affect hundreds or even thousands of LSPs. Unordered release, re-routing and secondary Signaling will lead to overload of the switch control system, resulting in a long network restructuring time. Therefore, there must be an elegant Restructuring Mechanism in MPLS, and then the restructuring policies of LSP groups with different priorities must be specified. The advantage of doing so is that it can quickly spread the occupied network resources after the reorganization to the relevant nodes of MPLS to ensure that the reorganization process still has the consideration of optimizing network resources.
Mark a stack
MPLS's label stack capability can significantly improve the recovery time of the network core. For example, when the business volume is aggregated in the core network, the number of LSPs in the core network can be reduced by adding another layer of tag. Once the network core fails, only a small number of LSP routes need to be reorganized, simplifying the reorganization task.
Recovery/Path Optimization
When the faulty device in the MPLS network is restored, the network should still be restored to the original optimal resource configuration status. In a connectionless network, the business volume is automatically restored to the shortest path. In a connection-oriented network, a similar mechanism must be adopted, recovery/path optimization is a frequently used method in connection-oriented networks. The LSR In the MPLS network can regularly check whether there is a better path than the existing LSP. If yes, the old LSP should be switched to the new LSP. A typical example is that after a network fault is restored, the temporary LSP is restored to the original LSP. In another case, the temporary LSP is restored to the path specified by the user.
Signaling Performance
Although MPLS is a topology-driven network, most LSPs have a long time and have relatively low requirements on signaling performance. However, as MPLS is designed as a backbone network technology, it is necessary to quickly and effectively recover faults, in fact, the quality of signaling performance has a crucial impact on the automation level of MPLS Traffic Engineering.
Network Service Quality
MPLS Traffic Engineering needs to provide service quality, specifically for different services, especially for future multimedia services. This requires that each LSR of MPLS can allocate a specific buffer and scheduling priority for the selected LSP, and reserve relevant network resources. Obviously, a dynamic multi-service network service quality optimization system is very important to the ultimate success of MPLS.
Conclusion
MPLS is an important tool for providing traffic engineering in IP networks. Although the Integrated Services (Integrated Services before MPLS) and Differentiated Services (Differentiated Services) can solve some service quality problems, only MPLS is the most comprehensive service quality assurance system. Through MPLS traffic engineering technology, traditional Best Effort IP networks can smoothly evolve into controlled multi-service networks. In addition, we must be aware that MPLS technology is actually an embodiment of ATM technology, while Marconi, a pioneer in ATM technology, has accumulated a lot of experience in traffic engineering, these experiences can be well reflected in MPLS switches.