Product Garaj

WorkAboutLifeContact

PM Portfolio - Project 01

Multi-Path WAN
Route Intelligence

From maritime fleet operations to enterprise NaaS: designing a BGP multi-path scoring and routing policy platform that lets operators express business intent without touching routing configs.

Origin

Shoft Shipyard

Role

Lead Product Manager

Timeline

3 years

Fleet Scale

30+ vessels

Domain

Maritime WAN / NaaS

Period

2019 - 2022

The Challenge

Maritime WAN infrastructure is fundamentally multi-path: VSAT, LTE, and shore fibre all running simultaneously, with BGP deciding which carries traffic. But the tooling to manage it was built for network engineers, not operators.

At Shoft Shipyard I owned the 3-year product roadmap for a B2B maritime WAN platform serving 30+ vessels and 5 enterprise port operators. I designed the Route Intelligence layer: a path-scoring and policy engine that translated business intent into BGP preferences without requiring operators to know routing protocols.

BGP Multi-PathRoute WeightingFailover AutomationSLA ObservabilityNaaS Platform

Research and Discovery

MK

Marina - Network Ops Lead

9 years, fleet network ops

Manages 22 vessels across 3 ocean regions. Manually monitors BGP sessions daily. Burned twice by outages traced to failover misconfiguration nobody caught in advance.

Core Pain

No visibility into which path takes over until the switch already happened

Design Goal

Confidence that failover will work correctly before the ship leaves port

TW

Thomas - WAN Architect

14 years, enterprise WAN design

Designs multi-path WAN topologies for high-value vessel fleets. Knows BGP deeply but must translate policy decisions into raw config with no abstraction layer for non-BGP stakeholders.

Core Pain

Clients ask for cost vs. reliability tradeoffs with no tool that maps business intent to routing policy

Design Goal

A layer between business requirements and BGP preference values

Key Insights

Operators want to trust failover, not configure it

Every interview surfaced the same tension: network engineers understand BGP path selection deeply but cannot translate it into business language for their operators. The product gap is not configuration - it is confidence.

Cost vs. reliability is a dynamic tradeoff, not a setting

Fleet operators weigh path cost differently in port versus at sea. Static routing policies fail because business context changes. Intent-based policy is what is actually needed.

Visibility is a pre-condition for trust

Teams without real-time path health had built shadow tooling: ping monitors, CSV exports, manual BGP log tailing. A platform that surfaces this data natively removes that friction entirely.

Explore the Platform

Three tabs show the operator experience: real-time path health across all WAN links, the intent-based route weighting simulator, and the failover event log.

Route Intelligence

MV Pacific Horizon - BGP Multi-Path

LiveVSAT Active

Path Health

Route Simulator

Failover Log

VSAT Primary

Satellite

ACTIVE

Latency

650ms

Pkt Loss

0.4%

Bandwidth

10 Mbps

Jitter

45ms

RELIABILITY94%

BGP AS6500199.1% uptime

LTE Cellular

4G / LTE

STANDBY

Latency

85ms

Pkt Loss

0.1%

Bandwidth

50 Mbps

Jitter

8ms

RELIABILITY97%

BGP AS6500299.7% uptime

Shore Fibre

Fibre Backhaul

STANDBY

Latency

12ms

Pkt Loss

0.0%

Bandwidth

100 Mbps

Jitter

2ms

RELIABILITY99%

BGP AS6500399.9% uptime

PRD Artifact: Route Intelligence v1

PRODUCT REQUIREMENTS - ROUTE INTELLIGENCE v1.0

01 - Problem Statement

Enterprise WAN operators managing multi-path connectivity cannot express routing policy in business terms. Configuration requires BGP expertise unavailable to operations teams, creating a dependency gap that delays policy changes and prevents proactive route management.

02 - Target Users

Primary: Network Operations Leads who configure path policy and respond to degradation events. Secondary: WAN Architects who design topology and validate policy correctness. Tertiary: C-suite stakeholders who need SLA reporting and cost visibility.

03 - Key User Stories

As a network ops lead, I want to see all path health metrics in one view so I can assess failover readiness before departure. As an operator, I want to adjust routing priorities using business-language controls without needing to know BGP local-preference values.

04 - Success Metrics

Time between path degradation and failover under 45 seconds, baseline was 4 minutes 12 seconds. Operator route policy changes without engineering involvement at 80% or more of total changes. Support tickets related to routing events reduced 30% within one quarter of launch.

05 - Out of Scope v1

ML-based predictive path selection, integration with third-party NOC platforms, per-application traffic steering, and IPv6 multi-path. These are scoped for v2 pending v1 adoption data.

Key Design Decisions

Single Primary Path vs. Scored Multi-Path

Before

Operators configured one primary and one fallback. No health scoring. A degraded primary stayed active until packet loss hit a hard threshold, often too late.

After

All three paths are continuously scored on latency, packet loss, jitter, and uptime. The router acts on relative health, not binary up or down.

Measured Impact

Mean time between outage and failover dropped from 4m 12s to under 30s in pilot testing

Hard-Coded Thresholds vs. Operator-Configurable Weighting

Before

Failover triggered at a fixed packet loss percentage set by the network engineer at deploy time. No mechanism to adjust for vessel context such as at sea vs. in port, cargo priority, or cost sensitivity.

After

Route weighting sliders let operators express business intent and the system maps those weights to BGP local preference values automatically.

Measured Impact

91% of operators in pilot said configurable weights made them more confident in policy decisions

Post-Event Reports vs. Live Path Health Monitor

Before

Uptime data was available as a CSV export the following day. Operators learned about failover events through alarms or client calls, not proactively.

After

Real-time path health dashboard with per-metric scores, rolling uptime, and BGP session state. Failover event log is persistent and filterable.

Measured Impact

Support tickets related to routing events dropped 38% in the quarter after launch

Outcomes

62%

reduction in unplanned outage duration after operators adopted pre-departure path health checks

Shoft Shipyard fleet data, 2020-2022

4 days

enterprise customer onboarding reduced from 3 weeks with Terraform-automated VPC and route table provisioning

Shoft Shipyard deployment records

$2.1M

ARR in renewed SLA contracts directly attributed to verifiable uptime data and automated failover reliability

Shoft Shipyard commercial outcomes, 2022

99.4%

delivery reliability across 30 plus vessels and 5 port operators over 3 years of platform operation

Platform SLA reporting, 2019-2022

Product Portfolio - Rebecka Raj

Back to Projects