You are on page 1of 18

ATM flow control

Flow control in ATM is used to optimize bandwidth utilization by having feedback mechanisms. One mechanism is GFC used to control at UNI. Other is ABR flow control that is used to utilize leftover bandwidth in the network for connections belonging to ABR service categories.

GFC-Generic Flow control


Low level flow control mechanism applicable at UNI in forward direction. Scope of GFC depends upon configuration of endsystem --Controlled connections: connections whose cell emission is controlled by GFC signals no resource is reserved for controlled connections. --Uncontrolled connections: cannot control cell emission except for HALT signal. Resource is reserved.

GFC procedures
HALT Signal: used to stop the transmission of assigned ATM layer cells for both controlled and uncontrolled ATM connection-to limit the amount of cell generated at end system. SET: issues set signal to set the credit counter to a specified integer value-specifies maximum data a controlled device can send. NULL: sends NULL signal to show that no action to be taken.

ABR flow control


The source controls its data injection rates according to loading conditions of network. RM cells carries information regarding bandwidth available, state of congestion to the source. Two flow control mechanisms:--Credit-based flow control --Rate based flow control

ABR flow control


Credit Based Flow Control: Proposed to hop by hop feedback flow control mechanism --credit specifies amount of data a source can send --every time credit decrements --and when credit is zero the sender is prohibited to send any more data.

ABR flow control


Rate based flow control: proposed to be and end-to-end feedback flow control mechanism. --destination sends control information to source which can be changed at intermediate node --source adjusts the rate according to control information.

Pros and cons


Credit based greater responsiveness to transient changes in network.in temporary overload conditions, a node has to stop sending credits to previous hop. --requires per-link-per VC monitoring Rate based schemes-simple scalable and does not require state information of the VCs to be stored in intermediate nodes. --there may be a significant delay before the control information reaches the source

ABR Flow control model


End to end mechanism Forward RM cell generated at source which is looped back by dest. And sent back to the source known a backward RM cell. A node can directly insert information in forward or backward RM cells. In case of congestion the destination can update the backward RM cell A node can directly generate a backward RM cell-no more end to end communication.

ABR Flow control model

RM Cell contents
Used to carry feedback information. No increase bit-specifies user to increase rate or not. --if 1 source cannot increase the rate --if 0 the source can increase the rate by RIF. Congestion indication bit --if 1 then rate is decreased by RDF

RM Cell contents
Single DIR bit-indicates that the RM cell is travelling from source to dest.(0) Or looped back(1). RM cells also specifies Explicit Rate at which source is meant to sent data and destination update it by its data rate and source can decrease it data rate or left unchanged. Backward notification CCR-current cell rate MCR-minimum cell rate

ATM congestion Control


Measures to be taken to prevent the problem of congestion in the network. CAC is an mechanism to prevent congestion. There are three techniques 1.Selective discard Cell 2.Frame discard 3.Explicit forward congestion indication

Selective Cell Discard


Selective Cell discard
A congested network may selectively discard cells which meet either or both the following conditions:
Cells which belong to a non-compliant ATM connection Cells which have CLP = 1

This is to protect the CLP = 0 flow as much as possible

Frame Discarding
The ATM Adaptation Layer (AAL) segments higher layer frames into small fixed-size cells for transporting over the ATM network A cell discarded by a switch causes the loss of the entire frame and eventually requires end-to-end error recovery through frame retransmission A small congestion problem could potentially escalate to a more serious one To prevent congestion escalation, Early frame Discard (EPD) and Partial Frame Discard (PPD) can be used to discard cells on a frame basis EPD and PPD are applied for ABR and UBR traffic of AAL-5 connections EPD
When congestion occurs and buffers are filling, EPD discards all cells associated with a new frame arriving at a queue

frame Discarding (continued)


The remaining buffer space can then be used for cells belonging to frames that already have entered the queue EPD maximizes the chances for already queued frames to leave the queue successfully PPD If EPD does not remove congestion and cells arriving at a queue have to be discarded because of buffer overflow PPD is applied PPD discards all subsequent cells associated with the same frame rather than just a few cells within the frame during buffer overflow PPD minimizes the number of frames becoming invalid in the queue

EFCI
Explicit Forward congestion Indication (EFCI)
A network element in an impending congested state or a congested state may set an EFCI in the cell header This indication may be examined by the destination end-system The end-system may adaptively lower the cell rate of the connection

Dynamics of TCP traffic


Delivery of inactive cells or dead cells-since a frame is fragmented into cells, a cell dropped from the frame renders remaining cells of the frame as dead or inactive. Link idle time-throughput also suffers when congested link is occasionally idle. links are idle when the TCP connections try to achieve equilibrium using slow start mechanisms.