A study of distributed database system architecture and its retrieval method

本研究的目的是找出分布式数据库系统访问异构分子生物学数据库所需的功能。为了实现这一目标，我们在第一年建立了一个原型数据库系统，用于Genbank、Swiss-prot、PIR和PDB。在原型系统中，这四个数据库存储在不同的网络工作站上，并且可以通过OQL访问（对象查询语言）通过面向对象的数据库模式。每个数据库都有一个用C和Perl编写的数据库代理，代理使用MPI与用户代理通信（消息传递接口）。我们还使用Agent Tcl开发了一个移动的代理。代理自主地在数据库中四处走动并收集相关的数据条目。在第二年，我们已经改变了数据库代理程序，使用Java applet和Horb。用户可以通过支持Java的Web浏览器轻松访问数据库。数据库代理将检索到的数据临时保存为HORB对象，并将最小数据发送给用户代理，例如检索到的数据的条目名称。这种体系结构有助于抑制不必要的数据传输，因为检索到的条目的全部内容可以按需发送到用户代理。通过向CPU服务器工作站发送数据，可以在CPU服务器工作站中完成对齐和聚合等耗时操作。我们有良好的系统响应时间。目前，数据库服务器中HORB对象的生命周期是在会话期间，但是，它可以更长。数据库代理本身可以通过使用同源搜索来组装根据同一基因的多个条目。安全问题还有待研究。几乎所有使用代理的系统都将面临这个问题，因为不清楚代理软件是否确实保证操作系统的安全性。

The objective of this research is to find out necessary fuctions of distributed database systems for accessing to heterogeneous molecular biology databases.To achieve this goal, we have constructed, in the first year, a proto-type database system for Genbank, Swiss-prot, PIR, and PDB.In the proto-type system, these four databases are stored in different networked workstations, and can be accessed by OQL (Object Query Language) through an object-oriented database schema.Each database has a database agent written in C and perl, and the agents commnunicate with the user agent using MPI (Message Passing Interface).We also developed a mobile agent using Agent Tcl.The agents autonomously go around the databases and collect relevant data entries.In the second year, we have changed the database agent programs to use Java applet and Horb.Users can easily access databases through web browsers that support Java.A database agent temporarily saves the retrieved data as HORB objects in itself and sends minimum data to the user agent, for example, entry names of the retrieved data.This architecture helps to suppress unnecessary data transfer, because the whole content of the retrieved entry can be sent to the user agent on demand.The time consuming operations such as alignment and aggregation can be done in CPU server workstations by sending data to them.We have good system response time.Currently, the life time period of the HORB objects in the database servers is during the session, however, it can be longer.A database agent itself can assembly multiple entries in terms of the same gene by using homology search.The security problem is yet to be studied.This problem will be faced in almost all systems using agents, because it is not clear whether the agent software surely assures the security of the operational systems.