Finextra, Tullett Prebon Files SEF application with CFTC, here.
The SEF, registered as tpSEF Inc. and headquartered in New Jersey, is a wholly owned subsidiary of Tullett Prebon. It has been established to ensure the Company’s compliance with Dodd-Frank legislation, enacted on July 21, 2010.
Shawn Bernardo, Tullett Prebon’s Senior Managing Director of e-broking and member of Tullett Prebon’s North American Executive Committee and Chairman of the Wholesale Markets Brokers’ Association (WMBA), is named Chief Executive Officer of the SEF.
DynaRack assumes there is an advantage to building an application specific rack of hardware to use in a colo. What features does your application need to have for DynaRack to breakeven or show a profit. It’s a lot of work to build a motherboard, test parts, get the right power supplies, assemble the rack test it and roll to production. There better be some significant upside to justify picking up all this work normally just left to Dell or HP.
To the first approximation, for DynaRack to breakeven your app needs to have several million floating point operations to execute between the time new market data is pulled off the wire and the time the first new order or cancel, dependent on that specific market data is issued to the exchange matching engine. The more millions of floating point executions in this interval the better the performance advantage for DynaRack over a serial code implementing the identical computation. It isn’t quite true that any collection of millions of floating point executions are sufficient for DynaRack to break even. Additionally these floating point operations need to be somewhat independent so that they can be executed in parallel. Also these Flops need to generate sufficient alpha or we never get to breakeven.
Broadly the argument is going to be if your competitor needs 100 millis to execute the required floating point ops and with a faster microprocessor you can retire the same floating point operations in under 1 milli with a sufficient alpha level you break even with DynaRack. We will fill out the argument and outline the obvious constraints. m2m for a equity option via Black Scholes closed form expression valuation is about 30 nanoseconds and a CDS m2m is about a millisecond in the worst case ( maybe 140 mics in the average case) discounting off a cooked credit curve. I expect an interest rate swap m2m would be a shade less that a CDS m2m but more than a Black Scholes evaluation. Given sufficient alpha, you can see why the DynaRack breakeven is easier with Fixed income products with a non trivial term structure.