The Brute Force Approach

Wall Street executes financial operations every single day. However, most of those tasks do not generate revenue yet they still incur cost. That’s a problem; obviously, especially when you consider that these tasks are done more than once. I know, I know that's a definite understatement. The point is that cost keeps increasing.

A few years ago banks started looking for ways to solve this problem. The answer they came up with was “grid computing”. In fact they invested 120 million dollars in it just last year. They needed more power and they got it. Makes sense right? Wachovia and JP Morgan have been investing in this technology since 2001, even the Chicago Stock Exchange got in on the act as early as 2002 when they purchased Oracle 9i RAC. Celent and Hersh predicted that investment in “grid technology” would increase to 500 million by 2010. Banks were able to decrease cost by leveraging their relatively inexpensive hardware. Problem solved right? Wrong. (For more info see: < http://www.wallstreetandtech.com/showArticle.jhtml;jsessionid=Q0U5FSIPLYWAMQSNDLOSKHSCJUNN2JVN?articleID=187203197 >)

That was 2006. This year Wall Street’s computational needs have increased and so has its cost. Consider this: Each Intel chip requires roughly 75 watts of power, multiply that by the thousands needed for the operations they perform and that’s a lot of money…way more if you throw in cooling costs. So what new answers have they come up with? One of those answers is “multicores”. Basically Intel and AMD have just increased the number of processors they give these banks. Intel has even developed an 80 core-chip, which they claim uses less energy than a household appliance. However, whether it will actually hit the market is still not certain. Accelerators are also popular and they too come in the form of chips. But whether they are GPU’s or FGPA’s they are still aimed at achieving the same goal...increasing your computational power. But that just seems like throwing money down the drain- it's hopeless! (For more info: < http://www.wallstreetandtech.com/showArticle.jhtml;jsessionid=BRQCSHECZUQO4QSNDLPCKH0CJUNN2JVN?articleID=198001925 >)

I’m exaggerating, but I do think that there have got to be better ways to approach this problem. I think Merrill Lynch has a good idea. They’ve already outsourced some of these processes (See above link). So has Wachovia. Well, not completely. They’ve actually developed a system to allocate system capacity based on how “mission-critical” a process is using IBM software. (For more info: < http://www.wallstreetandtech.com/showArticle.jhtml;jsessionid=3T0VLFLYHOBJUQSNDLRSKH0CJUNN2JVN?articleID=198001931 >) Why keep beating around the bush? I know, it’s not like these banks are designing an information system, but the costs are still significant. Rather than invest in processors, or blade racks, let someone else do it. SOA…or even SaaS, I’m just throwing things out there. But of course there is no perfect solution. As my friend Benson was kind enough to enlighten me, there are still problems that can arise with resorting to web-services; take DoS for example. That could be an issue when dealing with sensitive financial data. But does that mean that businesses are slaves to high performance computing and its insatiable lust for power?