Transaction multicasting scheme for resilient routing control in parallel cluster database systems

Authors
Citation
Ih. Jung et S. Moon, Transaction multicasting scheme for resilient routing control in parallel cluster database systems, J SYST ARCH, 46(8), 2000, pp. 699-719
Citations number
15
Categorie Soggetti
Computer Science & Engineering
Journal title
JOURNAL OF SYSTEMS ARCHITECTURE
ISSN journal
13837621 → ACNP
Volume
46
Issue
8
Year of publication
2000
Pages
699 - 719
Database
ISI
SICI code
1383-7621(200006)46:8<699:TMSFRR>2.0.ZU;2-A
Abstract
A disk cluster environment (DCE) refers to a distributed architecture for h igh performance transaction processing in which the computing nodes are loc ally coupled via a high-speed network and share a common database at the di sk level. In the DCE, it is crucial to determine at which node the incoming transactions are processed. This is called transaction routing. The aim of disk sharing in DCE is not only to achieve high performance by distributin g the workload among the processing nodes but also to obtain fault-toleranc e against possible system failures, like a single node failure. Although a number of transaction routing schemes have been reported for DCE, it is tru e that most of them are not sufficiently resilient against system dynamics, which inevitably requires changing the routing information. In this paper, we propose a new dynamic transaction routing scheme for DCE, called multic ast transaction routing scheme, MTR for short, that is able to change the t ransaction routing information in the presence of critical events without i mposing too much overhead to the transaction processing system. In our sche me, when it is required to change the routing information dynamically, the routing algorithm sends multiple clones of a transaction to a group of cand idate processing nodes and selects the processing node that first completes the multicasted transaction as a new processing node for re-routed transac tion. The selected processing node is expected to be a best affinity node w hen the system load is evenly distributed, or a relatively unloaded process ing node that is idle enough to process a transaction faster than other nod es. The novel aspect of MTR is that it automatically achieves an optimal ba lance between affinity-based routing and load balancing. The simulation stu dy shows that MTR rapidly stabilizes the system and produces an optimal rou ting information so that it finally guarantees faster response time. (C) 20 00 Elsevier Science B.V. All rights reserved.