http:^^www.cs.sunysb.edu^~gerstl^group^hptp.html

This data set contains WWW-pages collected from computer science departments of various universities

Date: Tue, 26 Nov 1996 00:40:57 GMT
Server: NCSA/1.5
Content-type: text/html
Last-modified: Mon, 29 Apr 1996 23:38:25 GMT
Content-length: 2746

<TITLE>High Performance Transaction Processing Group</TITLE>
<!WA0><img src="http://www.cs.sunysb.edu/~gerstl/group/name.xbm" ALT="High Performance Transaction Processing Group">
<h4>State University of New York at Stony Brook</h4>
   <!WA1><a href="http://www.cs.sunysb.edu"> Department of Computer Science</a> </h4>

<hr>

<P>Traditional algorithms for controlling concurrent access to shared
data, by requiring serializable execution, may unduly restrict
concurrency, and hence performance, in situations where data is
accessed heavily or distributed, or where transactions run for long
periods of time.  We are currently investigating the extent to which
the semantics of transactions (expressed as proofs in a formal system)
can be exploited to improve performance.  We use semantics in several
ways: to define a new correctness criterion for concurrent
(non-serializable) transaction execution, to decompose transactions
into smaller units so that locks can be released early, and to design a
new concurrency control that guarantees correct execution when these
units are interleaved.  We are taking two approaches to transaction
decomposition.  In the first we decompose a transaction into a sequence
of steps.  Steps are atomic and isolated and release all conventional
locks when they complete.  A new concurrency control and lock mode is
required to implement this approach.  The second approach guarantees
correctness using only conventional locks, but in a non-two-phase
fashion.

<P>The algorithms are being implemented in the context of a commercial
database system and a test bed has been constructed so that a benchmark 
transaction load can be applied in order to evaluate the ideas.

<P>We are also interested in the use of transaction semantics to better 
understand problems associated with federated databases and compensation.


<hr>
<h2>Current Projects</h2>
<!WA2><a href="http://www.cs.sunysb.edu/~gerstl/group/acc.html">Assertional Concurrency Control</a><br>
<!WA3><a href="http://www.cs.sunysb.edu/~gerstl/group/modular.html">Modular Analysis of Transactions</a>
<hr>
<h2>Members</h2>
<h4>Faculty</h4>
<!WA4><a href=http://www.cs.sunysb.edu/csdept/Faculty/art/art.html>Art Bernstein</a><br>

<!WA5><a href=http://www.cs.sunysb.edu/~pml>Phil Lewis</a>

<h4>Graduate Students</h4>
<!WA6><a href=http://www.cs.sunysb.edu/~wleung>Wai-Hong Leung</a><br>

<!WA7><a href=http://www.cs.sunysb.edu/~gerstl>David S. Gerstl</a><br>

<hr>
<h2>Recent Papers</h2> 
<LI> High Performance Transaction Systems Using Transaction Semantics,
A.J. Bernstein and P.M. Lewis, <I>
Distributed and Parallel Databases, </I> Volume 4, Number 1,
January 1996




<br>
<br>
<!WA8><a href="http://www.cs.sunysb.edu">Return to the department home page</a>
<hr>
If you have problems with this page, please send E-mail to:
<address><!WA9><a href="mailto:[e-mail]">[e-mail]</a></address>
</BODY>
</html>