• United States
Contributing Writer

PeopleSoft upgrade drives consolidation

Apr 05, 20045 mins
Computers and PeripheralsEnterprise ApplicationsIBM

Hospital turn to server partitioning and clustering to reduce maintenance costs.

Oakwood Healthcare System • Dearborn, Mich. • Oakwood Healthcare System had the classic recipe for a server consolidation: Technology costs were rising as applications and IT initiatives grew, server volume was becoming cumbersome to manage, data center  space was at a premium, and the workload was dramatically increasing.

So when its 5-year-old PeopleSoft financials and human resources system needed to be upgraded, the regional healthcare organization took the opportunity to transition from aging IBM  RS/6000s to new IBM eServer pSeries servers.

Intro: Extreme server makeovers

Beyond the Superdome

Escape from D.C.

The big box theory

“We’ve proven we can work with fewer physical platforms and handle them with fewer people,” says Brian Perlstein, lead systems architect in the design and infrastructure group at Oakwood, a network of four hospitals, 15 healthcare centers, two skilled nursing centers and numerous healthcare facilities in southeast Michigan.

Of the four servers, two have eight CPUs, one has four, and one has 16. Perlstein plans to add two more eight-processor pSeries servers to consolidate another five RS/6000 servers. All are connected to two IBM TotalStorage Enterprise Storage Servers.

By reducing its server count, Oakwood has reduced its ratio of technical staff to servers from 17-to-1 to 13-to-1. (That figure also includes Windows servers.) And because of the reduced labor requirement, the staff has been freed to learn about automated monitoring tools, which has further reduced IT costs. “We’re taking the skill set of scripting to automate system administration tasks so we can manage more systems than in the past,” Perlstein says. “Consolidation has assisted in that automation because it opens up a window of time.”

The four pSeries servers are divided into 10 logical partitions (LPAR), and therein, Perlstein says, lies the magic. Through LPAR, the applications appear to be running on their own physically separate pieces of hardware even though they’re sharing physical components with another application. “If I have four CPUS, I can split them off into four different LPARs – but the only thing that’s really separate is the I/O cards and the disk,” Perlstein says. This saves physical space and the cost of running two separate pieces of equipment.

The only disadvantage has been educating some of the medical application vendors that require their applications to be run on separate equipment. “We built an LPAR test box and let them dial in to see if they saw problems with extraneous overhead,” Perlstein says. As it turned out, the vendors couldn’t tell the difference.

Perlstein warns that maintenance scheduling gets tricky when you’re running multiple applications on one box. “The complexity comes with a full-system microcode upgrade or something that’s going to affect all LPARs across the board,” he says. “It’s one of the risks you take when you do consolidation.”

Perlstein has prepared for the more routine maintenance tasks by using clustering, which lets applications fail over to another set of processors when they need to be upgraded. This not only reduces the amount of application downtime, but it also opens up a larger maintenance window for the IT staff. Clustering is less expensive with consolidated servers because with LPARs, the application does not need to fail over to a separate box – you can accomplish it within one server.

All of this is very important to Oakwood, which is attempting to increase system availability 20% this year, from five-nines to seven-nines. “Healthcare systems are becoming more critical, and everything has to be highly available,” says Ken Garner, manager of client/server at Oakwood.

Oakwood has seen so many benefits from consolidating its Unix servers that it is shifting its consolidation focus to the Windows NT side of the house. “Because Microsoft is such a target of viruses and hacker attacks, we spend more time on the maintenance side and need more people to monitor those servers,” Garner says.

All told, consolidation has become an intrinsic part of Oakwood’s IT culture. Before doing business with any application vendor, for instance, Perlstein and Garner tell them upfront that their environment involves clustering and consolidation. “We ask them, ‘Will you support this, and if not, why?'” Perlstein says.

Previous architecture:
25 RS/6000 servers
Consolidated architecture:
15 IBM eServer pSeries servers of which four, totaling 36 CPUs, use logical partitioning. Oakwood plans to add two more pSeries servers with 16 CPUs to consolidate five more servers.
Two IBM TotalStorage Enterprise Storage Servers
Two Cisco 9509 SAN Directors.
Consolidated applications:

PeopleSoft HRMS and financial management

Main reason for consolidation:
With staff constraints, Oakwood needed to reduce the labor required to manage servers. It was also maxing out real estate in the data center and wanted to decrease the cost of IT.
Up to 15% annual reduction in technology costs.
Up to 20% increase in application upgrades and runtime.
Reduced month-end processing on the financial system by at least 20%.
Decrease in payroll processing time from three days to two.
Went from 17 to 13.5 servers per technician.
Lessons learned:
“We’ve gotten a lot more organized with the questions we ask our application vendors,” says Brian Perlstein, lead systems architect in the design and infrastructure group at Oakwood. “We tell them upfront, this is what we’re doing with clustering and consolidation — will you support it?”