About the Customer
MITH is a creator‑first social platform where users publish content, join niche communities, and interact in real time. A blockchain‑based rewards engine underpins the experience, letting creators earn Web3 tokens that can be traced and distributed transparently.
Customer Challenge
As MITH’s user base grew, live‑stream audiences began to spike unpredictably. The platform suffered:
- Content‑delivery latency frequently breaching the 500 ms p95 target (peaks ~620 ms).
- Video‑session uptime slipping below the 99.9 % experience goal (measured 99.2 %).
- Reward‑payout errors during high‑traffic events, shaking user confidence and driving support tickets.
The team needed deeper observability, automated governance across multiple AWS accounts, and tighter cost control—without slowing the release cadence.
Partner Solution & Architecture
ZirconTech used a pure‑code approach—Terraform stored in Git and applied by CI pipelines—to deploy a fully tagged, multi‑AZ micro‑service stack that meets Cloud Ops best practices:
- Compute & Networking – Amazon AppRunner services and an ECS Fargate cluster inside a two‑AZ VPC with public/private subnets and a NAT gateway.
- Event‑Driven Core – Amazon EventBridge and Kinesis Video Streams fan out notifications and live‑video feeds.
- Data Stores – Amazon DynamoDB (PAY_PER_REQUEST) for millisecond user‑content reads; Amazon RDS PostgreSQL for session metadata.
- Identity & Security – Amazon Cognito for user auth plus least‑privilege IAM task roles, WAF web ACLs, organization‑wide SCPs, AWS Security Hub and Config rules with auto‑remediation.
- Observability – CloudWatch Container Insights, custom metrics, and AWS X‑Ray traces wired into every service; alarms for CPU, error rate, and custom business KPIs.
- FinOps Controls – Mandatory tagging, Cost Categories, CUR + Athena + QuickSight dashboards, and anomaly detection.
Because everything is defined in Terraform, a single terraform apply can recreate the environment and its guardrails end‑to‑end.
Results & Success Metrics
Two optimisation sprints delivered measurable impact:
- p95 content‑delivery latency: 620 ms → 340 ms.
- Video‑session uptime: 99.2 % → 99.93 %.
- Reward‑transaction success: 95.8 % → 99.4 %.
- Support tickets related to performance dropped 60 %.
- Daily active users rose 12 %; average session length grew 8 % in the following quarter.
Automated alarms now surface anomalies within two minutes, and monthly FinOps reviews map spend to features so product and finance teams act on cost drivers quickly.
Lessons Learned
- IaC as the contract – Keeping preventive (SCPs, WAF) and detective (Security Hub, Config) controls inside the same repo as application code guarantees consistent, testable governance.
- Trace everything early – Instrumenting X‑Ray and custom metrics before scaling allowed rapid bottleneck isolation when traffic surged.
- Business KPIs matter – Latency, uptime, and token‑payout accuracy resonated far more with stakeholders than raw CPU graphs.
- Cost visibility drives culture – Tag enforcement and unit‑economics dashboards turned cloud spending into a shared metric, not just an ops concern.
About the Partner
ZirconTech is an AWS Advanced Tier Services Partner specialising in Cloud Operations, DevOps, and blockchain solutions for high‑growth digital platforms. The team’s infrastructure‑as‑code and FinOps frameworks help customers like MITH achieve reliability, security, and cost efficiency at scale.