Skip to main content

Traffic Management for High-Speed Networks (1997) / Chapter Skim
Currently Skimming:

Credit-based Flow Control
Pages 6-13

The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.


From page 6...
... Designing flow control systems for bursty traffic is obviously much more difficult than designing control systems for smooth traffic. In supporting computer communications, which are generally bursty, we will have no choice but to face the challenge of designing effective flow control for bursty traffic.
Read the entire page →
From page 7...
... Before forwarding any data cell over the link, the sender needs to receive credits for the VC from the receiver. At various times, the receiver sends credits to the sender indicating availability of buffer space for receiving data cells of the VC.
Read the entire page →
From page 8...
... Assume that the receiver uses a fair scheduling policy between VCs with Crd_Bal > 0, when forwarding cells out from its output link. Then, if there are N active VCs competing for the same output link, the maximum average bandwidth over RTT that the VC can achieve is -7 BW = Buf Alloc / (RTT+ N2*
Read the entire page →
From page 9...
... Adaptive Buffer Allocation Adaptive buffer allocation allows multiple VCs to share the same buffer pool in the receiver node adaptively, according to their needs. That is, Buf Alloc of a VC is automatically decreased if the VC does not have sufficient data to forward, cannot get sufficient scheduling slots, or is back-pressured due to downstream congestion.
Read the entire page →
From page 10...
... In referring to Figure 10, let RTTbe the maximum of all the RTTs and Mbe the size, in cells, of the common buffer pool in the receiver. For each allocation interval, which is set to be at least RTT, the receiver computes a new allocation and an N2 value for each VC according to its relative bandwidth usage.
Read the entire page →
From page 11...
... holds if those VCs that are blocked downstream do not occupy much buffer space at the current node. The adaptive buffer allocation scheme is indeed designed in such a way that inactive or slow VCs will be allocated the very minimum or a small buffer space, respectively.
Read the entire page →
From page 12...
... This is overallocation in the sense that if traffic is 1 00 percent steady state, M need only be RTT for sustaining the peak bandwidth of the output link. However, for bursty traffic, M needs to be larger than RTT to allow high link utilization and reduce transmission time.
Read the entire page →
From page 13...
... RTTwill provide "headroom" for each VC to increase its bandwidth usage under the current buffer allocation. If the VC does increase its bandwidth usage, then the adaptation scheme will notice the increased usage and will subsequently increase the buffer allocation for the VC t123.
Read the entire page →

Key Terms

  • buffer allocation
  • credit control
  • bandwidth usage
  • adaptive buffer
  • buffer pool

  • relative bandwidth
  • output link
  • output port
  • receiver sends
  • bursty traffic

  • data cells
  • static credit
  • adaptive credit
  • bandwidth allowed
  • buffer space

  • transmission time
  • credit balance
  • smooth traffic
  • downstream congestion
  • allocation scheme

  • cell loss
  • common buffer
  • credit record
  • allocated buffer
  • sends credit

  • peak bandwidth
  • fair scheduling
  • credit records
  • receiver node
  • allocate buffer


    • This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
    More information on Chapter Skim is available.