Intervention
We redesigned the dispatcher as OTP applications with back-pressure aware queues, plus observability hooks that expose delivery lag in near real time.
Telecom · 4 months
Erlang Backend Rewrite for Bulk SMS Provider
Rewrote the messaging backend in Erlang, tripling throughput headroom and cutting monthly compute cost by 30%.
Client not named under confidentiality terms; metrics from the engagement closeout report.
3×
Throughput headroom
30%
Monthly compute cost reduction
99.95%
Delivery SLO achieved
Before
A monolithic Node.js service was throttling high-volume SMS bursts, forcing the team to over-provision Kubernetes clusters in three regions.
Intervention
Observed result
- 3× throughput headroom during carrier promotions.
- 30% lower monthly compute spend after right-sizing the nodes.
- 99.95% delivery SLO with automated failover drills.
Limits
These measures describe this engagement only.
Selected records include work delivered before Anystack Engineering was incorporated in May 2024. Each record states its provenance; these results are not evidence of current commercial models.
How deterministic verification works.
These historical architecture patterns informed the invariant mapping and mutation testing models powering Pramāṇa.