html5-img
1 / 48

ACES at Duke

ACES at Duke. Re-inventing a custom web-based student system application. Part 1. Making decisions. What is ACES?. Duke’s online student system 24 hour access to registration, grades, bio/demo data, financial data, advising tools. What is STORM?.

lotus
Download Presentation

ACES at Duke

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. ACES at Duke Re-inventing a custom web-based student system application

  2. Part 1 Making decisions

  3. What is ACES? • Duke’s online student system • 24 hour access to registration, grades, bio/demo data, financial data, advising tools

  4. What is STORM? • Duke staff and faculty web access to student systems information • Access to student data, class information, applicant information • Ability to perform common course and student data management functions

  5. Why? • HTMLAccess (the PeopleSoft product) was immature with unknown performance • IVR wouldn’t be able to handle registration with a PeopleSoft backend • Difficult interface and cumbersome navigation within the PeopleSoft client • Easier to keep customizations outside of PeopleSoft

  6. Why? • Desire for added functionality • Registration • Curriculum/Advising • Online Application Entry • Differences across professional schools

  7. Bookbag allows for easier advising, less load during registration time, simpler course planning

  8. Previous Technology • HAHTSite/VisualBasic front end • Access to PeopleSoft Oracle 8i database via ODBC • Extra tablespace to support added functions • PeopleSoft Message Agent used to save data, so business logic is executed by PeopleSoft Application server

  9. Hardware • Cisco CSS load balancer • 3 IBM H70 AIX 4.3 Web Servers (iPlanet) • 3 IBM Netfinity Windows NT backend servers

  10. Downside • High support costs for HAHTSite and iPlanet • HAHTsite version no longer supported • Very specialized programming knowledge required for maintenance • Expensive hardware

  11. Step 1: Selling Java • A chance to clean up and consolidate code • OO design makes for a simpler design and easier maintenance • Shared code libraries reduce redundant code • A chance to rethink and redesign application pathways based on user feedback

  12. jBoss 3.0.3 Jetty 4.2 Jdk 1.4 Jdom b7 Ant Jedit cvs Total Cost: $0 Step 2: Selling Free Tools

  13. Step 3: Architecture • Remove business logic from display logic • Achieve greater flexibility and fault tolerance in system design • Facilitate a larger programming team • Maintain as much insulation from PeopleSoft patches and fixes as possible • Increase usability of custom business logic

  14. Making the move • Redeveloped business logic to work against PeopleSoft 7.6 • Business logic all contained within a combination of Stateless Session, CMP and BMP Entity EJBs • All business logic returned as XML with an eye toward a future portal • XML converted to HTML with XSLT

  15. The Next Steps • Changing business logic to use PeopleSoft 8.1 table changes • Changing message agent calls to use the component interface • Exposing business logic to other Duke applications – generating PDFs, Excel Spreadsheets, reports

  16. New Hardware • IBM xSeries rack-mount servers • Inexpensive • Not overkill for the job • Having a farm makes maintenance potentially more tedious, but reduces negative impact of any system failures and increases scalability • Minimal machine room space and lower power requirements

  17. The Future • Drop ACES in favor of PS-delivered web client? • Use business logic for Duke-specific functions and incorporate it into the PS web client

  18. Part 2 Using the PeopleSoft Component Interface with Java

  19. What’s the Component Interface? • PeopleSoft’s way of allowing 3rd party programs access to business logic in version 8.x • Replacement for the version 7.x Message Agent

  20. What do you get? • Data entry with the Component Interface simulates entering data directly into the pages • Order of parameters set was not guaranteed under Message Agent • All PeopleCode is fired, but order is not guaranteed

  21. Step 1 • Within App Designer, create a component interface mapped to the component you want to access

  22. Components in version 8 are like panel groups in version 7

  23. Step 2 • PeopleSoft begins generating the component interface by selecting all fields from the component referenced • Fields can be added or removed at will, but less dangerous to leave them all and only set the ones you care about in the client program

  24. Step 3 • Build the API (the .jar file to use with the client)

  25. Make sure Java Classes Build Box is checked

  26. Step 4 • Generate the client template

  27. Coding the client application • Add the following imports: import java.io.*; import java.math.*; import psft.pt8.joa.*; import PeopleSoft.Generated.CompIntfc.*; • The joa package comes with PeopleSoft • The .Generated package was built by the previous steps

  28. Coding the client application • Setting the server name and port String strAppSeverPath = “//153.4.555.1212:9000"; String strOperatorID = “username"; String strPassword = “password"; • App server path is IP (or DNS name) of application server machine, followed by a : and the JOLT port address

  29. Coding the client application • Create the session ISession oSession = API.createSession(); • Connect to the server oSession.connect(1, strAppSeverPath, strOperatorID, strPassword, null); • Get the component IEnrlRequest oEnrlRequest = (IEnrlRequest) oSession.getComponent("ENRL_REQUEST");

  30. Coding the client application • Set the keys oEnrlRequest.setEnrlRequestId(“0000000000”); oEnrlRequest.setEmplid(“0039646”); oEnrlRequest.setAcadCareer(“UGRD”); oEnrlRequest.setClassNbr(1022); oEnrlRequest.setEnrlReqAction(“E”); oEnrlRequest.setInstitution(“DUKEU”); oEnrlRequest.setStrm(“0960”);

  31. Coding the client application • Create the new request oEnrlRequest.create();

  32. Coding the client application • Adding classes IEnrlRequestEnrlReqDetailCollection oEnrlReqDetailCollection = oEnrlRequest.getEnrlReqDetail(); IEnrlRequestEnrlReqDetail oEnrlReqDetail; for (int i = 0; i < oEnrlReqDetailCollection.getCount(); i++){ oEnrlReqDetail = oEnrlReqDetailCollection.item(i); // Add an item to the collection for each class. oEnrlReqDetail.setClassNbr(1022); oEnrlReqDetail.setEnrlReqAction(“E”); }

  33. Coding the client application • Save the request oEnrlRequest.save(); • Get the generated enrollment request ID String enrl_request_Id = oEnrlRequest.getEnrlRequestId();

  34. Part 3 Design and performance considerations

  35. Lessons learned from previous version of ACES • Database tuning (indices, hints) is critical • Network traffic accounts for 50-75% of the time spent waiting for results • Co-locate a web-only PS app server and database • Use connection pooling wherever possible • End-to-end performance testing is critical, the whole is not always the sum of the parts

  36. Some EJB Decisions • BMP used much more heavily than CMP for read-only data • Flexible caching algorithms and parameters • Database hits can be greatly minimized on common data at the expense of slightly stale results

  37. Some EJB Decisions • Stateless session beans used for up-to-the second read-only data and to save infrequently used data that does not need to be cached • CMP used for frequently updated Duke-specific data controlled entirely through the web application (the bookbag)

  38. Consolidation of Machines and Processes • Co-locating EJB, Servlet, and web server on the same machine, and if possible within the same JVM greatly increases performance

  39. Connection Pooling • Database • App server connections

  40. Exposing EJB methods to clients • Use one stateless session bean to expose the API to EJBs • Reduces the number of JNDI lookups that need to be done to acquire an EJB

  41. Investigate Clustering and Cache Options • Jboss is highly configurable • Clustering may decrease performance because of the introduction of network traffic and coordination, but not using clustering can cause skewed results if sticky sessions are not used

  42. Contact Information • Matt Young (matt.young@duke.edu) • Richard Outten (richard.outten@duke.edu) • http://www.jboss.org • http://jakarta.apache.com • http://www.jedit.org

More Related