Help Butterfly Effect Refactoring,Migration andOptimize E2B infra(AlAgent) Platform

About customer

Butterfly Effect is a Chinese startup focused on AI agent technology, with its core product Manus being a universal artificial intelligence agent platform. Manus Agent and related systems are developed and built based on the E2B framework. Due to the fact that E2B currently only provides deployment solutions based on GCP, customers have encountered multiple technical challenges and business pain points during use. To address these issues and improve system performance, the customer needs to refact and migrate the E2B infrastructure platform to the AWS cloud environment.

Customer pain points

The workload and difficulty of E2B infrastructure layer transformation are high: differences in API of cloud service providers, complexity of resource mapping, and differences in network architecture require rewriting automation scripts and using CICD technology stack to improve the automation level of application deployment and reduce long-term maintenance costs

E2B system architecture transformation is complex and difficult to maintain: official Orchestrator 、envd 、Template-manager 、API、Docker-reverse-proxy、Session-proxy、client-proxy When the service encounters high concurrency on GCP, its stability is poor and it is difficult for the customer team to effectively grasp it

Poor stability of GCP environment: Encountering D-state issues in GCP environment, customers urgently need a solution that can perform hot migration of sandboxes to avoid system interruptions

Solution

The E2B Infra platform adopts the multi tier cloud native architecture of AWS Northern Virginia region, builds infrastructure through CloudFormation/Terraform/Packer, deploys three-tier networks (public/private/database subnets) across dual availability zones, uses Nomad orchestration service clusters and configures ALB load balancing and Auto Scaling elastic scaling, and uses multi AZ RDS PostgreSQL and ElastiCache Redis for the data layer. At the same time, it integrates CloudFront CDN and Datadog+Loki+Grafana+OpenTelemetry full chain monitoring system to achieve a highly available and scalable enterprise level platform architecture.

Sandbox Hot Migration Solution:
Based on in-depth analysis of customer business scenarios, a complete Sandbox hot migration solution was designed and implemented. This solution effectively addresses the core pain points of customers in terms of business continuity, ensuring the stability of the business during the migration process, while significantly improving user experience and service quality

Flexible automatic expansion and contraction solution:
Based on AWS ASG, more intelligent capacity management has been implemented, and the system can automatically scale according to load, cope with traffic fluctuations more calmly, and achieve significant cost optimization effects

Project results

  1. Significant improvement in development and operation efficiency: Customers can focus on the core business development of AI intelligent agents without having to invest energy in dealing with the complexity of underlying infrastructure. Through automated deployment, intelligent operation and maintenance tools, and real-time monitoring and alarm platforms, the development cycle has been shortened by more than 40%, and the operation and maintenance efficiency has been improved by more than 50%, significantly reducing technical barriers and maintenance costs
  2. Comprehensive guarantee of business continuity: Build a multi regional disaster recovery architecture to achieve uninterrupted operation of business 24/7. Provide heat migration capability, completely solve the D-state problem in GCP environment, equipped with a 24/7 professional technical support team to ensure business stability and reliability
  3. Excellent high concurrency processing capability: supports large-scale concurrent sandbox creation, easily responding to sudden traffic demands of AI applications
  4. Significant performance improvement: The modified sandbox outperforms the original GCP environment in core operations such as creation, pause, and restore, providing users with a smoother experience

About Ultrapower

Beijing Ultrapower Software Co., Ltd. (referred to as "Ultrapower") was founded in 2001 and is one of the first companies listed on the Growth Enterprise Market in China (stock code: 300002). Ultrapower adheres to the core values of "mutual respect, trustworthiness, and enabling others" and the development strategy of "innovation-driven and global layout", aiming to become a leading enterprise in the digital economy with continuous innovation capabilities, create industry-quality products, support customers' improvement, and promote industry development.

Ultrapower has always been focused on providing professional consulting services and technical support services for Amazon Web Services, including cloud consulting, migration, billing services, CDN services, cloud hosting services, and enterprise overseas expansion services. Moreover, it provides professional solutions for industries such as manufacturing, media, gaming, and e-commerce, and offers high-quality services to customers based on the concept of delivering value.