Big Data Analytics

Using Your Existing Data to Analyze Users’ End-to-End Network Experience

Voyance employs both streaming as well as batch analytics. By leveraging machine learning techniques, such as clustering and regression on all of the input data sources, enterprises now have at their fingertips a detailed end-to-end analysis of the user experience with the ability to predict outcomes. No other analytics solution delivers this breadth or depth of big data network analysis.

Service Baselining

Automatic baselining (i.e. what’s “normal”) for user experience for every service and application in the environment. Detection of deviations with very low probability of false alarm and thus alert operations when sudden changes occur.

Recommendation Engine

Observations of historical data in order to make predictive recommendations. “If you take action X, you will get a benefit of Y.” Additional information is correlated with each recommendation to give starting points to further exploration.

Multi-dimensional Analysis

Insights into extremely complex questions, such as “Which Windows laptop users on the corporate SSID are experiencing poor Skype for Business user experience. Correlation, indexing and visualization of every client incident across the relevant dimensions such as root cause, location, SSID, VLAN, OS, etc.

Cross Stack Analytics

Detailed data correlation of incidents with all of the contributing factors across stack,(eg. when client experienced poor Web experience, at the same time was Wi-Fi, the WAN or any other network service experiencing problems?) Eliminate potential root causes and point IT in the right direction for solving problems and improving user experience.

Incident Detection

Automatic identification and characterization of network incidents. When a service or application deviates from its baseline this ‘incident’ of varying priority is created. Once a high enough priority incident occurs, an alert is sent with important information such as the most common root causes, the specific client devices affected as well as the most common attributes associated with those clients. Alerts can be sent via e-mail, SMS, or automatically entered into a ticketing system such as Servicenow and Slack.

Root Cause Analysis

Correlation of any relevant data sources that apply to root cause of a particular type of incident to systematically call out or eliminate potential root causes. (eg. if issues related to client devices seeing high radius protocol latency are detected, the system correlates whether, at the same time the server CPU or memory utilization is high). Correlation is also performed on reason codes sent by the server to see if those could be the cause.

Problematic Clients

Machine learning detects clients that are outliers in terms of poor performance on any metric (Wi-Fi, Application, Connectivity, etc.). These outliers are labeled as ‘problematic clients’ and are surfaced for proactive troubleshooting.

Minute-by-Minute Correlation

Analysis and correlation all of the data sources at very short time intervals. These correlations yield root causes about what happened at the same time that could have caused the problem. For WAN flow analysis, per minute correlations yield the number of high link utilization minutes as well as the specific users or applications that contributed to those high link utilization minutes to help answer the question question “what caused the Internet to be inaccessible at a certain time on a given day?