Description

Do you want to shape the future of virtualized (SDN) networking in the world’s biggest public cloud?

The Amazon Elastic Compute Cloud (EC2) VPC Packet Pipeline team owns the packet pipeline that runs right beneath all our customer's EC2 VPC instances, adding features like firewalling (security groups), billing and monitoring as we touch every single packet on every single host across our massive worldwide fleet. Our vision is to combine the performance of bare metal networking while maintaining all the benefits of the cloud, including delivering features not possible on bare metal i.e. true Software Defined Networking (SDN).

We are seeking a Software Engineer to join our EC2 Nitro Fairness team, where you'll help ensure equitable resource allocation in our multi-tenant cloud infrastructure while maintaining the single-tenant experience customers expect. In this role, you'll develop sophisticated algorithms and systems that intelligently manage shared compute, memory,
and network resources across diverse customer workloads. You'll design and implement sophisticated fairness algorithms, real-time resource schedulers, and performance isolation systems that operate at microsecond latencies across millions of concurrent EC2 instances within the Nitro hypervisor. This position requires deep understanding of virtualization technologies, distributed systems, and performance optimization, as you'll be directly impacting how millions of EC2 instances share underlying hardware while appearing completely isolated to end users. You'll collaborate closely with hardware teams, hypervisor engineers, and service owners to implement fairness policies that balance resource utilization efficiency with strict performance guarantees.

Key job responsibilities
Technical Leadership & Development:
• Lead architecture decisions for fairness algorithms and resource allocation systems within the Nitro hypervisor
• Mentor junior engineers on low-level C/Rust programming and embedded system development practices
• Conduct design and code reviews for performance-critical fairness mechanisms
• Drive implementation of resource isolation and scheduling policies in embedded Nitro environments

Performance Analysis & Optimization:
• Analyze customer workload patterns to identify fairness violations and noisy neighbor scenarios
• Design and implement hardware-aware fairness algorithms leveraging CPU cache locality, memory bandwidth, and network resources
• Collaborate with hardware engineering teams to optimize packet-pipeline software and hardware offloads for equitable resource sharing
• Set and achieve fairness performance goals for new instance types and Nitro hardware generations

Cross-Team Collaboration & Strategy:
• Partner with hardware/firmware teams on co-designing fairness enforcements
• Collaborate with other engineering vertical teams to ensure fairness features integrate seamlessly across the stack
• Communicate fairness requirements and technical designs to senior management and product teams
• Evaluate emerging technologies for ML-based instance types and their impact on resource fairness

System Design & Implementation:
• Develop real-time fairness monitoring and enforcement systems operating at microsecond latencies
• Implement dynamic resource allocation policies that adapt to changing workload characteristics
• Design telemetry and observability systems to measure and validate fairness across multi-tenant environments