CS-534: Packet Switch Architecture Spring 2003

Department of Computer Science © University of Crete, Greece

2.1   Output Contention, Internal Blocking, HOL Blocking

[Up: Table of Contents] [Prev: 1.2 TDM, Circuit Switching]

[Next: 2.2 Multiplexing]

2.1.1   Packet versus Circuit Switching

2.1.2   Output Contention

Time-Window Width versus Buffer Space:
In this transparency, the rates lamda may be "instantaneous" or average rates. Truly instantaneous rates are not very interesting, because the flows under consideration are not "smooth" in time (they are not "fluid flows"): multiple flows share a link, and they are multiplexed on a packet-by-packet basis, not on a bit-by-bit basis. When applying the above inequalities to average rates, the averaging may be over a narrow time-window or it may be over a wide time-window. The narrower the window, the more stringent the equalities become, because they require the absence of output contention in each and every short "time-slice" --not just "on the average". When output contention is absent (i.e. the second inequalities hold) over a wide window, but occurs over shorter sub-windows (i.e. the inequalities are violated there), buffering is needed to absord the excess amount of information during these short-term traffic fluctuations. The size of this buffering is proportional to the width of the (shortest) time-window over which the inequalities hold. Compare this situation to the buffering needed in serial-parallel converters (section 1.1), or to the size of voice-sample memory in the time-slot interchanges (section 1.2).

2.1.3   Internal Blocking

2.1.4   FIFO Queueing and Head-of-Line (HOL) Blocking

---

[Up: Table of Contents] [Prev: 1.2 TDM, Circuit Switching]

[Next: 2.2 Multiplexing]

---

Up to the Home Page of CS-534

© copyright University of Crete, Greece. Last updated: 16 Mar. 2003, by M. Katevenis.