Progress - past, present and future

Best Execution Guide to Low Latency 2010 33
LOWER LATENCY TRADING
So how has this market matured
over the last 2 years? We know that
certain technology elements are
now playing best on the race circuit.
Like the Formula 1 car, a complete
solution stack needs to be tuned to
interoperate at maximum efficiency –
beginning with the latest multi-core
x86 industry standard microprocessor
and a tuned operating environment in
power-efficient hardware blades where
horsepower can be dialled up on
demand – all networked via advanced
non-blocking and optical techniques
light years removed from a connecting
copper rod.
This is computer science at its leading
edge – and the parties behind this
paper – Hewlett-Packard, Intel
Corporation and BT all have synergistic
expertise in this space which is playing
efficiently together. A quick walk
through the major considerations
across the landscape is valid – before
unprecedented deep dive hands-
on testing and simulation of the
interdependencies in lab and then
production environments
At the heart of trading is market
data, and hence the ability to move,
handle and manipulate it is a basic
foundation. Data Fabrics and message
bus technologies have come to the
fore in this space as the backbone of
trading solutions – with NYSE-Euronext
(Wombat Data Fabric), Thomson Reuters
(RMDS and TRADS) and 29 West (LBM)
leading the charge at present – joined
by interesting innovation from the bank
driven open source initiative of AMQP
by Rabbit MQ, iMatix, Red Hat and
29West. The messaging fabric is the
vehicle which takes market data from
the point where it reaches the firm
from the publisher (exchange, venue
etc) and makes the price information
ubiquitously available to consuming
analytical applications. Ideal performance
is blistering speed and no variation in
reliability of delivery – “jitter”.
Real time operating systems have
become a new option in this battle
for competitive advantage. As the
controller of the overall architecture,
the OS has to be at least as capable
as all other elements – otherwise it
is an obvious bottleneck. The real
50 msgs/sec
100 msgs/sec
150 msgs/sec
200 msgs/sec
IO Acceleration technology reduces latency
FIX to XML, BizTalk routing, XML to FIX
Message volume
200
150
100
50
0
Latency mSec
Latency improvement
without I/O Acceleration technology
with I/O Acceleration technology