Managing demand driven infra cost regarding product growth is essential to a company’s expansion. Demand and Efficiency management is key to enable services to scale reliably and efficiently. COVID-19 caused unexpected growth on user traffic and shortage on hardware supply, making demand and efficiency management more critical than ever.

Multi-tenant services have its unique challenges in demand and efficiency management in terms of attribution (especially interweaving social graph services) and holding all users accountable.

Meta has tens of large multi-tenant services, and each is used by hundreds of Meta teams. Using Meta web tier as an example, every day there are thousands of code changes and feature roll outs. Managing demand and efficiency at this scale and development pace is challenging. To tackle that, we built a one-stop shop to cover end to end management flow including (1) quota management/enforcement and admission control as a safe net to manage the overall demand (2) pre-production and production regression detection to prevent adding new cost (3) optimization framework to reduce existing cost. In this presentation, we will introduce this toolkit, how it supports web demand and efficiency management, and how we scale the tooling and process to manage demand across Meta large multi-tenant services.

Meta’s graph store TAO, like many other distributed data stores, has traditionally prioritized availability, efficiency, and horizontal scalability over strong consistency or isolation guarantees to serve its large read-dominant workloads. As product developers built diverse applications on top of this system, we have seen an increasing desire for transactional semantics. In this talk, we first characterize developer desires for transactions that emerged over the years and describe the current failure-atomic (i.e., write) transactions offered by TAO. We highlight the need for atomic visibility in a read transaction API with a measurement study on potential anomalies that occur without stronger isolation for reads. We then present the RAMP-TAO protocol, a variation based on the Read Atomic Multi-Partition (RAMP) protocol, that ensures atomic visibility for a read-optimized workload and an eventually consistent system. We demonstrate that this protocol results in minimal overhead, and our design can be feasibly deployed in production.

Security is everyone’s job, not just those who have security in their job title. In this session you’ll hear how a positive security culture helps you scale the efforts of your security humans and accelerate the thing that we are all trying to do. Ship securely. You’ll hear some approaches that AWS takes in their distributed environment & how you can apply those to your own organization.

Effective incident management is key to limiting the disruption caused by an incident and restoring normal business operations as quickly as possible. If you haven’t gamed out your response to potential incidents in advance, principled incident management can go out the window in real-life situations. This talk walks through a portrait of an incident that spirals out of control due to ad hoc incident management practices, outlines a well-managed approach to the incident, and reviews how the same incident might have played out if handled with well-functioning incident management. The talk introduces protocols for incident management, typical roles and heuristics for when to declare an incident and form a command post.

LinkedIn uses Terraform as a service(TFaas) to enable infrastructure-as-code uniformly across its teams. In this talk, learn why LinkedIn built its own service. We'll dive into the architecture and see how we solve complex issues at scale around storing state, drift detection, and remediation as we discuss the company-wide adoption strategy. Finally, we'll show you some of the common use cases, including bootstrapping new fabrics and migrating on-prem infrastructure to Kubernetes using this platform.

Featuring All Speakers & Moderated by Liane Praza