highlights and lowlights of the daily business of network engineers

Download als PDF File

Purpose

This note describes the separate iMaster NCE-CampusInsight product in independent deployment mode:

• product positioning
• logical architecture
• data model
• deployment modes
• high availability / DR
• typical hardware and sizing guidance

Product Positioning

• CampusInsight is an independent intelligent O&M platform.
• It uses existing O&M data such as device metrics and client logs.
• Its goal is intelligent troubleshooting and the digitization of user experience.
• Core idea: move beyond purely rule-based monitoring toward data-driven analysis with big data and AI.

Logical Architecture

According to the product documentation, CampusInsight uses a Huawei big-data analytics platform and analyzes network data through intelligent algorithms.

Architecture layers:

• Data and analysis result visualization
  • visualization from the network, application, user, and optimization perspectives
• Data analysis
  • collects, processes, and distributes data and manages analysis tasks
• Big data analytics platform
  • uses components such as Kafka, Spark, HDFS, and Druid
• AI engine
  • provides the AI/ML framework and algorithm libraries
• Device management and data collection
  • SNMP, Telemetry, and Syslog as the main inputs

Data Model / Data Collection

The documentation clearly separates data collection:

• SNMP
  • for device addition / device management
• Telemetry
  • for metric data
  • according to the documentation, packets are encoded with ProtoBuf
• Syslog
  • for log data

Practical meaning:

• SNMP = mainly management / onboarding
• Telemetry = the actual performance and analytics data foundation
• Syslog = event and log perspective for issue analysis

Core Product Characteristics

• Intelligent Analysis based on Real Service Flows
  • metrics and logs are collected through Telemetry
  • analysis is based on real service flows
• Big Data Processing Capability
  • centralized collection, storage, and analysis of large data volumes

Deployment Modes

Supported installation modes:

• Preinstallation
  • OS already installed, product preconfigured, installation through EasySuite
• Manual OS installation
  • servers/environment prepared manually, then product installed through EasySuite
• One-click installation
  • OS and product installed in one step when hardware and network planning meet the requirements

Typical Networking

• CampusInsight consists of at least three analyzer nodes.
• In networking design, different network segments are recommended for the external access network and the southbound network.
• The goal is to avoid link congestion caused by traffic pressure.

Protocol Stacks

Supported protocol-stack scenarios:

• northbound and southbound both IPv4
• northbound IPv4, southbound IPv4/IPv6
• northbound and southbound both IPv6

Important note:

• when northbound and southbound are both IPv6-only, DR is not supported.

High Availability and DR

Single-site cluster HA

• CampusInsight supports cluster deployment.
• A cluster has at least three server nodes.
• If a single node fails, services can be restored quickly and automatically.

Dual-site DR

• Nodes in the same cluster must not be distributed across sites or subnets.
• For higher resilience, two geographically separated sites are deployed.
• Each site runs its own cluster.
• DR is implemented between the two clusters:
  • one site is Primary
  • the other is Secondary
• If the primary site fails, services are automatically switched to the secondary site.

Sizing / Hardware Baseline

CampusInsight supports:

• physical servers
• VMs

Important baseline rules from the documentation:

• different hardware configurations limit the management scale
• Huawei servers are recommended for physical deployment
• customer-provided servers are also possible if they meet the requirements
• clusters typically require stronger NIC and storage characteristics than single-node deployments

Typical Minimum / Guidance Values for Customer-Provided x86 Servers

Single-node with 128 GB RAM

• CPU: at least 32 physical cores, 2.2 GHz+, hyper-threading
• RAM: 128 GB+
• system disk: 900 GB after RAID
• data disk: 3000 GB after RAID
• NIC: at least 1 GE, recommended 2 NICs with 2 GE ports each

Single-node with 256 GB RAM

• CPU: at least 32 physical cores, 2.2 GHz+, hyper-threading
• RAM: 256 GB
• system disk: 900 GB after RAID
• data disk: 3000 GB after RAID

Standard single-node with 256 GB RAM

• CPU: at least 40 physical cores, 2.2 GHz+, hyper-threading
• RAM: 256 GB
• system disk: 900 GB after RAID
• data disk: 3000 GB after RAID

Cluster with 128 GB RAM per node

• CPU: at least 32 physical cores, 2.2 GHz+, hyper-threading
• RAM: 128 GB+
• system disk: 900 GB after RAID
• data disk: 3000 GB after RAID
• NIC: at least 1x 10GE, recommended 2 NICs with 2x 10GE each

Cluster with 256 GB RAM per node

• CPU: at least 32 physical cores, 2.2 GHz+, hyper-threading
• RAM: 256 GB
• system disk: 900 GB after RAID
• data disk: 3000 GB after RAID

Standard cluster with 256 GB RAM per node

• CPU: at least 40 physical cores, 2.2 GHz+, hyper-threading
• RAM: 256 GB
• system disk: 900 GB after RAID
• data disk: 5000 GB after RAID
• NIC: at least 1x 10GE, recommended 2 NICs with 2x 10GE each

Additional Hardware Notes

• random read/write speed of the disks: at least 100 MB/s
• RAID controller must support WriteBack
• for x86 PM deployment, the documentation lists operating systems such as Huawei EulerOS V200R012C00 or SUSE Linux Enterprise Server 12 SP5
• NVMe SSDs are not supported according to the documentation
• for non-Huawei servers, Huawei evaluation is recommended before implementation

Meaning for Our CampusInsight Notes

• The existing CampusInsight note can now clearly distinguish between CampusInsight as a function around NCE and CampusInsight as a standalone product.
• For architecture topics, the dedicated product documentation is the better source.
• For network, user, and application views, it is also more accurate and more authoritative.

Key Takeaways

• CampusInsight Independent Deployment is a standalone analyzer product.
• Architecture: SNMP + Telemetry + Syslog -> Big Data Platform -> AI Engine -> Visualization.
• A production cluster starts at 3 nodes.
• For real site resilience, dual-site DR with primary and secondary sites is supported.
• Cluster networks should be cleanly planned, especially external access and southbound.

Sources

• 001_Docs/IMasterNCE/Campus_insights/profile.xml
• 001_Docs/IMasterNCE/Campus_insights/resources/toctopics/en-us_topic_0191059516.html
• 001_Docs/IMasterNCE/Campus_insights/resources/toctopics/en-us_topic_0191059515.html
• 001_Docs/IMasterNCE/Campus_insights/resources/toctopics/en-us_topic_0191059544.html
• 001_Docs/IMasterNCE/Campus_insights/resources/toctopics/en-us_topic_0191059520.html
• 001_Docs/IMasterNCE/Campus_insights/resources/toctopics/en-us_topic_0210921168.html
• 001_Docs/IMasterNCE/Campus_insights/resources/toctopics/en-us_topic_0000001371293128.html
• 001_Docs/IMasterNCE/Campus_insights/resources/toctopics/en-us_topic_0000001421972497.html
• 001_Docs/IMasterNCE/Campus_insights/resources/toctopics/en-us_topic_0201135485.html

Samuel Heinrich

Senior Network Engineer at Selution AG (Switzerland)

Arbeitet in Raum Basel (Switzerland) als Senior Network Engineer mit über 15 Jahren Erfahrung im Bereich Netzwerk