Network congestion avoidance
Network congestion avoidance ( " avoiding network congestion " ) is to prevent traffic jams, a process in telecommunication networks. Causal problem, the limitations of all resources, especially the available channel capacity and the available processing power of the individual routers. Effects of stress on available capacity addition, an increase in the propagation delay, more or less strong fluctuations of the term ( engl. jitter) and the loss of data packets.
The deliberate overloading of one or more network components is referred to as (Denial of Service) attack, and serves to make individual hosts or parts of a network unusable.
Method for avoiding overload situations put on at various points; each has specific advantages and disadvantages.
- 2.1 Routing Prokoll
- 2.2 Bufferbloat avoidance
- 2.3 Active Queue Management ( AQM )
End -to-end policies
Traffic is generated at a source host and finds its path through the network through hops to its destination host. Both the target host and each hop can be the source host to inform the sending rate to slow. Connection-oriented protocols such as TCP monitor in general the number of erroneous or lost data packets and delays in order to adjust the transmission speed adequately.
TCP / IP-based avoiding overload
This effect occurs when the number of parallel TCP connections are deleted in a router queue, because the buffer does not support classes of service, but simply deletes the last element of the queue. Then throttle many (all) TCP connections almost simultaneously its transmit power so that the jam is cleared. However, the compounds increase after some time its transmission rate so that there may be an overload situation again. This oscillation ( wave-like on and off ) of all TCP connections TCP Global Synchronization is called and used as a synonym for queue tail-drop, so deleting packages at the end of the queue without consideration of classes of service.
- See Explicit Congestion Notification (ECN )
- See TCP Global Synchronization
Hop - based strategies
Each network element is not transparent, is called a hop. The router is a network element; on each router runs an operating system which controls the hardware and has to be included according to options in the process of network congestion avoidance. On one hand, by the choice of the routing protocol, on the other hand by AQM:
Routing Prokoll
According to the used routing protocol, a router decides which of the available and known him a router every packet is forwarded.
Bufferbloat avoidance
Buffers are necessary and useful, they are used to collect load peaks. However, there are a plurality of buffers in each router, and its size is increased over time. In the event of a jam to fill all buffers and it comes to the designated undesirable effects, it is necessary the size of each buffer sent possible to adapt to the specific needs.
Active Queue Management ( AQM )
The network scheduler manages the located in the send queue buffer data packets. For small data buffers lead to packet loss during peak loads to large buffers lead to increased propagation delay when they fill up. The network scheduler can according to the algorithm used both targeted discard packets in the buffer ( = delete ), as also change the order of units in buffer packets. Usual, for example, the prioritization of data packets belonging to a real-time connection, e.g. IP telephony data packets or packets that belong to a SSH connection.