Investment banks have been leading early adopters of grid computing, and in the third 451 Grid Adoption Research Service (GARS) report, published at the end of 2004, 451 analysts looked at the current deployments and plans of this group. Now, more than a year later and in the tenth report, The 451 Group has looked at the progress of investment banks and how their deployments have changed since 2004, as well as how they will evolve their Grid strategies throughout 2006 and beyond.
Investment Banks See New Benefits
While in the third 451 GARS report only a few players in this sector had grid deployments, such implementations are now becoming far more widespread. The view from most investment banks is that, for the first time in a long time, IT looks capable of driving sustainable, long-term, and linear cost savings and performance improvements for this sector. Grids, plus supporting technologies such as blades and virtualization, appear to be leading this trend.
Where recent conversations differ from those in the past is in the raft of new benefits seen for grid deployments. These include:
- Internal collaboration/sharing.
- Quicker response times (moving toward real-time activities).
- Delivering stable and horizontal IT environments on which to more efficiently deploy business processes.
One bank observed that grids do not reduce TCO or save money, but they do enable the bank to make more money.
Overcoming Challenges and Taking on New Ones
Many of the specific challenges faced by investment banks today are related to moving beyond the compute grid. The problems identified in the third 451 GARS report – unmanaged code, lack of Grid-enabled enterprise applications, peak loads — have largely been solved, or at least overcome, by many investment banks.
This progress among investment banks comes partly through experience. For example, JPMorgan says that after four years of work on its Compute BackBone (CBB), it is now able to throw CPUs onto the CBB grid, which has been the key to solving most problems.
The following is a look at some of the more specific issues:
- Excel models
Investment banks still face the challenge that most models are based in Excel, and they do not want to redesign them for Grid deployment. This can still be a problem today, but commercial software vendors-such as Platform, DataSynapse, GigaSpaces and Digipede have improved their offerings to deal with the issue. Banks have also deployed in-house software teams to address the problem and, through experience, have worked out how to minimize the problem through the way they schedule jobs.
- Software licensing
Software licensing is always a popular choice among the obstacles to grid deployments! However, banks are atypical users, as they don't tend to have concerns about licensing models for commercial applications used on grids — they can afford it or can bring sufficient organizational scale (size) to bear down on the problem.
- Cultural issues
A particular challenge in the investment banking sector is the autonomy of the individual lines of business and the politics between them and central IT. Cultural issues particularly revolve around the issues of prioritization (who gets the resources and when?) and cost (who pays?). There are also lots of issues around the challenges of different groups working together, and some banks have been happy to leave different groups, particularly in different geographies, to use different vendors and deploy grids in different ways rather than try to enforce a company wide strategy. As investment banks go beyond the compute grid, these issues become ever more complex and challenging.
Security is clearly an overarching concern for the CIO of an investment bank, yet Grid security isn't — the activity is confined within the firewall.
A number of banks pointed to a shortage of skilled IT staff as a potential chokepoint. Interestingly, we are seeing a number of enterprise IT specialists with “Grid” in their job title now — something that was very rare at the time we researched the third 451 GARS report.
- Data management
The issue of data management was covered in depth in the eighth 451 GARS report, and although the in-house IT teams and large IT budgets of investment banks can mitigate the challenges in this area, it is still widely seen as an obstacle.
Moving Beyond the Compute Grid
Nearly half of the financial organizations profiled in this report have siloed grids. And the majority of these companies are either linking or plan to link their grids. At this stage, 451 analysts are seeing a significant advance in technology deployment (such as WANs) from siloed grids, as well the overcoming of organizational/cultural issues (crossing departments/LOBs, sharing resources, ownership and control of those resources). In most cases, the latter effort is being driven specifically by top-down CXO involvement.
Grid computing is generally already seen as a corporate success story at investment banks, although the grid itself isn't an answer to anything, but rather a means to an end. Certainly, for investment banks, there is a sense that the compute grid is mature (or at least well understood) and that it will gradually be absorbed into the fabric of next-generation distributed computing environments. The R&D group of one leading European investment bank is no longer involved with the grid on a week-by-week basis, and it has moved on to other strategic objectives using grids as the foundation. On the arc of emerging technologies, Grid computing has moved well beyond hype or experimentation and into implementation for tier-one banks.
Whether banks are using grids to support or enable high-performance computing, service-oriented architecture or shared internal utilities, or they are examining outsourced Grid-based capacity, they all have long-term and far-reaching plans to extend their grids.
All banks interviewed for this report are looking for additional internal applications-or indeed, entire groups/businesses-that are suitable for Grid-enablement.
About half of the organizations interviewed for this report have a plan to implement a service-oriented environment in some fashion, with SOA being the key enabler here. Not all organizations will be using grids to implement the environment, however. But to be clear, as far as those users we have spoken to are concerned, SOA equals Web services, which are lent fault-tolerance and scalability by running on a grid and using virtualization. Indeed, in general, we're seeing grids being perceived as the technology infrastructure and SOA being perceived as the application infrastructure.
SOA is regarded as a way to implement services that are componentized, reusable and interoperable; are developed independently of the architecture they run on; and can take advantage of loosely coupled resources that are also provisioned independently of the applications. SOA also provides a better platform on which to implement business processes that can span disparate business units, supporting the pursuit of improved business-IT alignment.
One example might be that instead of having to log into multiple applications to get a single view of a bank's risk or a client's exposure, a risk model based on the use of Web services components can be executed against an entire portfolio and deliver the desired view in near-real-time instead of minutes.
Some users don't talk about SOA at all in their plans, or they regard Web services as too immature to support a production SOA environment on grids at this time. Communication between Web services components is unreliable, and the danger is that the services won't be available when an application that is constructed by assembling these discrete services is executed.
Moreover, for a bank like UBS, the reality is that for the foreseeable future application services will continue to be of an intra-departmental nature. That's because there is, and will remain, tremendous value in the custom applications used by each department.
At this point, it appears that SOA may have more to do with new application development and implementation strategies — certainly where they cross departments — while grids are perceived primarily as a way of chopping up existing applications.
The bottom line is that users expect grids will allow them to run Web services better, faster and cheaper because they are cross-platform, can scale and offer better utilization.
Grid Economy … IT Economy
HSBC with a shared utility model, a major European investment bank with a virtual resource market and JP Morgan Chase are perhaps the most advanced in terms of plans to create internal, company-wide shared IT utilities. Shared utilities using grids imply significant technical challenges, including linking grids over LANs and WANs and creating a virtual resource pool from a diverse range of IT assets.
The creation of what might then be regarded as “Grid economies” or “IT economies” using grids will also require the establishment of marketplace principals and measurements for the trading of available resources, with the attendant metering, billing and charge-back elements.
Utility models furthermore imply that the organization and its groups will need to embrace the concepts of using non-dedicated resources and sharing those resources that are dedicated. The key technical challenges for users extending their infrastructures in these ways are bandwidth (networks), latency (time) and storage (data volumes).
While our interviews indicate that there is still a mix of grid strategies, with some driven top-down from the CXO and some driven bottom-up by lines of business, it seems that specific direction from the CXO is a requirement for establishing shared utility models.
For more information on this or any other reports in 451 GARS, visit
About Steve Wallage
Steve Wallage is a director of research at New York-headquartered The 451 Group — an independent technology industry analyst company focused on the business of enterprise IT innovation.