Skip to content

ABean

Sections
Personal tools
You are here: Home » Arctic Beans publications » Management of Large Distributed Transactions
Log in
 

Management of Large Distributed Transactions

Document Actions
Weihai Yu. Position statement, High Performance Transaction Systems Workshop, Asilomar, USA, October 2001.

Abstract

Transaction processing suffers from two scalability problems for large transactions.

  • Concurrency control for transactions that access large amount of data and last for long period. This will cause large amount of data to be locked for long period and thus hinder concurrent access of them. Moreover, the more data each transaction accesses, the higher chances of deadlocks among transactions.
  • Management of distributed transactions that involve many server sites. The most noticeable tasks for transaction management are transaction enlistment and commitment processing. These tasks may involve extra remote invocations and eventually also disk accesses, which are the main causes to poor performance.

This position statement is about the second scalability issue.

(pdf)

« November 2007 »
Su Mo Tu We Th Fr Sa
        1 2 3
4 5 6 7 8 9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30  
 
 

Powered by Plone

This site conforms to the following standards: