MIME-Version: 1.0
Server: CERN/3.0
Date: Sunday, 01-Dec-96 19:31:08 GMT
Content-Type: text/html
Content-Length: 2297
Last-Modified: Thursday, 31-Oct-96 05:28:42 GMT
Transaction Server - Bart
Transaction Server
Libby Lin, Sunil Srivastava, Eric Vitrano
Write-Up: 1
Goal:
The goal of our project is to create a fault tolerant Transaction Server
for multiple clients using TCPIP as the basis for communication.
Interfaces:
We will be providing service to the ECash Application, who will be using our function
stubs, and we will be using the services of Stable Storage, Encryption and possibly
Private Email.
Schedule:
The schedule with appropriate methods for achieving the desired goals is as follows :
Week
|
Goals
|
Methods
|
1.
|
Single Client and Single Server
Fault Tolerance.
|
2 or 3 Phase Commit.
TCPIP with encryption layered on top of it.
|
2.
|
Multiple Client and Single Server
|
Forking on the Server side.
|
3.
|
Multiple Client and Multiple Server
|
Randomized load balancing on the client's side.
i.e. Clients will randomly choose a Transaction Server.
|
4.
|
Testing/Debugging. *
|
|
5.
|
Preparation for Presentation. *
|
|
* If time permits, we may implement a Private Email communication with
ECash.
Design Decisions:
Communication Protocol - TCPIP was chosen to be the communication protocol between our
function stubs and transaction servers because SSL is not fully standardized, and may have legal problems
with RSA. Privte Email is our secondary choice because of excessive overhead, but may be implemented
later to provide a higher degree of service.
References:
Dale Skeen and Michael Stonebraker. A Formal Model Crash Recovery in a Distributed System. IEEE Transaction of Software Engineering, May 1983.