Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Quasi-Static Window Flow Control for High Speed Data Networks

IP.com Disclosure Number: IPCOM000039540D
Original Publication Date: 1987-Jun-01
Included in the Prior Art Database: 2005-Feb-01
Document File: 3 page(s) / 42K

Publishing Venue

IBM

Related People

Chen, MS: AUTHOR [+2]

Abstract

The present publication describes a proposed scheme wherein window sizes are designed to react only to the fluctuation of the number of virtual circuits (VCs), instead of the instantaneous traffic fluctuation. Hence, only moderate complexity is required in this scheme. Basically, without prior traffic information, every VC is treated equally and has the same window size, which is determined by the process described below. In a packet-switched data network, the key resources that are shared to provide efficient and cost-effective communications are the links and the communication controllers. If the sharing is not carefully controlled, undesirable situations, such as heavy congestions at intermediate nodes, store and forward buffer deadlocks, and unfairness of resource allocation, can occur.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 3

Quasi-Static Window Flow Control for High Speed Data Networks

The present publication describes a proposed scheme wherein window sizes are designed to react only to the fluctuation of the number of virtual circuits (VCs), instead of the instantaneous traffic fluctuation. Hence, only moderate complexity is required in this scheme. Basically, without prior traffic information, every VC is treated equally and has the same window size, which is determined by the process described below. In a packet-switched data network, the key resources that are shared to provide efficient and cost-effective communications are the links and the communication controllers. If the sharing is not carefully controlled, undesirable situations, such as heavy congestions at intermediate nodes, store and forward buffer deadlocks, and unfairness of resource allocation, can occur. Flow control is the network function that manages resources to achieve the following three criteria [1]: 1) to properly utilize network resources so that a network can achieve high throughput, 2) to ensure deadlock-free data path, and 3) to be fair to all users in terms of resource allocation. The process referred to above is as follows: Let Nc be the number of VCs, B and Bc the total number of buffers and the number of buffers dedicated for each VC, respectively. If Nc*(2WL-1)<B, every VC has the window size WL, where WL is the ideal maximum window size [2]. Otherwise, buffers are allocated equally among all VCs. Or every VC gets buffers of BC = floor(B/Nc). Since (2W-1) should not be larger then Bc, we have The worst-case throughput of this quasi-static (QS) scheme can be evaluated. The worst-case scenario is that all VCs have zero traffic but one, which has infinite traffic. In this situation, the link efficiency can be estimated as the ratio of the window size to WL. Using the numerical example of a T3 link, propagation delay is 1.6 msec (about 300 miles), and packet size is 1 K bits, normalized worst-case throughput, W/WL, is plotted against the number of VCs and total buffer size in the figure. It can be seen that even the worst-case performance is reasonable: 50% and 100% for 8-Mbyte and 16-Mbyte buffers,

respectively, when Nc=400.

22% and 45% for 8-Mbyte and 16-Mbyte buffers,

respectively, when Nc=1000. It is important to remember that this is only the worst-case performance, which is a very rare event. When there are two VCs that are active, the throughput is 90%. When there are more than two VCs that are active, the throughput is 100%. Hence, optimum throughput is achieved most of the time and in very rare occasions is throughput degraded down to 90% or 45%. The QS scheme is designed to have the best balance in complexity and throughput performance for the very high speed communication environment. The QS scheme is better than static schemes because a QS scheme can achieve whatever static schemes can achieve, but also has the flexibility of extending to larger number...