Cost of latency – choosing wisely

Trading System Latency

Latency in trading systems has always been a design/cost consideration however, in today’s global trading markets, the focus on latency has never been greater.  Banks and traditional exchanges have seen their market share threatened by new entries into the marketplace, namely hedge funds and alternative trading venues with their automated trading systems.  The rise of multi-asset trading and the increased focus on risk management has also placed a greater emphasis on developing trading systems that can keep up with these ever increasing trading volumes.

Hardware-based Messaging

From this reality, a new breed of messaging technology has emerged.  Middleware messaging has morphed into “messaging appliances”. These hardware-based messaging devices provide lower latency throughput
than software-only solutions and are the platform of choice for those high demanding, high-volume trading applications.

A “keep up with Joneses” mentality has permeated Wall Street. After all, who wants to see their market share erode as a result of under-performing trading systems.  This has led to a modern day “arms race”, at times, with little regard for the long term costs of maintaining such systems.

Focus of TCO

The flip side to an under-performing system is an over engineered system.  A system that is more expensive and complex than it needs to be. Understanding the total-cost-of-ownership (TCO) to develop, deploy and maintain a global trading system should be in the forefront of any technology architect or trading desk head.  Knowing your business, accurately forecasting message rates, and designing a system that scales easily as your business grows are the key challenges faced by today’s IT experts.

For example, the TCO of hardware-based messaging appliances will usually be higher than a software-only solution.  When you take into consideration the additional costs of power consumption, rack space, cooling requirements, and installation of these appliances, the added latency gains need to be measured against these additional expenses.

The value of messaging appliances is further reduced if the recipients of the data are not in a position to take advantage of this lower, more deterministic latency. Unless we are talking about co-located automated trading solutions, it is likely that the added costs associated with these devices will not be justified.  When the internet is the medium for data delivery, the value proposition of these devices will certainly be negated.

There is a place for both software and hardware-based messaging – understanding the business needs and picking the right tool for the job is the interesting challenge we face.

2 Responses

  1. Interesting topic jcaldazo,

    I agree with your assessment that performance is one of the reasons behind the growth in messaging appliances, however, lower cost and turnkey simplicity also key reasons. Regarding power consumption, network processors and FPGAs generate higher performance at lower clock rates thus the power cost is actually far lower (measured by msgs/sec/watt). A single appliance scales to the equivalent of many servers running software, so likewise, datacenter footprint is also lower.

    Lots of details on appliances vs software on servers are available here:

  2. Oops, pressed post too fast: lower cost and turnkey simplicity *are* also key reasons…

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: