CCR accelerates computations by a factor of 900

Improving the modeling of natural disasters with a solution from IBM

Published on 07-Nov-2012

"The IBM solution is helping us and our clients to better assess the potential financial consequences associated with natural disasters and thereby to better manage the insurance risks." - Antoine Quantin, Director of Modeling, Caisse Centrale de Réassurance

Customer:
Caisse Centrale de Réassurance (CCR)

Industry:
Insurance

Deployment country:
France

Solution:
Business Resiliency, Information Integration, Virtualization, Virtualization - Storage

Overview

CCR (Caisse Centrale de Réassurance) was founded in 1946, and is one of the top 25 global reinsurers. Wholly owned by the French state, CCR employs 252 people and turned over €1.385 billion in 2011. Backed by a state guarantee, the company can offer unlimited insurance cover for certain types of risk specific to the French market, including natural disasters, acts of terrorism and the reinsurance of ships and nuclear sites.

Business need:
Aiming to model the financial impact of up to 10,000 simulated natural disasters each month, CCR required a powerful yet efficient computational cluster that would also be scalable and easy to manage.

Solution:
Worked with IBM to redesign its data center and deploy a cluster of 40 IBM® System x® 3550 class servers, providing more than 400 computational cores, with an IBM Storwize® V7000 virtualized disk array.

Benefits:
Offers 30 times increase in raw computational speed, combined with optimized software to deliver a 900 times improvement; powers a web-based flood-risk map for insurers, providing accelerated insight.

Case Study

To read a French version of this case study, click here

CCR (Caisse Centrale de Réassurance) was founded in 1946, and is one of the top 25 global reinsurers. Wholly owned by the French state, CCR employs 252 people and turned over €1.385 billion in 2011. Backed by a state guarantee, the company can offer unlimited insurance cover for certain types of risk specific to the French market, including natural disasters, acts of terrorism and the reinsurance of ships and nuclear sites.

As a reinsurer for potentially unlimited risks, it is vital for CCR to be able to understand and model these risks. The company uses both deterministic and probabilistic approaches to create models of the likely cost impact of natural disasters such as floods. In simple terms, the deterministic approach consists in simulating the impact of a given scenario, while the probabilistic approach depends on a very large number of possible scenarios, each one being characterized by its probability.

CCR’s existing computational platform was no longer meeting the company’s needs in terms of performance, as Antoine Quantin, Director of Modeling at CCR, explains: “For the deterministic approach, we take a single event such as a flood, build a physical model, then intersect that with the insurance portfolios to see the potential financial losses for CCR and our clients – the insurance companies.

“With the existing hardware and software, we had the capacity to model at most 10 or 20 events a year. We also wanted to extend our use of probabilistic modeling, which required us to simulate up to 10,000 events each month. Our major challenge was to be capable of going from a handful of simulations to modeling tens of thousands of events.”

Balanced solution
For both deterministic and probabilistic approaches, CCR uses a topographical model of France with a 50-meter mesh – meaning that the model is based on the elevation above sea level for a set of points at 50-meter intervals across the entire country. To model a flood, CCR takes several different sets of rainfall data from the French meteorological office, then simulates the flow of water across the model and its impact on river levels. The portfolios of insured properties can then be superimposed to reveal the height of the flood water at that location, ultimately allowing an estimate of the cost of damages.

“The probabilistic approach is exactly the same, but instead of modeling a single event that affects one region, we try to build what we call a catalogue of events,” says Antoine Quantin. “It can never be perfect, but we try to cover all of the possible cases of natural disaster in France: a very large number of scenarios. With the old servers, it would have taken several years to run the models that we wanted to run.”

Following a request for proposals, CCR compared offers from two short-listed companies before choosing IBM to provide its new computational platform. “We had been approaching the limits of the air-conditioning capacity in our data center, so IBM included a study into the physical arrangement of our servers,” says Antoine Quantin. “As a result, we were able to lower the temperature by a couple of degrees so that we could accommodate the new servers.”

He adds, “The solution proposed by IBM met our requirements around performance and energy consumption, and it also represented a good balance between price and performance.”

Rapid processing
The heart of the new supercomputer at CCR is a cluster of 40 IBM System x3550 class servers, each with two 6-core Intel Xeon 5690 processors and 48 GB of memory. The Xeon Processor 5600 Series offers industry-leading performance and energy-efficiency with the ability to intelligently adapt performance to diverse needs. An IBM Storwize V7000 disk arrays provides virtualized storage for results, and the servers are networked using dedicated 10 Gb/s Ethernet. In addition to deploying the IBM servers, CCR worked to optimize its Matlab codes.

“We tried to make gains on both sides: by optimizing the codes and by increasing the computational power,” explains Antoine Quantin. “In each case, we improved by a factor of about 30, so that the simulations run up to 900 times faster overall. Our original goal was to be able to simulate 10,000 events in a single month, and we’re already at about 8,000 in fifteen days, with further optimization to come. Without the new IBM System x solution, we’d have no hope of achieving this objective.”

Understanding risk
The tremendous increase in performance provided by the IBM solution is helping CCR to improve the quantity and quality of risk information. Internally, this helps the company – and, as the insurer of last resort, the French state – understand its financial exposure. It also allows CCR to provide better information to its external clients, the insurance companies.

The IBM solution has also enabled the creation of a new service from CCR: an online probabilistic map for flooding. Recently launched and available exclusively to CCR clients, the map provides a quantified estimate of the risk of flooding at any location across the entire French territory.

“As far as we know, this is the first online solution of its kind in France,” says Antoine Quantin. “It’s based on information that we simply couldn’t produce in the past, and we think it’s going to help the insurers to manage their portfolios much better.”

The IBM System x landscape is also proving to be very reliable, running week-long computational runs without any problems. The solution can easily be scaled up to cope with increased demand in the future, which will allow CCR to achieve its goal of increasing the resolution of its terrain mesh from 50 meters to 25 or even 5 meters. The company may also extend the scope of the solution to cover the risk of drought.

“We have ample performance to meet our current requirements, and this solution can be extended in line with our future plans,” says Antoine Quantin. “The IBM solution is helping us and our clients to better assess the potential financial consequences associated with natural disasters and thereby to better manage the insurance risks. Even though the new computational cluster is much larger and more powerful than the one it replaced, it is more efficient and easier to manage – enabling us to focus on higher-level tasks.”

Products and services used

IBM products and services that were used in this case study.

Hardware:
Storage, Storage: Storwize V7000, System x, System x: System x3550 M3

Service:
GTS Technical Support Services: Hardware Maintenance

Legal Information

© Copyright IBM Corporation 2012 Compagnie IBM France 17 avenue de l’Europe 92275 Bois-Colombes Cedex Produced in France November 2012 IBM, the IBM logo, ibm.com, Storwize and System x are trademarks of International Business Machines Corporation, registered in many jurisdictions worldwide. A current list of other IBM trademarks is available on the Web at “Copyright and trademark information” at: ibm.com/legal/copytrade.shtml. Intel, Intel logo, Intel Inside, Intel Inside logo, Intel Centrino, Intel Centrino logo, Celeron, Intel Xeon, Intel SpeedStep, Itanium, and Pentium are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. Microsoft, Windows, Windows NT, and the Windows logo are trademarks of Microsoft Corporation in the United States, other countries, or both. Other company, product and service names may be trademarks, or service marks of others. References in this publication to IBM products, programs or services do not imply that IBM intends to make these available in all countries in which IBM operates. Any reference to an IBM product, program or service is not intended to imply that only IBM’s product, program or service may be used. Any functionally equivalent product, program or service may be used instead. All customer examples cited represent how some customers have used IBM products and the results they may have achieved. Actual environmental costs and performance characteristics will vary depending on individual customer configurations and conditions. IBM hardware products are manufactured from new parts, or new and used parts. In some cases, the hardware product may not be new and may have been previously installed. Regardless, IBM warranty terms apply. This publication is for general guidance only. Photographs may show design models.