Cell loss priority

From Hill2dot0
(Redirected from CLP)
Jump to: navigation, search

The cell loss priority (CLP) bit allows the user or network to indicate the desirability of discarding specific cells, assuming that the network is operating in the CLP-significant mode. This is a particularly important feature because an ATM network discards cells if the network is congested or if a user exceeds the negotiated bandwidth. By indicating the cell loss priority, a user can at least indicate a preference as to which cells should be discarded and which should not (although the network is not obligated to comply with the indicated preference!). Furthermore, if a single cell is discarded, the entire CS-PDU is lost, possibly along with the entire service PDU. If used judiciously, the CLP can minimize higher layer PDU retransmissions by trying to force discarded cells (for the most part) to come from the same higher layer PDU.

The CLP bit can be set to 1 by the user or network to indicate lower priority cells, which might be discarded by the network under certain load conditions. Use of the CLP bit could limit the adverse effect of losing cells due to a buffer overflow condition within the network. One application for the CLP mechanism could be for variable bit rate services where high-priority cells (CLP=0) carry the synchronization information necessary to reconstruct the signal. Losing one of these cells is more damaging than losing a single data cell.

At its option, the network can set the CLP bit in a cell, no matter how the originating equipment has set it. If a user sends cells at a rate just over its negotiated bandwidth, the network can set CLP=1 to tell downstream switches that this cell is a candidate for discard in case of congestion. Because the CLP bit is normally set to 0 by the origination equipment, the equipment receiving a cell with the CLP bit set can infer that this end-to-end channel is nearing or exceeding its sustainable CLP0 bandwidth capacity and that the overall network traffic level is low enough to provide excess bandwidth to some users (e.g., a cell that could have been discarded during congestion was delivered).

While the QoS might differ for the CLP=0 versus CLP=1 cell flows, both cell types are considered compliant with the connection’s traffic contract if they pass all generic cell rate algorithms (GCRA), which test that cell’s CLP state. In other words, a CLP=1 cell simply complies with a different aspect of the connection’s contract.

The issue of the network discarding traffic is of legitimate concern to users. The discarded cell might represent only a tiny fraction of the entire upper layer protocol data unit, given the small size of the cell payload. However, discarding a single cell would cause the reconstructed PDU to have many bit errors at the receiving end. This, in turn, would cause the sender to resend many cell payloads to reconstitute the upper layer PDU at the receiver.

If the goal of discarding traffic is preventing network congestion, this is not a good strategy. Alternatively, the ITU-T allows ATM network service providers to offer assured operation or assured delivery of cell payloads. In this case, even if a cell payload were discarded, the ATM network itself—not the sending user—would be responsible for resending the cell payload. A similar approach was followed with X.25 and required elaborate tracking and buffering protocols. The same would be needed to implement assured delivery in ATM, so the ATM Forum and almost all ATM networks today limit cell conveyance to nonassured delivery, which means that end-users are responsible for detecting and recovering from missing cell payloads.