On-demand computing smoothes out highs, lows

Maintaining your highest level of service to customers even during peak times can be an expensive proposition if it means purchasing resources that sit idle when the spiked demand is over. In the mutual fund industry with its massive peak season, that could mean equipment is idle 95 per cent of the time.

Mississauga, Ont.-based Unisen Inc. provides investor services to about 160 investment firms and insurance companies. The wholly-owned subsidiary of AGF Management Ltd. already prides itself on its flexible service relationship with these clients. That flexibility is expected to be enhanced when it completes a current infrastructure upgrade to more cost-effectively offer transfer agency and fund accounting services even during the peak RRSP season.

Unisen was created about a year ago as a composite of four firms that were either acquired or other AGF affiliates. Robert Smuk, Unisen president and CEO, reports that the amalgamation resulted in multiple platforms.

They selected one system to use which is based on the IBM iSeries. The company is in the process of consolidating 10 servers down to two and adding applications in clients onto those two new boxes.

The upgrade began the first week of August of this year. Smuk expects it will take about six months to complete the migrations and consolidations.

Smuk says consolidation alone will save the maintenance and replacement costs of implementing software patches or upgrades across 10 machines. He adds that since the new system is faster and more efficient, it can do more for less cost which will lower transaction costs. The new servers also offer increased security and greater scalability, but it was the on-demand processing capability that drove this upgrade, he says.

The IBM iSeries comes with a feature called logical partitioning (LPAR) which IBM introduced in 1999. Steve Henderson, IBM Canada’s territory manager for finance, insurance and services, describes this as multiple CPUs sitting on one physical machine, ready to be turned on when additional capacity is required. Resources can be allocated dynamically to a specific logical machine. This provides more processing capacity along with the flexibility to shift resources to where the workload is greater, he says. “So when you do need it, you no longer have to place an order, have the capacity built and have it shipped. Instead, it is there and ready to be turned on.”

IBM’s on-demand arrangement builds on this capability by allowing Unisen, in this case, to turn on additional processors whenever the workload for their clients jumps significantly, such as on a heavy trading day or during RRSP season. Unisen can then turn them off when the spiked demand returns to normal loads, so it only pays for the capacity as it uses it, rather than buying the full capacity outright and having it sit idle most of the time. Unisen has the option to rent or buy the additional capacity.

Smuk says they can run 75,000 financial transactions per month on 30 to 40 per cent of one of the new servers. “In the old stepped world, you start at 60 per cent utilization, you go up 80 to 85 per cent utilization and you suddenly have to have a large fixed upgrade cost, a large capital outlay,” Smuk explains. “Now we’ve eliminated that. It’s a bit of a one-time up front but now it is very much a variable and we should be able to model our revenue and the iSeries expense inline, thereby not negatively impacting our margins. That was something that was very important to us.

“We could have continued to leverage the technology we have in place for a period of time but it didn’t provide us the flexibility to tie our expenses to our revenue as tightly as the on-demand processing capability does,” he adds.

Unisen ran a financial analysis to ensure that there was more value in having a higher incremental cost that is variable than having a lower incremental cost that is not variable. Smuk says the company charges clients based on activity and the number of fund accounts. The number of fund accounts drives the quantity of transactions which in turn dictate the computing power required.

“If we price to our clients on something that is variable, then we want a linkage of our systems cost to be tied closely to that so we can continue to have the same margins,” he explains. “The on-demand process provides us an opportunity to leverage the capabilities of the system for that peak and not continually have to pay for it the rest of the year. It comes back to the same thing for clients: they want peace of mind and to ensure that you can meet your deadlines.”