Skip to content

About

Self-hosted sFlow collector and traffic analytics for RUCKUS ICX switches: top talkers, flow maps, port utilization, trends, alerts. Offline installer and VMware appliance (OVA).

Topics

Resources

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Repository files navigation

sFlow Analytics

Self-hosted sFlow collector and traffic analytics for RUCKUS ICX switches (and any sFlow v5 device). Answer "who is saturating this uplink?" and "how is my network used?" in a few clicks, with no license, no agent and no extra hardware: the switches already export sFlow.

Dashboard

Features

  • Traffic Explorer: top talkers, applications, protocols, conversations (one-way or both directions), VLANs, per switch, per port or per group; every filter combines and every view is a bookmarkable URL; CSV export.
  • Flow map: who talks to whom as a Sankey diagram (hosts, services, ports, VLANs, groups), for the whole network, one switch or one interface.
  • Port health: exact utilization from interface counters (average, 95th percentile, peak), discards and errors, broadcast / multicast storm detection.
  • Trends: day, week, month or year compared with the previous period (or a year before), from a 3-year hourly history per interface.
  • Alerts: utilization, discards, broadcast storms, silent switches, host or group traffic thresholds; e-mail, Microsoft Teams / Slack webhooks, syslog.
  • Reports: weekly and monthly PDF and Excel reports (monitored links, busiest ports, applications, talkers, trends, alerts), e-mailed and kept 90 days; "Report now" on any period.
  • RUCKUS One or SmartZone integration (optional, read-only): switch and port names, LLDP neighbors, client names. Standalone switches: SNMP v2c / v3 polling (sysName, ifName, ifAlias, LLDP), or port names from the ICX numbering.
  • Names everywhere: host aliases (IP, MAC, subnet), reverse DNS, MAC vendors (bundled IEEE registry), IP and interface groups.
  • Multi-user: administrators and read-only accounts, sign-in log, HTTPS only, your own certificate (PFX, PEM or CSR generated on the appliance).
  • Appliance-friendly: OVA for ESXi and VirtualBox or one-command offline install, updates from the web interface (backup first), 90 days of detailed data by default, automatic disk protection, backup script.

Flow map

Install

Two ways, from the latest release. First sign-in to the web interface: admin / sflow; a new password is required at once (sign in before exposing the appliance).

Option A: virtual appliance (OVA)

  1. Deploy sflow-analytics-<version>.ova (ESXi 6.7 or later: Deploy OVF Template, thin provisioning; VirtualBox 7: File → Import Appliance). Defaults: 4 vCPU, 8 GB RAM, system disk 30 GB, data disk 100 GB.
  2. Power it on and open the VM console: an assistant asks network (DHCP or static), host name, time zone, NTP and the password of the system account sflow, then installs everything (~2 minutes) and shows the URL.
  3. Later: log in on the console as sflow and run sudo sflow-console (status, network, admin password reset, restart).

If the ESXi Host Client fails to upload the OVA, check that your PC resolves the ESXi host name, or extract the .ova (a tar archive) and deploy the .ovf + .vmdk files.

Option B: existing Linux VM

On Ubuntu 22.04 / 24.04 or Debian 12 (x86_64), 2 vCPU, 4 GB RAM, 50 GB disk or more:

tar xzf sflow-analytics-*.tar.gz && cd sflow-analytics-*/ && sudo ./install.sh

The script installs Docker if needed, loads the images, generates the configuration and prints the URL. No switch at hand? sudo ./install.sh --demo simulates 5 switches. Everyday commands, backup and uninstall: packaging/QUICKSTART.md.

Upgrade

Administration → System shows when a new release is published: Update now makes a backup, installs it and restarts the application (about one minute; data and settings are kept). Without Internet access: extract the new package on the host and run sudo ./install.sh again.

Send sFlow

RUCKUS ICX:

sflow destination <appliance-ip>
sflow enable
interface ethernet 1/1/1 to 1/1/48
 sflow forwarding

sflow enable is required. For small or quiet networks add sflow sample 512 (the default sampling rate gives few samples). Switch and port names come from RUCKUS One / SmartZone, from SNMP (snmp-server community <name> ro, or SNMPv3) or can be typed in. More in docs/sflow.md.

Traffic Explorer

Sizing

Detailed flows and counters are kept RETENTION_DAYS (90 by default); the hourly history used by Trends is kept 3 years. Disk needed for flows ≈ samples/s × 86 400 × days × 25 bytes:

Sustained samples/s Per day 90 days
100 0.2 GB 19 GB
1 000 2.2 GB 194 GB
5 000 11 GB 972 GB

When the disk fills up, the oldest days are dropped automatically. Put the data on a second disk and grow it at any time: docs/deployment.md.

How it works

ICX switches ──sFlow v5 / UDP 6343──▶ collector (Go) ──▶ ClickHouse ──▶ API (FastAPI) ──▶ web UI (React)
                                                                 ▲           │              behind nginx HTTPS
                                 worker: history, alerts, reports, DNS, RUCKUS One / SmartZone sync, SNMP

Traffic volumes are estimates (sampled packet size × sampling rate), as with any sFlow tool; port utilization comes from the switches' interface counters and is exact. Details: docs/architecture.md, REST API: docs/api.md (OpenAPI at /api/v1/docs).

Development

Requirements: Docker with Compose v2 (Go 1.25+, Python 3.14 and Node 24 only to run tests outside containers). See CONTRIBUTING.md and SECURITY.md.

scripts/dev-up.sh          # .env with a random ClickHouse password, builds and starts the stack
scripts/dev-up.sh --test   # same, plus synthetic traffic from 5 fake switches
make test                  # collector unit, fixture and UDP end-to-end tests
make api-test              # API tests against a throw-away ClickHouse database
make integration           # collector <-> ClickHouse tests
scripts/ui-screenshots.sh  # every UI page in headless Chromium, fails on console errors
scripts/build-package.sh   # offline installation package in dist/
Directory Content
collector/ Go collector: UDP listener, sFlow decoding, exporter discovery, ClickHouse sink, disk guard, synthetic sender
api/ FastAPI service and background worker (alerts, reports, history, DNS, controller sync, SNMP)
frontend/ React + TypeScript web UI
nginx/ HTTPS reverse proxy
packaging/ offline installer (install.sh, compose file, quick start)
appliance/ OVA build (Ubuntu 24.04 cloud image, first-boot console assistant)
db/, docs/, scripts/ ClickHouse config, documentation, tooling

Roadmap: TODO.md · Changes: CHANGELOG.md

License

© 2026 Frédéric Muffat Es Jacques. All rights reserved — see LICENSE. You may install and use the published releases free of charge, including for your customers; copying, modifying or redistributing the source code requires written permission.

sFlow Analytics is an independent tool for RUCKUS ICX switches (and any sFlow v5 device). It is not a RUCKUS product and is not supported by RUCKUS; RUCKUS, ICX, SmartZone and RUCKUS One are trademarks of their respective owner(s), and this project is not published, endorsed by, or affiliated with RUCKUS Networks, Belden, or their affiliates. It is provided "as is", without warranty of any kind, express or implied — see LICENSE for the full disclaimer.

About

Self-hosted sFlow collector and traffic analytics for RUCKUS ICX switches: top talkers, flow maps, port utilization, trends, alerts. Offline installer and VMware appliance (OVA).

Topics

Resources

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages