The technology infrastructure underpinning central counterparties (CCPs) is being reinforced or replaced to support a jump in workload. CCPs are now building the capacity to process approximately seven million trades a day per asset class, up from 2000-3000 a day currently, according to Kevin McPartland, senior analyst at research firm TABB Group.
The increase in the volumes will stem from regulations being put into place across the Group of 20 countries (G20), such as the Dodd –Frank Act in the US and the European markets and infrastructure regulation (EMIR), which mandate central clearing of trades for standardised OTC derivatives contracts.
In Asia, Hong Kong Exchanges and Clearing, the exchange and infrastructure operator for the special administrative region, has plans to launch an OTC clearing facility in late 2012, while Singapore Exchange (SGX) set up its clearing facilities in November 2010.
“There is clearly quite a race on from a CCP point of view to implement the systems that are required,” says Keith Bear, worldwide sales leader for financial markets at IT provider IBM.
Tabb estimates that CCPs have increased spend on IT by 9.6% compound annual growth rate from 2010 to 2011, taking it to a total of US$293 million globally, with the US accounting for US$119 million, Europe for US$118 million and Asia for US$56 million. Of this, 44% will be spent on external IT resources.
Some parts of the CCPs’ operations require greater investment than others. The technology stack within CCPs can be roughly divided into systems that deal with trade capture, margin calculation, risk management and settlement.
The framework
The greatest challenge for the trade capture element, which involves the storage and retrieval of large amounts of complex data, is the ability to scale reliably.
“In the early days of migration from the floor to electronic trading we first encountered the need to look at our technology stack and its capacity – we saw 30% CAGR in the first five years,” says David Hoag, managing director, clearing technology at CME Group. “The theme we developed was horizontal scalability in our systems.”
The technology to provide this is relatively commoditised. Servers can be clustered together to provide a distributed computing model, which allows a database to be run across multiple servers. More can be added to increase the database’s capacity, to replace failed hardware or improve system performance with minimal effort and cost. For example CCP ICE Clear uses Intel-based Linux servers running database supplier Oracle‘s Real Application Clusters (RAC) with the Oracle Database 11g Enterprise Edition.
ICE Clear can add an RAC node or swap out existing hardware for newer, faster, commodity hardware from HP, DELL, or IBM without disrupting its systems. HP’s 3Par storage array is used to manages the workload distribution across the available computing resources.
Once the infrastructure is able to handle the load required for managing the data, the firm’s calculation engines need to operate at the same speed.
“Trade management is pretty well covered in most markets,” says Per-Anders Hall-Bedman, deputy chief executive of technology provider Cinnober.
“But on the risk management and trade processing side, there is a lot to be done. Traditionally these areas have dealt with standardised products, not OTC products. Where CCPs are dealing with standardised products they can have standardised formulas for calculating risk. That is still a problem.”
Crunching the numbers
The challenge for the clearing houses is the need to deliver real-time risk and margin calculations. The ability for market participants to understand their position and collateral requirements in real-time becomes a necessity under the new rules if risk is to be properly managed, but the existing systems within CCPs tend to make calculations over larger time periods or are batch driven; very few are dealing with risk management close to or in real time. The calculation complexity depends upon the assets in question, says McPartland.
“The margin calculations on relatively illiquid CDSs are much more complex mathematically than for IRS,” he says. “Valuing IRS you are basically looking at the cash flow of the yield curve... for CDS you are predicting the likelihood of default of the underlying, so the calculations become much more complicated.”
Some CCPs will be able to adapt their existing internal systems to cope. ICE Clear Europe and ICE Clear Credit, which clear OTC CDSs and energy derivatives, already offer intraday risk management providing the clearing house and all clearing members with trade, position, profit and loss and margin reports every five minutes, although internally within ICE this is available every one minute.
A spokeswoman for ICE Clear said that the rising demand for central clearing by market participants has already been reflected in the growth of ICE's cleared OTC energy volume, which has increased from 2% in 2002 to 80% in 2007 to 96% today, well ahead of the enactment of financial reform in the US and EU.
“We have a group of very talented programmers, database administers, and system administrators,” she said. “We optimise the system from application design, implementation, database optimisation and system optimisation. Because of that, the risk system as a whole works very efficiently and achieves the highest level of operation excellence."
Hoag says CME uses a distributed computing model, with servers connected to function as a grid that can combine the computing power of several servers into a single resource to process complex calculations. This provides the flexibility to cope with increased processing tasks as well as storage and retrieval.
“Our stack allows us to partition message processing and data processing, so we can distribute that load across multiple servers and aggregate it when it comes back to core processing,” Hoag says. “That model works for trade capture, it works for running risk and margin analysis and it works for settlement, the latter being simpler as the limited number of settlement banks reduces the scale challenges.”
On the market
There are also commercial technologies that will allow a CCP moving to develop its capabilities in this area quickly.
In April, LCH.Clearnet announced its SwapClear platform, which clears interest rate swaps, had adopted system provider Murex’s MX3 risk and collateral management platform to allow it to handle 25,000 IRS trade sides per hour and a total portfolio of 4 million trade sides, doubling SwapClear’s previous capacity. It has been designed to scale to 10 million trade sides in anticipation of global market developments in centralised clearing.
“This is where we see the applicability of more appliance-based database technology such as Netezza,” says Bear. “It bundles together an integrated database with the server and the storage, to achieve performance of ten to 100 times faster than what has traditionally been gained which is required as we get into these hugely data intensive calculations.”
Cinnober’s Hall-Bedman says: “We've done a lot of testing in that space, it's perfectly possible to handle very high transaction load in close to real time but the more complex the instrument gets, the more complex the data and the model gets.” However, he adds that the issue of valuing some instruments stretches beyond simple number crunching.
“You end up with a number of fundamental questions in terms of how you really evaluate some instruments. Look at the CDS market; after the Lehman crash it died, in that you had no market prices, and you very quickly realise that there are some challenges in valuating risk for CDS in the proper way,” he says.