Miniquiz, CIS 451, Fri 10/31/03. 10 minutes only. 1. What are the main purposes of Sliding Window Protocols? A. Flow Control: Sliding Window protocols ensure that when two computers of unequal capability communicate, the more powerful one (or the one with less other work to do) can not overload the other one. Not asked: The destination can slow down the source by not sending an acknowledgement until it has space in its receive buffer for more data. B. Error Control: Sliding Window protocols make sure that missing data and data damaged during transmission is re-transmitted. Sliding Window Protocols also ensure that data arriving more than once, and data arriving out of order, are handled correctly. 2. We have Voice over ATM. Every ATM cell contains 48 data bytes. The voice is PCM encoded, with 8000 samples / sec and 1 byte (8 bits) per sample. No fancy coding or data compression. What is the ``cellification'' delay? Each cell contains 48 Bytes, e.g. 48 samples. At 8000 samples/sec that represents 48/8000 = 6/1000 of a second on voice signal. The ``cellification delay'' is 6 msec. One student said that the ``serialization delay'' is 6 msec. That is incorrect. Suppose the ATM cells are shipped over a link with datarate 45 Mbit/sec. At the source (voice encoder) it takes 6 msec to build the ATM cell. Once it is ready, it is shipped out in (8*53)/(45,000,000) = .000,009,422 sec, a little bit less than 10 mu-sec. That is the serialization delay. (And then there is propagation delay, depends on how far the cell must travel). The serialization delay depends on the size of the packet or cell and the data rate of the transmission link. (The link carrying the packet or cell). The packetization (cellification delay) of a stream of real-time data depends on the intensity of the data stream being carried and the size of the packets or cells. A few students said the ``cellification delay'' is 8000/48 = 166.27 seconds. If you are paid $8000 per year, and you get $48 per payday, is there 48/8000 year between paydays? Or 8000/48 ? A correct argument would have been: there are 8000/48 = 166.67 cells per second, so the time between cells is 1/166.67 = .006 sec = 6 msec. Therefore, it takes 6 msec to collect enough voice samples to fill a cell. --- Aside: Under the USA proposal (64 data bytes per cell) the cellification delay would have been 8 msec. Under the European proposal (32 data bytes per cell) the cellification delay would have been 4 msec. Not discussed in class: This extra delay might have caused a need for many more echo-cancellers in the European telephony system, had cells of 64 data bytes been accepted. The USA already had those echo cancellers installed. (Not to be asked on any test!) (At least not in THIS class.)