Enterprise grid computing boosts efficiency

Zatni Arbi, Columnist, San Francisco, California
zatni@cbn.net.id

Let's examine the monthly computing activities at an average company.

Granted, during the last week of every month, the human resources department (HRD) will be more active than during the other three.

Those in the payroll section will have to work out the wage that each employee should receive in his bank account. Overtime, bonuses, commissions, etc. will have to be carefully calculated.

Some of the employees may have opted to have their monthly car installments, insurance premiums, mortgage payment, etc. automatically deducted by the company directly from their take- home pay. Each of the employees may also have to be notified how much medical allowance they still have outstanding till the end of the year, etc.

All of these activities will certainly require a lot of computing resources. If the HRD uses its own servers to do the job, the servers will generally be put to work for around seven days each month and will essentially remain idle for the remainder.

Would it not be better to share their computing capacity, when not in use, with the company's other divisions, such as the research and development or sales departments?

Talking about sales, here in Indonesia retail businesses typically have their busiest sales activities of each year during the month of Ramadhan, leading to Idul Fitri.

Hypermarkets and supermarkets are visited more often as households prepare for the festive occasion. Supermarkets and hypermarkets will then have to track their sales closely to find out which items are in high demand and must therefore be replenished on the shelves more often.

In addition, with increased sales, the accounting department will have to work harder to keep its records up to date.

When Idul Fitri is over, the accounting department will have to recapitulate all the figures, including how much it owes its suppliers so that the accounts payable section can begin to make the necessary payments.

When these tasks have been completed, however, the computers may have a slack period again until the time comes to pay taxes.

These two examples show how the utilization of computing resources may vary significantly from time to time. It is not difficult to find other examples where the workload is always changing.

Take airline companies as another example. We all know that the travel industry has its own peak and low seasons. During the peak season, airline companies will require more computing power to accept seat reservations and ticket sales. During the low season, however, they may not need that much.

Another very interesting case that we have here in Indonesia at the moment is all the IT resources -- 64-bit servers, terminals and IT infrastructure -- that are being procured by our General Elections Commission (KPU).

There has been controversy over the specifications issued by the commission's team of IT experts. Regardless of the specs, however, a crucial question remains: Who will use all this expensive machinery and gear once votes have been counted and our new president and vice president have been sworn in next year?

"g" for Grid

Grid computing was the highlight earlier this month as Oracle Corp. launched its newest family of enterprise software. It was intended to serve as the software layer for the enterprise grid, which was discussed in my previous article.

The Oracle 10g, however, is not ready to enter the market yet.

Although it was the highlight of the entire week at San Francisco's Moscone Center, it was actually a sequel to an earlier preannouncement. Larry Ellison, Oracle's Chairman and CEO, told the audience during the Q&A after his keynote speech that he expected the product to be ready for the market by the end of this year.

Price announcement is expected in October.

However, so far, a number of customers have been using a prereleased version. Oracle's own Oracle University is already running on a grid of hundreds of LinTel (Linux and Intel) machines.

Oracle promises that the grid capability of its new database and application server products will enable companies to save money by redirecting the unused computing capacity in their enterprise to the tasks at hand and avoiding having to buy more systems.

However, as if to put things back into perspective, Carly Fiorina, the CEO who has so far successfully orchestrated the megamerger between HP and Compaq, reminded everyone during her keynote speech on the final day of OracleWorld that grid computing architecture was in danger of being overhyped.

She said that the technology would need four years to five years more to mature.

Nonetheless, I think the idea of being able to allocate computing resources to where they are needed at any given time is great. It will help companies -- and countries like Indonesia -- save money on computers, storage and network devices.

Keep in mind that the industry agrees that, typically, between 40 percent to 60 percent of computing capacity lies unused. It will also boost outsourcing as a way of getting all the necessary computing power in a much more cost-efficient manner, which is an option that companies and countries with scarce financial resources should choose.

This is how computing resources should be acquired by KPU, which, reportedly, is planning to procure multiple units of very expensive 64-bit servers.

Too bad that, at this juncture, grid computing technology is not yet fully ready for prime time. Otherwise, this country could save itself a lot of money by buying only systems with standard Intel processors.