For communication service providers (CSPs), understanding how their network is utilized by their subscribers (and the resultant quality of service – QoS) and how the subscribers experience the network and services are the significant factors influencing their revenue and business value.
Guavus, a pioneer in AI-based analytics, today announced Guavus-IQ — a comprehensive product portfolio that provides a unique, multi-perspective analytics experience and real-time insights for CSPs in the areas of their network behavior and subscriber experience with respect to network and services provided.
Guavus provides an “operator-friendly” analytics experience, which is a simplified on a single pane of glass where CSPs need not be data scientists or big data experts – this allows them to focus on their core business and value.
Break through areas
Multi-perspective Analytics Experience:
Guavus’ real-time “outside-in/inside-out” perspective helps network operators identify subscriber behavioral usage patterns and better understand their operational environments, enabling them to increase revenue opportunities through data monetization, deliver an improved customer experience (CX), and reduce costs through automated, closed-loop actions.
What are the top 3 requirements for CSPs to understand the subscriber’s pulse and needs?
High impact Requirements | Traditional Approach | Guavus-IQ Approach |
---|---|---|
Closed loop network impact, network impacting events or service for subscribers, such as load/activity impacting network performance or availability. | Custom methods (manual/semi-automated) which process customer care / service tickets, a plethora of alarms/events, static rules, subscriber churn, surveys, limited maintenance windows, to deliver action items for better QoS. | Ingest the data from network management system, enrich and consolidate via in memory streaming. Apply analytics with pre trained ML models to give clear insights on the factors aggregated, so that MTTU (mean-time-to-understand) and MTTR (mean-time-to-repair) are reduced significantly, thereby improving subscriber QoS. |
Subscriber Quality of Experience (QoE) | Process different data sources such as network metadata: EDR (event data record), CDR (call data record), DPI (deep packet inspection), billing system (BSS) and apply methods to join the data and create insights via traditional RDBMS (relational database management systems). | Ingest and aggregate all disparate data, apply disturbed query engine on all big data for correlation and insights, providing subscriber behaviour micro-segmentation via an intuitive user interface (UI) workflow. |
Subscriber Quality of Service (QoS) | Collect Network metadata, device location, device signatures, customer care / service tickets, and process via different DBs and consolidate to gain certain insights. | With features like device enrichment, multiple data source types support for ingestion, and location mapping - correlated pipelines that are created in memory to deliver insights on location-based network experience. |
“Operator-friendly” By Design:
The Guavus-IQ portfolio, powered by its underlying Guavus Reflex platform, comes with easy to understand and use visualization tools for both Ops-IQ and Service-IQ product areas. With Ops-IQ Network Analytics, you can create a next generation network operations center (NOC) without new UI, training, workflow or automation changes.
Guavus has an “operator-friendly” analytics dashboards experience which is based on mature open source tools like Grafana where in you can go and drill down the insights, play around and calculate results with click of few buttons. These analytics provide insights for a host of value-add use cases.
Guavus’ common data platform unites the product experience for both Service-IQ and Ops-IQ, as the system “ingests once” and allows for an “analyze many” mode of operation. This efficiency enables data reusability and decreases the amount of storage required for on-going analysis.
Ops-IQ Network Fault Analytics Dashboard
Service-IQ Marketing Analytics Dashboard
Enterprise Grade, Scalable, Resource Efficient – On-Premise, Cloud ready:
Guavus-IQ portfolio operates on a common unified data platform for both Service-IQ and Ops-IQ product areas, allowing for the reduction of hardware footprint compared to legacy, monolithic big data platforms. The concept is based on “Ingest once, analyze many” model, where in data is ingested once via common platform, sliced / diced, and analyzed in in a multi-perspective manner to offer diverse use cases for CSP to achieve the improved ROI.
Based on CSP needs, especially in large scale IoT scenarios, the data ingestion and analytics can be provisioned at edge. By using Guavus-IQ’s real-time streaming capabilities wherein pipelines can be orchestrated at the edge location, achieving low latency results, while delivering continuous data with sub-millisecond latency to the core. It is been observed that we have achieved 40% of footprint reduction with respect to memory and core compared to traditional systems for processing millions of records per minute. Storage savings are observed with consolidated and final aggregations going into HDFS/Hive.
Cloud Ready:
- Virtualized/hyperconverged and secure infrastructure is provisioned by cutting edge DevOps tools and techniques which power the cloud native microservices, decreasing the deployment time to few hours along with scale out run time environment.
- The Guavus-IQ Reflex data platform is orchestrated by Kubernetes, with in-memory efficient pipeline processing, insights derived via Presto, ML modules in Spark, augmented by Hive for big data execution.
- The resources needs are dynamically tunable for the required data sizing with containers running stateful pipelines spawned as when and required, based on performance needs.
- The CSPs can view and act on common KPIs for monitoring and logging via Grafana dashboard powered by Prometheus.
Example Scenarios
As a CSP, your data lake is a storage repository of various network events, subscriber information, device information, network meta data and alarms, etc.
How do you deal with such disparate type of data and how do you derive insights of such big data, and turn them into best customer experience you can offer, there by having a positive impact on your bottom line?
Alarm Noise Reduction and Fault localization:
NOC operators are overwhelmed with the storm of alarms, and over worked to drill down and identify the right set of alarms which are impacting the network’s operation (e.g. availability, performance, etc.).
The above scenario also comprises significant noise (e.g. false positives, maintenance cycles, etc.) due to huge number of alarms, which makes it difficult for operators to prune, identify and rectify the actual issue which impacts the QoS of the network towards subscribers. These numbers also result in true incidents being missed due to operators being overwhelmed by the sheer volume of alarms.
Guavus’ Ops-IQ Network Fault Analytics module solves the scenario by enabling the following workflows for the NOC (network operations center) operators.
As the data get ingested and validated via an “ingest once, analyze many” model, the data is modelled via random Forest Model and Hill Climbing Model, followed by the execution of analytics workflow (via Spark jobs) which mines all the alarms to cleanse, categorize and correlate the result.
As an operator, via the Ops-IQ user interface (UI), you would realize the value of alarm correlation, consolidation and noise reduction (suppression). Further down, when you drill into the “Search” button on the unified Ops-IQ 2.0 UI, you get landscape view of alarms which can be filtered on “Prediction to Incident”, “Confidence”, “Severity”.
The NOC operator is left with only focused set of alarms to work on to identify the actual cause and time to resolve becomes faster. The noise reduction is also achieved by alarm prioritization backed with real-time characteristics, potentially achieving faster MTTU for service quality degradation and fault isolation.
Topology Independent and Rules Agnostic:
Traditional Alarm detection and correlation is based on the static rules (specifying what actions need to be taken for which kind of alarms) being configured catering to a specific topology and the rules need to be tweaked each time based on the effectiveness on alarm detection, adapting to network topology changes and equipment changes, etc.
It is been observed these kinds of rules-based approach is highly complex, cumbersome and requires manual effort. Adding to that, the NOC operator cannot do this by themselves but need SMEs (subject matter experts) to maintain the static rules.
Guavus’ Ops-IQ Network Fault Analytics module solves the problem with the help of ML based algorithms which is been modeled on vast set of network family of alarms/rules along with on-premise topology fields for cross validation. Hence, the dependency on physical topology and static rules to predict alarm family and root alarms is minimal. The alarm noise reduction and correlation are done dynamically adapting to network configuration changes, operating well in SDN/NFV scenarios.
As an operator, you need not worry about updating the rules and maintaining them. You can use the unified ingestion framework of Guavus-IQ to ingest and validate the data, run enrichments and training on the newly ingested data using Ops-IQ Network Fault Analytics, achieving alarm correlation and deducing family of alarms via operator friendly UI workflows, without worrying about underlying topology changes.
Now, you are ready to explore the dashboard and drill down views to get the insights on reduced noise, alarm priority and incident prediction to allow you to achieve MTTR (mean-time-to-respond).
Audience Measurement Service:
With all the disparate data sitting at data lake, the anonymized subscriber data with precise location information mapped could give you an opportunity to:
- Deliver targeted marketing campaigns, improvements localized to the geography
- Derive subscriber “footfall” insights enabling you to focus on needs of subscribers
Service-IQ Marketing Analytics gives you Audience measurement service (footfall option) where in the anonymized subscriber information is pruned, enriched and mapped and associated with device location insights.
If you launch the “User Distribution” workflow, after securely authenticating yourself in Service-IQ Marketing Analyticsuser interface, the default view shall show the percentile distribution of all users during the last 7-day period.
There are drill down filters available based on your focus area. For example, if you want to get the audience footfall, click on “filters”, in the “Usage Filters”, you can select the ‘Video Streaming (w/o P2P)’ values in the Global Data Services, then ‘Apply’ to refresh the chart.
This would give you the view of user segments concentrated on “Video Streaming”. Likewise, there are further drill down filters available for user/subscriber behavior insights. With the data available in such dashboards as well as via database queries, you can take a targeted action, as well as sell anonymized subscriber information to partners.
Targeted Marketing Campaigns:
With insights regarding user segmentation and footfall measurement available, how do you leverage those insights to feed into your Campaign Management System (CMS) for best subscriber service experience as well as for monetizing the new insights via partners?
Service-IQ Marketing Analytics provides you with an ability to do targeted market campaign to your customers. This comes at cost of few clicks from operator perspective.
The opportunity is to leverage Service-IQ to gain insights about subscribers segmented by a specific service, for example, online video streaming service (e.g. Netflix) as well as a view of those users who are not utilizing video streaming services like Netflix. Following workflow explains how to achieve that
After authenticating and creating the video users’ segment as per previous scenario workflow, choose “Content filters” and select “Netflix”. This gives you a new segment of users who are into Video streaming and in that using Netflix service. Now with two segments created for “Video users” and for “Netflix v service users”, you can apply the “Auth” calculation on them, which will give you the # of users who are not using “Netflix”. Applying ODS filter as “Video interest” on this resultant data will give you an insight on what video services are used by this user segment. Service-IQ enables the operator to understand the common attributes or characteristics of both user populations to better understand their content interests and motivations – this is extremely powerful for CSPs seeking to better understand the usage behaviors of subscribers by type and to leverage to improve customer experience (CX) for key services and applications.
This drill down style view, integrated with leading campaign management systems, enables the CSP marketing team to make custom and personalized marketing campaign based on subscriber interests or to share and monetize the insights with OTT service providers.
Here is the typical workflow for processing subscriber related data within Service-IQ Marketing Analytics:
- Ingest data from all types of sources and types (EDR, CDR, DPI etc., with CSV, ORC formats) via SFTP (secure file transfer protocol) and Kafka plugins
- Enrich the data by combining and correlating multiple data sources using in memory streaming engine (Guavus’ SQLstream)
- Dynamic segmentation of subscriber behavior with offline and online ML model training
- Query the aggregated enriched data in Hive via distributed query engine, Presto.
- And these are manifested via the Operator friendly UI workflow which allows you to visualize the segments with few clicks and low latency. Users are authenticated and authorized via OAuth2/ODIC protocol offered by Keycloak
Summary
To recap, Guavus-IQ provides a comprehensive product portfolio that provides a unique, multi-perspective analytics experience and real-time insights for CSPs in the areas of their network behavior and subscriber experience with respect to network and services provided.
In addition, Guavus-IQ enables CSPs with “operator-friendly” advanced analytics and AI use cases, which is a simplified on a single pane of glass where CSPs need not be data scientists or big data experts – this allows them to focus on their core business and value.
As we look forward to our next blog, we will continue to drill down on the capabilities and use cases for Guavus-IQ, with a focus on explaining the technical aspects, value and benefits of the offering to transform how CSPs plan, deploy, and manage their network and service operations.
Additional Resources: