Please use this identifier to cite or link to this item: http://hdl.handle.net/10397/25606
Title: SDITPM : a novel transfer policy model to facilitate object mobility that shortens service roundtrip time by load balancing over the internet
Authors: Lo, JCT
Wong, AKY
Lin, WWK
Keywords: Load balancing
M3RT
Object mobility
Planar/horizontal integration
Primary metric
SDITPM
Secondary metric
Transfer policy
Transfer probability
Vertical/incidental integration
{0,Δ}2 objective function
Issue Date: 2007
Publisher: Market Harboroug
Source: Computer systems science and engineering, 2007, v. 22, no. 1-2, p. 57-69 How to cite?
Journal: Computer Systems Science and Engineering 
Abstract: The objective of the novel statistical distribution independent transfer policy model (SDITPM) is to reduce the service roundtrip time (RTT) in object-based computing effectively. As a result it shortens program execution and helps time-critical applications meet predefined deadlines effectively. The SDITPM achieves its objective through load balancing, which is a natural outcome of guided object/agent mobility. Applications running on the Internet are object-based, and the cognate objects interact in a client/server relationship over logical TCP (Transmission Control Protocol) channels. Service RTT is the interval between the point when a client has made a service request and the moment when it has received the result correctly. If the host of a logical server is congested, the average service RTT would be prolonged by delays of various origins, including queuing, context switching, and retransmissions. If the server can migrate from its congested host to a less busy node, the service RTT is automatically reduced due to the effect of load balancing, provided that the object mobility is properly guided. The SDITPM guides the object migration process by leveraging primary metrics (e.g. server's queue length). From every primary metric the corresponding secondary ones are derived for the sake of proportional (P), derivative (D), and integral (I) controls. The statistical leveraging operation is the responsibility of the M3RT module in the SDITPM framework, which treats every primary metric as a waveform. The secondary metrics to be derived are: (a) the "current estimated mean over the last estimated value" ratio for proportional (P) control, (b) the "current estimated rate of change" for derivative (D) control, and (c) the P and D deviation errors (DE) for integral (I) control. In every transfer policy decision making cycle the SDITPM combines the P, I, and D controls selectively to compute the overall transfer probability (TP0) that decides if object migration should occur. An affirmative transfer policy decision by SDITPM is the first guidance to a successful object migration, which must be substantiated later by the location policy that provides the necessary second guidance of pinpointing the suitable host destination. The motivation of the SDITPM is to effect sound transfer policy decisions.
URI: http://hdl.handle.net/10397/25606
Appears in Collections:Journal/Magazine Article

Access
View full-text via PolyU eLinks SFX Query
Show full item record

Page view(s)

35
Last Week
4
Last month
Checked on Aug 20, 2017

Google ScholarTM

Check



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.