A few years ago, Green Mountain Power Corp. found itself in an all-too-common predicament as it examined the state of its database systems. It had multiple systems running on multiple operating systems, spread across the company supporting specific enterprise applications.
It was a scenario that could cause even the most rock-solid database administrators to shudder as they tried to be experts on every configuration. Two years ago, IT officials at Green Mountain decided there had to be a better way.
They decided to alter a portfolio that had grown up sporadically as new applications came online. The goal was to consolidate nine databases into one and to do it in a way that would provide better performance and easier management. About halfway through the project, the Colchester, Vt., utility company is already realizing performance gains and time savings, officials said.
“Theres 101 little tasks a DBA has to do every day, and thats if nothing breaks,” said Mike Moore, DBA and senior systems developer. “To get all of that on one server, thats a big savings.”
Green Mountain was supporting seven different Oracle Corp. databases—with versions ranging from 7.3 to 8i—along with two running the former Digital Equipment Corp.s Rdb database (now owned by Oracle). Those systems were running on a combination of Windows and VMS operating systems.
With such a distributed database structure, any network disruption would ripple across the databases and hurt their performance as well, said Mike Bathalon, manager of IT operations at Green Mountain.
On top of that, the range of expertise needed to support the systems was straining an IT staff that in the past three years has been cut in half through attrition, Moore said.
“By going to a consolidated database model, we could focus our efforts on training and technical expertise on a single operating environment and a single database,” Moore said. “And thats what were driving to now.”
Critical to Green Mountains decision to move to a consolidated database was clustering technology from Oracle, of Redwood Shores, Calif. The consolidated database is running Oracle9i with RAC (Real Application Clusters), Oracles latest clustering technology that allows for sharing of disk storage. The database, which has a separate schema defined for storing the data for each application, operates on Hewlett-Packard Co.s Tru64 Unix operating system running on two Compaq AlphaServer ES45 systems, officials said.
“Without clustering and without RAC, Id feel very uncomfortable putting all our hopes and dreams onto one big database because youre risking quite a bit,” Moore said. “You can use the RAC technology, [and] you can build an extremely fault- tolerant system.”
So far, four of the databases, including both Rdb systems, have moved onto the consolidated database. Those databases run applications for a geographic information system tracking locations of field equipment, a system storing historical information on power loads and other data, a fixed asset system, and an accounts receivable system.
Two more databases are scheduled to move by early August—a work force management system and a 12GB data warehouse storing communications such as statements and notices to 85,000 customers, Moore said. In addition, a database supporting an inventory management system that tracks capital equipment should be converted by the end of the year.
The company is still working on the road map for converting databases for its customer information system, SCT Corp.s Banner CIS, and PeopleSoft Inc.s Financials. They are the largest systems and involve the most testing to be sure server-side code will operate correctly, Bathalon said.
One of the biggest improvements the company has seen so far is in the performance of some of the applications switched over. With the geographic information system, IT officials ran into performance hurdles when it was running on Oracle8i on Windows NT. The system already had the Global Positioning System information on overhead electrical equipment, but when officials tried to load the data on GPS coordinates for underground equipment, the database system bogged down. Thirty-six hours into the load, less than half of the data was transferred, Moore said.
So the officials decided to wait until the conversion to the consolidated database, and the data load for both sets of data was reduced to less than 8 hours, Moore said.
“Once you get past 36 hours, its real hard to find time to do it in a production environment,” Bathalon said. “Tru64 production environment doing it in [8] hours meant overnight, essentially.”
In another example, a recalculation application that redistributes the costs of a work order within the fixed asset database used to take about 6 hours to be completed. Once the recalculation capability was moved from Rdb running on a VAX 4000 to the consolidated environment, it took only seconds, Bathalon said.
Beyond performance gains, Green Mountain is counting on the new database model to save time and money. While officials didnt have cost-savings estimates, they said they believe it eventually will mean less money spent on maintenance for multiple database systems and operating systems, not to mention fewer headaches.