MEET DB2: Automated Database Migration EvaluationReynold S. Xin∗EECS, UC [email protected] DantressangleIBM [email protected] LightstoneIBM [email protected] McLarenIBM [email protected] SchormannIBM [email protected] SchwengerIBM [email protected] databases compete for market share, which iscomposed of not only net-new sales to those purchasing adatabase for the first time, but also competitive “win-backs”and migrations. Database migration, or the act of movingboth application code and its underlying database platformfrom one database to another, presents a serious adminis-trative and application development challenge fraught withlarge manual costs. Migration is typically a high cost ef-fort due to incompatibilities between database platforms.Incompatibilities are caused most often by product specificextensions to language support, procedural logic, DDL, andadministrative interfaces. The migration evaluation is thefirst step in any competitive database migration process.Historically this has been a manual process, with the highcosts and subjective results. This has led us to reexaminetraditional practices and explore an automatic, innovativesolution.We have designed and implemented the Migration Eval-uation and Enablement Tool for DB2 for Linux Unix andWindows, or MEET DB2, a tool for automatically evalu-ating database migration projects. Encapsulated in a sim-ple one-click interface, MEET DB2 is able to provide de-tailed evaluation of migration complexity based on its deepanalysis on the source database. In this paper, we presentMEET DB2, and discuss many aspects of our design, and re-port measurements from real-world use cases. In particular,we show a novel way to use XML and XQuery in this do-main for better extensibility and interoperability. We haveevaluated MEET DB2 on 18 source code samples, cover-ing nearly 1 million lines of code. The utility has providedbenefits in several dimensions including: dramatically re-duced time for evaluation, consistency, improved accuracyover human analysis, improved reporting, reduced skill re-quirements for migration analysis, and clear analytics forproduct planning.∗Work done while at IBM Canada.Permission to make digital or hard copies of all or part of this work forpersonal or classroom use is granted without fee provided that copies arenot made or distributed for profit or commercial advantage and that copiesbear this notice and the full citation on the first page. To copy otherwise, torepublish, to post on servers or to redistribute to lists, requires prior specificpermission and/or a fee. Articles from this volume were presented at The36th International Conference on Very Large Data Bases, September 13-17,2010, Singapore.Proceedings of the VLDB Endowment, Vol. 3, No. 2Copyright 2010 VLDB Endowment 2150-8097/10/09... $ 10.00.1. INTRODUCTIONThe relational database market is a worldwide businessof over 20 billion USD. No single company dominates thisspace, providing a fertile competitive arena for several keycompanies, including IBM, Oracle, Microsoft, Teradata, andothers. These companies compete for market share by at-tracting new business, and by encouraging companies toswitch from a competitive platform to one of their own. Thebusiness term used for the conversion from a competitiveplatform is often called a “win-back”. The process of mov-ing an application from one database to another is calleda database migration. This paper describes a prototypeand its automated migration evaluation, called the Migra-tion Evaluation and Enablement Tool for DB2, or MEETDB2.Despite the existence of international standards for databasedefinitions and language processing, such migrations rarelyoccur without development cost due to the large number ofnon-standard language and interfaces extensions that manydatabase vendors provide. Therefore, when an applicationcoded to operate against a specific databases needs to mi-grate to a new database platform, there are commonly anumber of these extensions used by the source database thatare unsupported in the target database, requiring some num-ber of database or application changes to be made before themigration can be completed.The costs of migration vary, and can range from 100%compatibility between disparate platforms to a significantengineering effort. Consequently, it never begins in prac-tice without first having an evaluation and review of thesecosts. These evaluations have historically been performed ina manual and laborious fashion by engineers highly skilledin the database platforms of both the source and the targetdatabases. Not only have such evaluations been time con-suming, subjective and prone to variabilities of all manualprocesses, but the process is further compromised by thechallenge of finding people deeply skilled in multiple combi-nations of databases. The introduction of automation intothis process dramatically simplifies the analysis in severalways, as follows:1. It significantly shortens the evaluation time.2. It reduces the skills needed by the evaluator.3. In many cases it will improve accuracy.4. It adds consistency to the evaluation results.14265. It can provide a level of detail in the assessment thatwould be difficult for human evaluators to collect inmost cases.6. It can provide critical information (i.e. what is notsupported but used extensively by customers) in futuredatabase system development’s requirement planning.MEET DB2 provides two types of interfaces: GUI andcommand line interface. By default, MEET DB2 presentsitself in a simple GUI interface that contains a source fileselection interface and a button to start the evaluation. Theutility takes a file as input that contains source code of thesource database’s DDL and procedural logic, and producesa report in HTML file as output. For expert users and thoseto wish to run MEET DB2 in batch mode, it can be invokedfrom a command line interface. Figure 1 shows a screenshotof the generated report on a test database.The following list summarizes the output of MEET DB2:1. A report of the percentage of lines of code (LOC) thatrequire modification in order to be migrated from thesource system to the target system.2. A report of the number percentage of database objectsthat require modification in order to be migrated fromthe source system to the target system.3. Detailed list of tables, triggers, indexes, views, func-tions,