Financial Services

Cray Solutions for Financial Services

Financial services firms operate in a complex and changing business environment. Cray solutions are a key ally in managing risk — and outsmarting the competition.

Connect With an Expert

Get Leading Technology...and a Powerful Ally

The financial services industry operates under ever-intensifying pressure. From meeting new regulatory demands to battling cyberattacks to capitalizing on big data opportunities, firms face increasingly complex problems. And they must respond quickly, through financial risk analytics and more, to manage risk and stay ahead of the competition.

While the industry has long relied on large compute grids, these commodity systems struggle to cope with today’s more complex and real-time workloads. Solving problems such as whole portfolio counterparty credit exposure calculations or detecting fraud using big data requires new techniques more closely aligned with supercomputing than commodity grids.

Arming yourself with a proven technology partner is the first step in adapting to a changing business environment. Cray brings over 40 years of supercomputing leadership to every product and solution — experience that helps you meet the exacting demands of your industry.

  • Risk management – Enhance risk management capabilities while significantly reducing TCO.
  • Trading strategies – Massively increase strategy discovery and back-testing throughput at the same time, keeping costs lower than with traditional commodity clusters.
  • Quantitative finance – The world’s most capable accelerated systems.
  • Regulatory compliance – Graph compute at massive scale, revolutionizing both cost to compliance and investigatory efficiency.
  • Financial crime – How do you find the needle in the haystack cost effectively, quickly and without too many false positives? This is the key question for AML, insurance fraud, trader surveillance and others. Cray’s graph technology enables complex anomaly detection from many data sources, moving beyond the inflexibility and inaccuracy of traditional rules-based systems, at the scale necessary to be useful and uncompromising.
  • Cybersecurity – Like financial crime, this is about finding needles in haystacks from many data sources. Software partners approach Cray because they understand that anomaly detection is easy with graph but tough to do at scale. With Cray cybersecurity anomalies can be rapidly spotted and investigated at lightning speed, transforming your defensive capability.

Featured Resources

Cray Gets Top Marks for Financial Services Analytics Benchmark

In this benchmark report from the Securities Technology Analysis Center (STAC), the Cray XC40 Intel Xeon Phi system outperformed all previous results for a benchmark that tests technology stacks used for compute-intensive analytic workloads used in pricing and risk management.

The Future of Grid in a Cloud-Enabled World

This TABB Group research report considers grid’s place in a cloud-enabled world, addresses options for optimizing onsite system performance, and discusses total cost of ownership (TCO) considerations for on-site grid components.

Cray Analytics Solutions for Financial Services

Find out how the Urika-GX system can provide a solution for trading analytics, fraud detection, anti-money laundering, cybersecurity and more.

See All Resources

Financial Services Solutions

Risk Management

A comprehensive approach to risk management is critical to the health of financial institutions. Yet relying on a system that can't keep pace with today's volume of data and increased performance requirements will impede risk management.

Read more (open/close details)

The Challenges of Risk Management

While the concept of applying analytics to risk management isn't new, there are significant challenges in applying analytics in today's environment:

  • Explosive data growth – Whether you're a bank, insurer, trader or other financial institution, keeping pace with the deluge of digital data is challenging. No longer are desk-level views enough. A trade affects not just the desk, but the entire bank's risk position in real time.
  • Near-real-time expectations – Risk models that have traditionally been run overnight are now being built out to include intraday, real-time and complex regulatory stress tests. Traders can no longer rely on last night's risk reports, but need to pivot the complex models on hundreds of variables in real time.
  • Increasingly complex models – Today's risk models are larger, incorporate increasingly complex algorithms, accommodate more scenarios and are being called upon to be increasingly forward-looking.

Increasing Risk Capability While Reducing Cost with Cray® XC™ Series Supercomputers

There is little doubt that the Cray XC system outperforms a commodity cluster — but lower TCO? When you look at the details of a typical Monte Carlo simulation it is easy to see how the XC system can reduce costs. These simulations are highly parallel, employing thousands of compute nodes. But when the work is divided up each node shares much of the same data (shared market, reference data and initial conditions). In a normal architecture this means data is duplicated to each node in hard disk or SSD. With Cray's DataWarp™ I/O accelerator, shared flash SSD sits directly on Cray's Aries™ interconnect to share fast storage between all nodes. This means not only do you not need to purchase storage for each node, but for each job you need to distribute far smaller amounts of data. This and numerous other efficiencies add up to faster job turnaround times and lower TCO.

Beyond Piecemeal: A Holistic Supercomputer Approach to Risk Management

Traditional risk software architectures are highly parallel based on each core's calculation (e.g., for a particular asset) being independent of other calculation. This works well in simple situations and for desk-level risk, but as banks start to look at the real-time impact on their entire portfolio, these become very large memory problems to do properly.

Instead of each core accessing a very large-memory single node, a more efficient structure is to have a single large-memory model shared across many nodes, where each core has access to it. Traditional compute architectures do not allow this and rely on each node keeping its entire copy of the model in its own memory. Supercomputer technologies allow the problem to be solved at much less cost by exploiting shared global address spaces.

Supporting Massive Data Storage, Movement and Processing

Built on the Aries interconnect, the XC series has the I/O to support data movement well beyond the capacity of commodity infrastructure, as well as the compute and storage support capacity to support tomorrow's largest risk models.

With a no-compromises architecture built to spend its lifespan performing at maximum sustained utilization, the XC series is your best bet for consistently providing near-real-time performance, without hiccups or unexpected surprises. And with an architecture that's been proven to provide near-linear scaling to tens of thousands of nodes, for a variety of applications, the Cray XC series is your best ally to support future growth.

The XC series also provides holistic performance. With its fast interconnects and robust compute capabilities, the XC series is capable of dealing with massive models holistically, instead of breaking them up piecemeal. This opens up new avenues, affording views of both the individual trees and the forests, of risk management.

Back to top


Trading Strategies

Quantitative traders and hedge funds live or die by discovering new smarter strategies with a greater likelihood of winning under a wider variety of market and economic conditions.

Read more (open/close details)

By combining Cray's Lustre® parallel file system and the XC supercomputer with DataWarp I/O acceleration, we can massively increase strategy discovery and testing throughput while keeping costs lower than with traditional commodity clusters.

Testing thousands of trading strategies is a very I/O-intensive task. The testing style falls into two categories:

  • The strategies share the same market data - often true for a quick pass analysis of many strategies
  • The strategies are more complex or important and demand their own set of data

In the first case, Cray's DataWarp technology on the XC series supercomputer can share this data at lightning speed across all nodes rather than your needing to pay for SSD storage per node to duplicate data to each node. This achieves the same or better performance, but at far lower cost.

In the second case, fresh data is loaded in per strategy and is extremely I/O bandwidth dependent. Many firms have to cut corners by reducing the number of months of market data they're looking at and/or the time granularity of that data. Quants will quietly tailor the data to the IO capability of the infrastructure. This means strategy confidence is directly proportional to I/O capability.

Eliminate I/O Bottlenecks with the World's Fastest Production File System

The Cray® Sonexion® scale-out Lustre storage system enables smarter trading by eliminating I/O bottlenecks. It operates in parallel, allowing you to load terabytes of data in moments. Cray Sonexion storage enables you to scale I/O performance from gigabytes to over 1 terabyte per second.

Cray systems have been deployed, tested and proven over and over again in the most demanding industries. In fact, Cray Sonexion scale-out Lustre storage holds the current record for the world's fastest production file system. You realize the benefit of this unmatched experience in the customized software, rigorous testing and expert support that means your storage solution is simple to manage, easily scales performance and capacity and protects your valuable data.

Parallel file systems vs. NAS performance

With the Cray Sonexion parallel file system, all you have to think about is maximizing your returns.

Back to top


Quantitative Finance

With greater regulatory compute loads, firms are starting to look at accelerators to flatten the growing cost curve of compute grids. Accelerators are also being used more with deep-learning algorithms for trading strategy discovery.

Read more (open/close details)

Why GPUs for Quantitative Finance

GPUs excel at the types of computational workflows often used by the most demanding quant jobs, including Monte Carlo algorithms, Murex analytics, etc. As parallel processors, GPUs excel at tackling the large amounts of similar data these applications generate because the problem can be split into thousands of pieces and calculated simultaneously.

Reliably Uncompromising GPU Performance

Built upon decades of supercomputing and HPC experience, Cray's CS-Storm family of GPU-accelerated systems are designed to reliably deliver the highest sustained levels of performance in the industry. This is significant for quantitative finance — it ensures they make the most of their investment — but perhaps most significant in that it helps ensure the system they implement will work as expected.

Back to top


Regulatory Compliance

Regulatory compliance is the largest expense for any firm, often involving tens of thousands of people. Thanks to Cray, there is a better way.

Read more (open/close details)

Graph technologies can be used to revolutionize both cost and efficiency. Working with our partners, we have developed a regulatory platform that allows graph to be used at the massive scale you need.

  • A regulation is interpreted and implemented as a set of policies.
  • These policies are used to specify IT applications for regulatory reports and transaction scanning for policy breaches.
  • Breaches are singled out as "exception reports"; analysts review them manually, and find that many of them are false positives.

When regulations change the whole process needs to start again. And many rules refer to other rules, which can mean transaction breaches go several levels deep. That makes rules complex and hard to debug, and breaches harder to prove.

Cray works with technology partner Mphasis NextAngles to represent rules and policies as graph ontologies directly. Transaction scanning can be done directly against this ontology, and the investigation platform can be far more efficient, representing very clearly how a complex breach occurred and automatically bringing all the data of that breach together to greatly speed investigation. If rules and policies change, it is a simple matter to change the ontology so that the system is immediately compliant.

In today's world of fast regulatory change and margin compression it is this kind of revolutionary breakthrough that can make you stand out against your competition.

Back to top


Financial Crime

What do insurance fraud, insider trading and money laundering have in common? Detecting them is often akin to finding a needle in a haystack — or even a needle in a needlestack. You collect and sort through massive amounts of irrelevant data and false-positive alarms to identify anomalies that can lead to bad actors.

Read more (open/close details)

Traditionally firms have relied on rule-based systems to spot problems. This has led to a number of problems, such as:

  • These systems tend to throw up a lot of false positives, requiring extra staffing to wade through all the exceptions, and reducing customer satisfaction when transactions are denied.
  • The systems are often batch based, so problems are handled after the damage has been done.
  • As attacks morph and become more sophisticated, it is hard to update the rule models.
  • As financial crimes evolve, analysts rarely know what to look for.
  • Massive volumes of data, and false-positive alarms, make identifying anomalies a challenge.
  • Analysts struggle to determine relationships between entities and datasets from various systems.
  • Analyzing data in siloed subgroups is time consuming and often unproductive.

A Platform for Detecting and Investigating Anomalies at Lightning Speed

Firms are now looking to graph technology to update old rules-based systems, because graphs are a more natural representation of complex relationships such as those leading to fraud. Graph-based systems are about twice as good at reducing false positives than rules-based systems. Because graph technology is all in-memory, traditional hardware architectures constrain good performance to models held within a single compute node. But Cray allows you to remove this constraint and fully exploit the power of graph technology.

  • Discover bad actors, instead of searching for them – Upon ingesting data from multiple sources (databases, logs, network analysis, etc.), the Urika platform explicitly captures and updates all of the relationships contained within the data. This eliminates the need to define or maintain relationships between entities, thereby ensuring that anomalies or suspicious activities are flagged, even if analysts don't know what they're looking for.
  • Ad-hoc search capabilities – The search for evolving methods of financial crime requires the ability to perform ad-hoc searches for patterns of relationships, for which graph databases are ideally suited. The ability to identify the nature of these relationships across massive datasets without portioning extends analysts' ability to identify and find bad actors.
  • Real-time data analysis – Leveraging Cray's decades of supercomputing and HPC expertise, the Urika platform enables a data analyst to iteratively explore and isolate suspicious activity in real time, while lowering incidences of false positives. This helps analysts work more efficiently, and enables them to discover malicious activity in time to minimize their impact.

Software partners approach Cray because they understand that anomaly detection is easy with graph but tough to do at scale. With the Urika platform, financial crime anomalies can be spotted rapidly and investigated at lightning speed, transforming your defensive capability.

Back to top


Cybersecurity

Regardless of the size or technical maturity of your business, your organization is likely to fall victim to a cyberattack.

Read more (open/close details)

Challenges Associated with Avoiding or Mitigating Attacks

  • Analysts rarely know which queries will find the right answers.
  • Massive volumes of data make identifying malicious activity difficult.
  • Analysts struggle to determine relationships between diverse datasets.
  • Processing data in subgroups is time consuming and often unproductive.

A Platform for Detecting and Investigating Anomalies at Lightning Speed

  • Discover threats, instead of searching for them – Upon ingesting data from multiple sources (logs, network analysis, etc.), the Urika platform explicitly captures and updates all of the relationships contained within the data. This eliminates the need to define or maintain relationships between entities, thereby ensuring that anomalies or suspicious activities are flagged, even if analysts don't know what they're looking for.
  • Ad-hoc search capabilities – The search for new cyberattacks requires the ability to perform ad-hoc searches for patterns of relationships, for which graph databases are ideally suited. The ability to identify the nature of these relationships across massive datasets without portioning extends analysts' abilities.
  • Real-time data analysis – Leveraging decades of supercomputing and HPC expertise, the Urika platform enables a data analyst to iteratively explore and isolate suspicious activity in real time. This helps analysts work more efficiently, and enables them to discover malicious activity in time to minimize their impact.

Software partners approach Cray because they understand that anomaly detection is easy with graph but tough to do at scale. With the Urika platform, cybersecurity anomalies can be spotted rapidly and investigated at lightning speed, transforming your defensive capability.

Back to top


Solution Briefs, White Papers & Reports

Cray Gets Top Marks for Financial Services Analytics Benchmark

In this benchmark report from the Securities Technology Analysis Center (STAC), the Cray XC40 Intel Xeon Phi system outperformed all previous results for a benchmark that tests technology stacks used for compute-intensive analytic workloads used in pricing and risk management.

The Future of Grid in a Cloud-Enabled World

This TABB Group research report considers grid’s place in a cloud-enabled world, addresses options for optimizing onsite system performance, and discusses total cost of ownership (TCO) considerations for on-site grid components.

Boost Confidence in Algorithmic Trading

Trade smarter with Cray. Balanced systems for computing, analytics and storage scale efficiently as your data — and trading opportunities — continue to grow.

Financial Crime: Fraud Detection, Anti-Money Laundering, Employee Surveillance and Cybersecurity

With the rapid increase in the volume and sophistication of financial crime, financial services firms can no longer afford to rely on yesterday’s technology.

Cray Urika-GX Agile Analytics Platform for Financial Services

A powerful platform for trading strategy development, risk analytics, compliance and fraud detection

Future Grid: Accelerating Risk Analytics While Reducing TCO

New pressures in today’s fast-moving capital markets are forcing banks to rethink their risk management systems architecture.

White Paper: Credit Valuation Adjustment

How to Solve the Computational Challenges of Credit Valuation Adjustment While Reducing Grid TCO

Win by Trading Smarter with the Cray Sonexion Parallel File System

Hedge funds and quant trading firms compete to a large extent based on the quality of their analytics. Success is determined by two things: strategy innovation and strategy confidence.

Videos & Webinars

The Future of Grid in a Cloud-Enabled World

The TABB Group explains how capital market firms are challenging their existing computing architectures, with the goal of increasing flexibility and performance while reducing total cost of ownership (TCO).

How to Take Advantage of Big, Fast Data

Agile analytics can change the financial services playing field. By applying advanced analytics enabled by the Cray Urika-GX platform, firms can reveal connections in big data – connections that lead to key insights and better, faster decision making.

Credit Valuation Adjustment

Staying competitive in financial services today requires more. Credit valuation adjustment (CVA) has emerged as a critical component to evaluating risk.