« The Debate on Dark Pools | Main | Rumblings in the Cloud »

Monday, February 22, 2010

Peas and Carrots

Posted by Louis Lovas

In the words of the auspicious Forrest Gump some things go together like peas and carrots. Truer words were never spoken. Some things just do go together well, sometimes by design, often by accident. I don't think anyone actually planned milk and cookies or popcorn at the movies but nonetheless these things are made for each other.  When it comes to technology the same harmonious relationships exist.

In the recent Aite report on High Performance Databases (HPDB),  the market for specialized databases is surveyed along with a handful of vendors in this space.  This is a cottage industry where the big database vendors don't play. It's hard to imagine in this day and age where database technology is so standardized and mature and a multitude of choice abounds from commercial products to open source that any other database technology and a gang of vendors would have a chance. Yet it is happening and it's thriving.  

I believe it has to do with a synergistic relationship to event processing. If CEP is the "peas" then HPDB's are the "carrots". These two technologies share two fundamental precepts:

  •  A focus on Extreme Performance
  •  Temporal Awareness

I. Extreme Performance, Speeds and Feeds
These HPDB's which are often referred to as Tick databases, are found in the same playground as event processing technologies. In the Capital Markets industry they connect to the same market sources, consume the same data feeds. Both technologies are designed to leverage modern multi-core hardware to consume the ever-increasing firehose of data. By the same token, once that data is stored on disk, database query performance is equally important.  The massive amount of data collected and is only as good as the database's ability to query it efficiently thus creating another (historical) firehose of data which an event processing engine would be the consummate consumer.  

II. Temporal Awareness, when is the data
Time is a basic principle in event processing technology, applications typically have as a premise to analyze data-in-motion within a window of time. HPDB's design center is to store and query time series data. Some of the database vendors even bring time to a higher level business function. They understand the notion of a business Calendar, knowing business hours, business week, holidays, trading hours, etc.  Imagine the simplicity of a query where you want 'Business hours Mon-Fri for the month of February' and the database itself would know the third Monday was Presidents Day, skipping over that, thus preventing analytic calculations from skewing erroneously.

Leveraging the Synergy
These two fundamental shared principles provide the basis for a unique set of business cases that are only realized by leveraging Event Processing platforms and High Performance Databases

  • Back testing algorithms across massive volumes of historical data compressing time
What if you could test new trading algorithms against the last 6 months or 1 - 2 years of historical market data but run that test in a matter of minutes? What if you could be assured that the temporal conditions of the strategies (i.e. timed limit orders) behaved correctly and deterministically matching the movement of time in complete synchronicity with the historical data? These are just a few of the characteristics that define the harmony between event processing and high performance (Tick) databases.
  • Blending live and historical data in real-time
Querying historical data in-flight to obtain volume curves, moving averages, the latest VWAP and other analytics calculations are possible with these high performance databases. Leading edge trading algorithms are blending a historical context with the live market and even News. The winners will be those that can build these complex algo's and maintain ultra low-latency.
  • Pre-Trade Risk Management
Managing positions, order limits and exposure is necessary, doing it in real-time to manage market risk is a mandate.  In addition to market data, these high performance databases can store pre and post trade activity to complement event-based trading systems and become the basis for trade reporting systems.

In the Trading LifeCycle, Event Processing and High Performance databases are partner technologies harmoniously bound together to form a union where the whole is greater than the sum of the parts. They are the peas and carrots that together create a host of real-world use-cases that would not be possible as individual technologies.

Myself along with my colleague Dan Hubsher we are doing a 3-part Webinar series entitled "Concept to Profit". The focus is on event processing in the trade lifecycle, but we include cases that touch upon high performance databases. You can still register for part 2: Building Trading Strategies in the Apama WorkBench where I will focus on the tools for strategy development aimed at the IT developer.

Once again thanks for reading, you can follow me at twitter, here.
Louie

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d83452154069e201310f296690970c

Listed below are links to weblogs that reference Peas and Carrots:

Comments

forex trading journal

There are lots of industries working all over the world but good to see there are lot of blogs also who are sharing industrial information. Great job keep this work up!

The comments to this entry are closed.

<-- end entry-individual -->