Skylight performance analytics

Overview

The Accedian Skylight solution is already known for its ability to produce some of the most precise network and application performance data available, and its our analytics platform that lets customers get the full value out of it.

• Developed in a DevOps fashion alongside a Tier 1 mobile operator, Skylight Performance Analytics allows you to monitor and analyze your network at a macro level, and dive deep into micro level details when investigating issues.
• Valuable performance information is now available in an accessible way to groups outside of Network Ops, people like product managers, network planners, customer service and management.
• Machine learning is leveraged to find insights in your data, with recommendations presented to the user.
• Analytics can also be extended to support non-Accedian time series data like you’d get from router telemetry or server performance statistics. Correlating this 3rd party data with network and application performance metrics elevates and enriches the actionable insights that Skylight Analytics provides.

The volume of performance data being collected is growing like crazy as vendors make it more available and operators design networks and services with QoE as THE KPI. Don’t drown in a sea of data, rise above it to take control - Skylight Performance Analytics is specifically designed to optimize your big data workflows, and has a rich pipeline of features to make sure you’re ready for the future.

Key Advantages

Architecture

Skylight performance analytics is a cloud native solution that is built to scale to support the largest of networks.
The following diagram illustrates our high level solution architecture.

Let's walkthrough this from the bottom up!

Data Ingestion

A containerized data collector called roadrunner runs close to where the PM data is generated. Roadrunner parses the data via the configured dictionary according to the reporting object type, then streams the result into the Skylight analytics solution over a secure connection.

Data Pipeline

Streams from all of the collectors hit a load balancer in front of the Analytics service. From here, data is distributed across multiple message buses where it waits for ingestion tasks to run analytical algorithms and store the data.

Data Storage

Ingested data gets stored in two locations:

1. A high-performance time-series datastore that provides exceptional query response times for complicated aggregated queries across 1000s of objects over variable time periods. This allows us to avoid pre-aggregation in the majority of cases and lets the user work with the data, slicing it however they want to view the world.
2. To disk in a big data format called Parquet. Our analytics engine utilizes these files for pattern matching and other advanced analytics use cases.

Data Visualization

Accedian prides itself on its UX - we work with customers closely to optimize workflows through the UI to make users more efficient. Our modern user interface enables our powerful monitoring and troubleshooting capabilities with lots of customizability.

Data Analytics

This component is where ingestion and post-ingestion algorithms run on the perform:

• data cleaning – remove records or metrics that are erroneous or match customer defined cleaning rules
• data aggregation (roll data up into pre-configured periods for faster query performance based on the granularity requested)
• monitoring for threshold crossing events
• and many more in the future

Analytics as a Service

Accedian offers Skylight performance as a managed service, meaning our customers get all of the following:

• 24/7 service monitoring by our SRE team
• In-app chat support from our experienced CSM team
• Automatic updates to the latest software