High-performance computing is moving deeper into the financial sector, reshaping functions such as marketing and longer-term investment strategies — areas where, up to now, the human mind has prevailed.
That is a shift from the recent financial industry past, where the impact of new computing and networking tools mostly was felt in the realm of faster trading, according to E. Paul Rowady Jr., senior analyst with the TABB Group.
The financial sector appears to be at the beginning of a broad transition to high-performance computing, which already has enabled tremendous growth in other sectors of the economy, most notably the massive consumer Internet companies that pioneered the technology for their own use. “I think we are still very much in the beginning. Not any of this is mainstream yet,” Mr. Rowady says. “We have seen a lot more interest in the topic in the last six months.”
High-performance computing can be measured in terms of processing speed and the complexity of data-processing and analysis. “You can move around those two axes,” says Mr. Rowady. It involves more parallel architecture — multiple processing cores, multiple servers, multiple clusters of machines and multiple, concurrent execution of threads, the basic units of instruction. Says Mr. Rowady “There are layers of parallelism.”
In earlier years of computing, advances in performance were measured in terms of higher frequency processors. Those efforts began to push against their upper limit, and newer approaches to innovation involved yoking together clusters of commodity machines. On the hardware side, databases were speeded up by storing information directly in processors, instead of on separate storage units, a concept known as in-memory storage. And conventional disk drives increasingly are replaced with faster solid-state memory, which powers smartphones and tablets.
Much of this technology emerged from companies such as Google Inc., Yahoo Inc., and Facebook Inc., which needed ways to manage their hyperscale communities, boosting their computing power while lowering cost, power consumption and the use of space at the same time. Such demands led to the development of Hadoop, an open-source software framework for storing and processing large amounts of data on clusters of low-cost machines.
While Internet companies and many other corporations have been using Hadoop for nearly a decade, “a small set of tier one banks are just now experimenting with Hadoop,” Mr. Rowady says.
Throughout finance, Mr. Rowady says, a broader range of use cases for these technologies is coming into being:
Risk and Pre-trade Analysis
Pretrade analysis that might have taken an entire night can now be run intraday, or even in real time. These problems include the complex pricing of fixed-income instruments that don’t have readily observed prices. The prices must be triangulated, based on prices of instruments in other markets. Typically, this work was started at the end of the day and ran overnight.
Increasingly sophisticated forms of artificial intelligence and automation will make it possible for computers to take a bigger role in managing strategies with longer holding periods, which are still dominated by human decision-making.
Client-facing Activities and Project Management
Increasingly, data will offer greater decision support, helping sales traders and brokers figure out what kind of resources to devote to a particular client, whether a phone call or a meeting or a more automated engagement is best.
Unified communication tools comprise a spectrum of messages, such as voice calls, email, chat and social communication. Video can be added, too. Financial firms that must demonstrate to regulators they are monitoring all communications will take advantage off high-performance computing platforms that recognize patterns. While the use of these surveillance tools is mostly defensive for the time being, they also open up new opportunities to generate revenue by analyzing stream of customer data, according to Mr. Rowady. But he warns that proper governance standards that take the need for privacy and transparency into account are critical.
While technology may be a source of competitive advantage, the benefits aren’t guaranteed. “There is a change management aspect to it. The technology doesn’t guarantee that you have the creativity to use the tools properly. Just because you have them, doesn’t make you more competitive, but it might give you a better shot at being competitive,” Mr. Rowady says.
Success depends on four factors: infrastructure, software and processing, data and human capital, he says.
“You need to optimize deployment of all four, which will depend on the use case. Some are reliant on, say, speed, but you need to optimize all four factors to be successful. Ultimately, human capital is the one that stands out,” Mr. Rowady says. “If you have the right people, you can tune hardware and software to use data and the output will be a strong competitive advantage. Not everyone will have the right people.”
Companies need to infuse their organizations with people who have high-quality engineering skills, and with more data scientists and managers. In some cases, they will need people skilled as process engineers and user experience designers. Presentation and visualization of data will be key, so that people who are not skilled engineers can understand and manipulate data.
The availability of high-performance computing over the cloud will lower costs and spread its usage, leveling the playing field and making it easier for startups to introduce innovation into financial markets. Instead of buying technology and talent, companies can rent it over the Internet, paying only for what they need, and scaling their capabilities up and down in real time.
These changes won’t occur overnight, according to Mr. Rowady. Architecture and business culture and process evolves slowly. “In general, there will be more incremental change, but some things may cause spikes in ability, spikes in innovation.”
Over time, though, the impact may be profound. Says Mr. Rowady: “High performance computing is likely to unleash a level of creativity. You could see the pace of innovation go parabolic, although it may seem incremental for awhile.”