Key insights from session with Reghunath B

Reghunath B

We recently hosted an exclusive session for our portfolio company VuNet Systems, bringing together industry leaders to discuss the growing criticality of observability in modern digital ecosystems.

Our guest speaker was Reghunath B, Head of Engineering at Atlassian, who previously built large-scale, resilient systems at Aadhaar and Flipkart share some interesting insights on Observability and its criticality for large scale systems.

Observability is Different from Monitoring:

  • Monitoring is tracking system-level metrics like CPU, memory, API status, etc.
  • Observability is about understanding why failures happen across complex systems — connecting dots across business metrics, application metrics, infrastructure health, and user journeys.

Observability Must Start at Product Definition, Not After:

At Google, the highest priority workload on their cloud (Borg) is monitoring services. Why? — “You can’t operate what you can’t see.” This is how critical Observability is but most organizations still treat observability as an afterthought.

  • In high-scale systems like Flipkart’s Big Billion Days and Minutes delivery, failures happened because observability wasn’t baked in early enough.
  • Ideal approach: Build observability as a fundamental requirement when designing the system.

Business-centric Observability is Crucial: Observability must be designed specifically for what keeps the business alive.

  • Observability must align deeply with the specific business goals and nuances. It’s not a horizontal tool like simple logging or monitoring.
  • Different industries (e.g., payments, quick commerce, lending) need different observability models to protect what matters most to their business.
  •  
  • Examples:
    • In Flipkart BBD: Observability focused on real-time sales, payment success rates, traffic handling.
    • In Aadhaar: Focused on enrollment throughput — 50,000/hour, 4 TB data processed daily.
    • In Quick Commerce: Ensured that milk is discoverable within 10 minutes delivery SLA. 

Observability Defines System Goals and Trade-offs:

Observability isn’t just about visibility — it helps define what “good” looks like for your system (throughput, latency, availability) and where you should optimize.

  1. Example: Aadhaar system needed to enroll 50,000 people/hour → so throughput became a system design goal.
  2. Quick Commerce needed deliveries in 10 minutes → so latency (time per order orchestration) became a critical design metric.
  3. Flipkart knew 100% uptime was too costly → so they accepted some trade-offs. For instance, while users are browsing for a product, they could tolerate slight inconsistencies, while when users added an item to the cart, it had to be 100% accurate and strongly consistent.

Future of Observability:

  1. Observability is not a plug-and-play tool.
  2. It needs customization for the business model (payments, quick commerce, etc.).
  3. As systems become distributed and fragmented, observability will become even more important — not just as a tech stack but as a business enabler.