According to the New York Stock Exchange, program trading encompasses a wide range of portfolio-trading strategies involving the purchase or sale of a basket of at least 15 stocks with a total value of $1 million or more. As electronic trading has made it easier to execute program trading strategies, the trading of large baskets of stocks in a coordinated way has become popular for both buy-side and sell-side firms.
- Brokers and buy-side institutions use algorithmic trading to break large orders into smaller chunks to avoid moving the market.
- In an effort to better measure trade execution costs in the "post soft dollar" world, buy-side firms look to measure performance against such benchmarks as value weighted average price (VWAP), time weighted average price (TWAP), and target volume (TVOL). With an emphasis on fast trade execution, these benchmarks often drive buy-side firms and their brokers to utilize program trading of baskets of stocks in order to stay close to the chosen performance benchmark.
- A typical large sell-side firm executes one to two million stock trades each day, and some trading strategies may take multiple days. Sell-side firms need the ability to efficiently store the necessary real-time and historical order and market data to monitor and execute their program trading strategies.
According to Financial Insights, an IDC Company, "The average, large sell-side firm receives about 250,000 equity orders, executes between one and two million trades, and captures about 5 to 10 gigabytes of equity pricing data during each six-hour trading day. Most equity traders require real-time monitoring capability on all outstanding customer orders and at least five days worth of market data, while program traders may require three months or more of market data. The existing data storage challenge will increase by a large order of magnitude over the next few years as program trading continues to increase trading volumes and as depth-of-book pricing data such as Nasdaq's SuperMontage screens and the NYSE's Open Book become more widely used."
Limitations of existing technology
According to Financial Insights, an IDC Company, "Program trading systems and order management applications cannot cache large amounts of real-time data without impeding performance. Relational databases and online analytical processing (OLAP) systems can allow traders to store and analyze large amounts of data, but these often degrade in performance when forced to update continuously. In addition, relational database and OLAP technologies make it difficult to implement new kinds of queries to analyze and act on the data."
Because of the performance limitations of relational database and OLAP technology, sell-side firms have built custom memory cached solutions to be able to meet the execution performance and scalability requirements of Program/Algorithmic Trading. Maintaining custom memory mapped solutions is costly, but has historically been seen as necessary to achieve performance requirements.
In addition to being costly to maintain, the nature of memory cached solutions typically results in a vulnerability to hardware failures. Transactions, if data is not persistent, are not immediately recoverable -- leaving firms open to significant liability if failures occur. In a fast moving market, the recovery time is critical when large volumes of program trades are being executed. Using a persistent database for program trading solutions drastically reduces the potential number of "lost" transactions due to a failure and dramatically shortens recovery time. The caveat is that the database must be capable of providing the necessary transactional performance.
InterSystems' Cache for Program/Algorithmic Trading
InterSystems' post-relational database, Cache, is uniquely qualified to provide algorithmic and program trading solutions with the transactional performance they require, along with persistence for instant recoverability in the event of failure. Cache's efficient multidimensional data structures allow applications to store and retrieve information without the processing overhead required to break it up into a series of relational tables. In addition, Cache's unique Enterprise Cache Protocol dramatically decreases network traffic between an application's business logic and the database.
The result? Cache demonstrates transactional performance that rivals or surpasses what can be achieved by using in-memory databases. But as a persistent data store, Cache also provides high reliability and low maintenance costs to algorithmic/program trading applications.
In a recent research paper, Financial Insights, an IDC Company, identified 5 Key Technology Building Blocks for Program Trading.
To execute program trading strategies, sell-side firms and their clients need the following types of technology:
- Connections to order management systems that buy-side firms can use to submit lists of stocks.
- Algorithms for optimizing the trading strategy.
- A facility for storing real-time and historical order and pricing data.
- A trade execution engine for executing trades at the best prices.
- An integration hub for coordinating trading strategies across trading desks and middle- and back-office systems.
One Memorial Drive
Cambridge, MA 02142
Contact: Rich Perry,
Vice President and General Manager, Financial Services Sector