Cloud Cost Optimization for Video Assessment Platform

Challenges

Our client, a B2B SaaS video assessment platform, faced a significant challenge with its rapidly growing, complex multi-region infrastructure, leading to a high cloud expenditure of approximately $300,000. EbizOn was tasked with reducing these costs while maintaining the application's performance, availability, and resiliency. Additionally, the infrastructure needed to accommodate load bursts due to variable application demand.

Solutions

Here are the key steps we took:

• Compute Optimization : The client was using a mix of Reserved Instances (RI) and on-demand instances, with compute costs making up over 50% of their monthly expenses. EbizOn analyzed the usage patterns and per-CPU utilization to optimize the infrastructure by right-sizing the EC2 instances and employing the optimal mix of RI, on-demand, and spot instances. This strategy reduced compute costs by 30% while ensuring high availability.
• Log and Artifact Storage Optimization : The client's extensive infrastructure led to daily builds across multiple regions, causing substantial storage consumption for build artifacts and log files. We optimized the storage footprint using serverless functions to intelligently manage the storage and deletion of artifacts based on current deployments. We automated the regrouping and compression of older logs, resulting in a 60% reduction in storage costs.
• Multi-Media Object Storage Optimization : We optimized the multi-media object storage lifecycle policies using various S3 storage classes, including Standard, Intelligent-Tiering, Infrequent Access, Glacier, and Deep Glacier. By analyzing usage patterns, we achieved the right balance between storage costs and object retrieval costs, reducing storage expenses by approximately 30%.
• Network Optimization : We analyzed the VPC and subnet network setup and identified inefficiencies, addressing them to reduce unnecessary costs associated with inefficient data transfer between different services.

Impact

• Enhanced the dollar value of per-CPU core compute power by 30%
• Reduced overall cloud spend by approximately 33%, cutting monthly costs from $300,000
• Implemented policies for build artifact storage, achieving a 60% reduction in storage costs
• Lowered S3 storage costs by 30%