Directory Service Deployment

   

After HugeCo completed the design of the extranet applications and directory, it embarked on the deployment phase of the project. HugeCo used a phased approach: Products were chosen , an extranet pilot project was deployed, and then the production service was rolled out to all retailers.

Product Choice

Because HugeCo had already deployed Netscape Directory Server for its corporate intranet, continuing to use it for the extranet applications made sense. Nevertheless, HugeCo analyzed the requirements of the extranet directory to verify that the Netscape product was suitable. This analysis detailed the types of operations that the extranet applications would perform against the directory. The following operations were included:

  • User authentication and authorization . Whenever a user logs in to the order entry and tracking system, his or her user ID and password are sent to the application server via the Web server, which verifies the information and looks up the corresponding entry in the directory. A browser cookie is returned to the client, which associates further activity with the user. Because no further interaction is required with the directory until the user logs in again, user authentication and authorization place a very light load on the directory.

  • User start page generation . A user's personalized start page includes a list of relevant product bulletins or announcements that the user has not yet viewed . To generate this list, two directory entries are retrieved. First the user's entry is retrieved and its hugeCoHrpRetailerID attribute examined. Then the corresponding retailer entry is retrieved from the directory, and the hugeCoHrpProductAuthorized attribute is extracted. When these pieces of information have been retrieved, the application has enough information to display the new product notices for only those products that the retailer is authorized to sell. When a new order is placed, the order entry and tracking system consults the retailer's directory entry to verify that the retailer is authorized to sell the ordered product.

The types of queries required by these applications are well within the capabilities of Netscape Directory Server. Hence, HugeCo decided simply to install additional instances of the Netscape server to support the extranet application. If, on the other hand, features required by the application were not provided by the Netscape server, HugeCo would need to evaluate other directory servers for suitability, or redesign the applications to work in the absence of those features.

Piloting

As with the original directory deployment, HugeCo rolled out the extranet applications and directory in an incremental fashion. The process began with a small-scale pilot that provided valuable feedback and validation of the directory design. The pilot involved three different types of retailers ”a home improvement superstore, a small-town hardware store, and a homebuilder ”in different parts of the United States. Selecting these three types of retailers provided a set of users with a wide range of computer experience. The large home improvement center, for example, has its own computer consultant (shared with the other stores in the region), whereas the homebuilder serves as his own computer consultant.

HugeCo staff visited each pilot site. They assisted the retailer at each site with hardware and software installation, created entries and issued passwords for the users, and then stayed for three days as the pilot users began to use the extranet system. During this time, any difficulties the users encountered while using the system were noted and fed back to the system developers at HugeCo. Of particular interest to the developers were the experiences of the local administrators as they created and removed accounts for their staff using the administration utility hosted on the extranet Web server.

Also of interest to the extranet developers was the quantity of directory load placed on the extranet directory servers. During the pilot the developers kept logs for the extranet directory servers and observed how each individual retailer generated load on the directory. Extrapolating from the pilot usage data allowed the developers to validate their assumption that three directory servers (a writable master and two read-only replicas) would be sufficient for the needs of the extranet applications. Of course, if any new applications are deployed in the future, the load on the directory might change significantly, so this assumption must be revisited periodically throughout the lifetime of the extranet.

After a pilot site was installed, the HugeCo staff moved on to the next pilot site but remained available by telephone for assistance. After all the pilot sites were installed and operational, the pilot staff returned to each pilot site in order and interviewed the pilot users. The collected feedback from all the pilot sites and the usage data from the extranet directory servers were reviewed by the developers, and changes were incorporated into the system. Most of these changes involved alterations to the user interface of the order entry system.

Putting Your Directory Service into Production

After the pilot was completed and applications were updated to incorporate feedback, HugeCo planned its production rollout of the service. Although HugeCo believed that the application would be easy to use, it decided to be conservative in the number of sites to be brought online, just in case the retailers needed a lot of telephone support. To avoid swamping the Help Desk with calls, only 25 retailers were initially brought online. One week after the retailers went live, follow-up calls were placed to make sure that they were successfully using the application.

After several weeks of enabling 25 retailers per week, HugeCo was confident that it could handle a larger volume. This conclusion was based on the frequency of support calls and the usage patterns on the extranet directory servers, which were increasing as expected but were well within the capacity limits of the servers. Volume was increased to 100 retailers per week, and all retailers were brought online within ten weeks.

To encourage use of the new application, HugeCo tracked the usage of the extranet applications and the telephone order center. Initial results were disappointing ”the call center volume actually increased for a period of time ”but the help calls eventually began to decline as retailers became more comfortable using the system.

Data was collected and analyzed during the entire process to determine if the directory was in danger of becoming overloaded. On the basis of the experience they had gained while deploying HugeCo's corporate intranet directory, staff members felt confident about their abilities to perform capacity planning for the new extranet applications. Basic performance metrics, such as average search response time, were developed and measured throughout the directory deployment phase, and directory load remained well within expected levels.

   


Understanding and Deploying LDAP Directory Services
Understanding and Deploying LDAP Directory Services (2nd Edition)
ISBN: 0672323168
EAN: 2147483647
Year: 2002
Pages: 242

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net