Multi-layer Traffic Engineering

“Internet traffic engineering is defined as that aspect of Internet network engineering deal-ing with the issue of performance evaluation and performance optimization of operational IP networks” [24]. According to this definition, traffic engineering are techniques that help the network operator to obtain a better network utilization with subsequent performance improvement. Usually the target of traffic engineering is to avoid congestion by control-ling how the traffic is routed. Figure 2.8 illustrates a schema with the main blocks in a Multi-layer Traffic Engineering (MTE) problem. These include five blocks:

• Traffic demands: These are the information that customers need the network trans-port. There are many features that determine a given traffic demand. A fixed matrix with end-to-end peak rates can be the input to the problem, but more complicated processes may be defined, for instance, with random traffic changes.

• Network Equipment: This is all kind of physical devices used in the process of transporting the information, including routers, ROADMs and the cables or fibers that connect them and their software. Operators invest on infrastructure and MTE tools help to this network capacity task, but once the equipment is deployed on the network, MTE algorithms must deal with the resources available only.

• Objective: This is a crucial aspect of every engineering problem. It is very important to properly define what aspects the MTE problem must either to solve or improve.

Congestion is usually the target of many TE problems but, in multi-layer networks, it is mandatory to define the objective in both layers.

• MTE module: this module is a software that is able to get the operator require-ments, traffic information and network status, and provide an optimum or feasible solution based on this information.

• Network configuration: this is the solution to the previous problem and it is signaled to the network equipment, so it can apply the proper changes.

Traffic Demands

Network Equipment MTE

Module

Objective Network

Configuration

Figure 2.8: Multi-layer Traffic Engineering Problems

The previous definition is also valid for Single-layer TE (STE). Multi-layer TE adds more flexibility to the network, at the expense of increasing the problem’s complexity.

Multi-layer capable nodes can establish end-to-end connections (light-paths) between any two nodes in the network on-demand. Thus, traffic can be groomed and sent over a direct light-path, either by-passing the intermediate nodes or following the hop-by-hop connections. Such traffic offloading is depicted in Figure 2.9.

...... _ROADM IP Router

DWDM Transponders

...... _ROADM IP Router

DWDM Transponders

Router Traffic

cut -through Traffic

Figure 2.9: Cut-through and hop-by-hop traffic

An important remark concerning traffic offloading in the optical domain is that it requires a reconfigurable optical technology. In the past, the first optical networks were static, such that network managers had to go to the location where the optical equipment was connected and then change the connection. This solution is not suitable for multi-layer networking. At present optical devices can remotely and automatically change its configuration and a control plane is defined, thus the technology has became mature enough to establish by-pass connections that do traffic offloading at the optical layer, as Figure 2.10

illustrates.

IP Transit

Optical Transit

Static DWDM network IP network

Dynamic Optical network IP network

ROADM ROADM ROADM

Figure 2.10: Traffic offloading over the optical transport network

While in pure IP networks, all TE mechanisms are exclusively done at the IP layer, in IP over WDM architectures these can be carried out either at the IP/MPLS or optical layer.

According to the previous definition of the MTE problem, there are many types of MTE problems. Every block has its own restrictions or features that makes a given MTE problem different. A taxonomy of Traffic Engineering Systems is done in [24], let us summarize the most important ideas:

• Time-dependent vs State-dependent vs Event-dependent. Time-dependent algorithms use historic information of the system in order to apply their policies.

State-dependent based algorithms use the state of the network to make the decisions.

Usually, this state information can not be obtained with the historical data. Event-dependent algorithms use learning models to find an optimal solution.

• Offline vs Online. Offline algorithms are used when there are no real-time re-quirements, while online algorithms are triggered periodically or when an anomaly occurs.

• Centralized vs Distributed. Centralized solutions have all the information in a unique entity, which usually makes it easy to find the optimum solution. However, these do not scale and lack robustness.

• Local vs Global Information. A TE is local when it does not require information from the whole network but a portion. Global algorithms needs information from the network, which can be a problem because of the delay in monitoring and measurement systems.

• Descriptive vs Prescriptive. Descriptive solutions just evaluate the possible op-tions to carry out, while prescriptive algorithms recommend an option among them.

• Closed Loop vs Open Loop. Closed Loop algorithms are those which not only make a decision but also use feed-back information to improve such a decision. Open loop algorithms do not use this feed-back information.

• Tactical vs Strategic. Tactical algorithms try to solve a certain problem, while Strategic algorithms make a systematic evaluation of the system to keep it in a proper state in the medium or long term.

Whenever there is a possibility of improving the network performance in a multi-layer network, there is a MTE problem. To solve such a problem, the MTE algorithm requires information concerning traffic demands and available network equipment and resources.

Once the problem is clear the MTE module must be designed paying attention to the alternatives shown in the previous classification. This will provide a new network config-uration if there is a feasible solution or a recommendation to upgrade the infrastructure and satisfy the objectives.