DEV Community

Sagar Joshi
Sagar Joshi

Posted on

Why Real-Time Communication Still Breaks — And What a 2014 Idea Got Right

Every time a Teams, Zoom, or Webex call freezes during an important moment, we’re reminded of something uncomfortable:

Real-time communication is still too fragile.

Even in 2025, we continue to see:

  • Regional outages
  • Overloaded media servers
  • Capacity limits blocking new joiners
  • Users forced to rejoin calls

Years ago at IBM, we explored a different path — one that didn’t depend on running idle backup servers.


A Simple but Powerful Idea (US Patent 10,051,235)

Instead of keeping mirrored servers on standby, the system could:

  • Monitor performance using SDP thresholds
  • Dynamically elect an active participant as a temporary media relay
  • Switch over seamlessly when the primary server failed

No reconnects.

No cold backups.

No duplicated cloud cost.


Traditional Redundancy vs. Our 2014 Approach

Traditional Redundancy

  • Requires mirrored idle servers
  • High cloud/infra cost
  • Failover takes seconds
  • Participants must reconnect

Our Approach

  • Elect from active participants
  • Zero idle infrastructure
  • Instant failover
  • Session stays alive

A 2025 Application: Scaling When Servers Hit Capacity

One of today’s biggest challenges:

Live events and webinars hitting concurrency limits.

When a media server is full, instead of rejecting viewers:

Let an existing viewer temporarily proxy the stream.

This gives:

  • Graceful scaling during spikes
  • Fewer “room full” errors
  • Continuity during partial outages
  • Better reliability in poor networks

The Hard Part: Graceful Failback

A modern design would need to:

  • Sync media + session state
  • Avoid ping-pong failovers
  • Time the switch during quiet windows
  • Possibly split media vs signaling paths

With today’s telemetry (latency, jitter, CPU), this can even be predictive.


Quiet Validation Over the Years

This idea was later granted as U.S. Patent 10,051,235, now cited by:

  • Unify
  • Vonage
  • VMware

(Full citation list available on Google Patents.)


Closing Thought

Failure in real-time systems is inevitable.

But failure doesn’t need to reach the user.

If you're working on real-time media, cloud systems, or distributed comms, I’d love to hear your perspective:

What does “never drop the call” look like in your world today?

References

https://www.linkedin.com/pulse/from-cloud-outages-built-in-continuity-rethinking-sagar-joshi-8xogc
https://patents.google.com/patent/US10051235B2
https://medium.com/@jo.sagar/from-access-cards-to-ai-presence-the-evolution-of-intelligent-communication-routing-5ad39dea698e

Top comments (0)