Understanding Latency and Its Impact to Trading Profitability
Recent events has significantly increased the level of volatility in capital markets, as witnessed by the historic drop in West Texas Intermediate oil futures. On Monday, March 23, NYSE leadership indefinitely closed the physical trading floor after multiple positive tests for infection, but there has been a historic increase in electronic trading. According to Options Clearing Corporation (OCC), more than 48 million options traded on Friday, Feb. 28, setting a record for total trading volume in a single day. The peak daily messages in the Options Price Reporting Authority (OPRA) feed reached a record-breaking high of 102B in February 2020 — more than triple that of February 2019 (31.4B). This is putting significant strain on global financial network infrastructures.
When it comes to electronic trading in highly fluid markets, access to real-time and accurate market data is critical, and nanoseconds count. The speed and sequence with which market participants place and execute orders at the matching engine of a given exchange venue depend on many technical variables. One of these variables is the latency in accessing real-time market data. There are several methodologies and places to measure latency, which will be discussed further in a later post. For the purpose of this article, latency is defined as being any delay or lapse of time between a request and a response. As it pertains to trading, latency directly influences the amount of time it takes for a trader to interact with the market. The timely reception of pertinent market information and the speed for market participants to act upon its receipt will gain them a competitive advantage or disadvantage. Further illustrated below is the latency footprint recently published by ICE. The diagram shows latencies between ICE Global Network consolidated feed ticker plants across major sites. Note that latency and contributing factors to market data quality vary considerably depending on the transport mechanisms and circuit provisioning employed at and between network sites.
ICE global network latencies (source: Intercontinental Exchange, see https://www.theice.com/
To better relate the issues of latency within a trading system it is important to understand how issues may occur and propagate within a system. When it comes to designing and operating an exchange network the need for near-instant identification of the cause of data errors and contributing factors is paramount. Take the example below with the traffic levels on the Arcabook Channel 2 on the 1st graph. Burst traffic during a certain period increased from around an average of 275Mbps (shown by the purple line) to around 450Mbps at around 11:44 AM. The latency (right hand graph) at the same time – one-way latency (OWL) jumped from 7,495 uSec to a peak of 7,510 uSec (15 uSec increase) and then settled back to 7,500 uSec (5 uSec). Clearly a close correlation. Was this caused by network congestion or something within the ticker plant/matching engine?
It is important to identify the potential issues within the system, as shown in the example above. While working with exchanges and market participants, Keysight Technologies met with IT teams focused on network infrastructure and operations to discuss such issues. For market data performance analytics, these teams articulated the clear need for near-instant identification of the root cause of data errors, as well as the contributing factors. Is the reported sequence gap on a market data feed linked to the exchange, the data distribution network, or the sell-side network where the feed handlers are? What went wrong and why? Where should these teams take measurements in the global network infrastructure, and how?
Some of the potential considerations are:
- Specifying latency and loss objectives in conjunction with bandwidth requirements for peak utilization periods;
- Continuously monitoring as market volatility conditions change rapidly, as volatility impacts average and peak message rates, causing higher than specified or expected levels of latency and packet loss because of microbursts;
- Monitoring infrastructure in place with trigger-based alert notifications is essential for proactive issue identification and problem resolution;
- Understanding and continuously verifying the global latency footprint from the exchange to remote colocation sites throughout network infrastructure, and between data centers or colocation sites that consume real-time market data;
- And, ensuring that transaction analytics tools provide detailed visibility into quality events.
These considerations will be further discussed in a following post, in addition to solutions for market data quality issues further exasperated by current market volatility. Also, for a more in-depth look at TradeVision check out recently published white paper Understanding Latency and Its Impact on Trading Profitability.