The challenge when measuring transaction and market data latency is to do so without impacting upon the latency, which applies especially in ultra-low latency and HFT environments. The Corvil appliance allows passive on-the-wire monitoring and provides latency root cause analysis with a high degree of accuracy.
The Corvil Appliance uses FPGA technology to capture and timestamp the data on the wire. Corvil use their own proprietary precision time protocol service to synchronise timestamps between appliances with sub-microsecond accuracy, depending on the distance between appliances.
Data can be acquired from the wire using mirroring, tapping and aggregation techniques allowing the data to be captured in raw packet capture format as well as analysed in multiple domains. The data is written to disc then decoded normalised, persisted, analysed and correlated before being written to the database for reporting purposes.
Measurement and analysis capability
The Corvil appliance supports most commonly used protocols including FIX, FAST, TIB-RV, LBM, Solace, EMS, MQ, and NASDAQ.
Once the events have been recorded, Corvil uses the data to produce a production profile of the monitored system.
A typical report would show the distribution of recorded latency and will identify outliers for attention. The exercise of visualizing the message flows through the system provides a more holistic view of performance management which helps to identify where the biggest gains can be made for the budget available, leading to a quick and pragmatic approach to where improvements can be achieved.
Persisted and normalised messages and performance metrics can be sent to third party systems for real time alerting of any significant deviation from normal behaviour, and can also serve as a source for risk, compliance and other analytic systems.
An added value of on the wire packet capture is the ability to use this data to perform transaction based APM – a feature that Corvil have branded as Streaming Analytics. An auto-discovery feature will identify applications by comparing network headers against a list of known ports, invoking an Analytics plug-in which will reconstruct messages and transactions to extract application specific metrics.
Visualisation and scope
The web-based Dashboard Viewer supports the creation of business-level dashboards that offer instant views of key performance and business analytics. Big Data Adapters enable integration with existing reporting systems such as Splunk, Hadoop, Tableau or Storm.
Additionally, Corvil provides management reporting to schedule regular reports on the network, applications or specific business analytics. For example, infrastructure trending data can be scheduled to run monthly and distributed to network engineering management, or end of day reports to managers on key application metrics, such as user experience performance summaries, server hit counts and application errors.
What could be improved: