Topics and Sessions

The main theme of exploration would be "How would X need to change to support the requirements of functions and serverless computing?"

Some of the broad set of questions to consider include, but are not limited to:

• New datacenter architectures and networking interconnects
• Synergy between serverless computing and disaggregated network architectures
• Routing and link layer protocols to support function/lambda performance, scale up, isolation
• Programmable fabrics and control planes that can support agility needed by functions
• Revisiting transport and network layer protocol design
• Revisiting old abstractions, such as the socket API
• Jointly scheduling functions/lambdas with network resources
• Function isolation and security, accounting, necessary new support at network and application levels
• Novel networking services built using serverless platforms, e.g., stateless network functions
• Programming language support for serverless networking needs
• Network subsystem support for ultra-low latency, high throughput, and rapid elasticity

The aim is to understand how the serverless paradigm may force us to rethink existing approaches to designing, implementing and deploying applications and infrastructure. Here is a rough categorization of the above topics into sessions:

Hardware (2 sessions)

(potential topics: switching; interconnects; AQM; disaggregation; optics; what abstractions should the network provide? role of programmable network hardware)

Protocols (2 sessions)

(potential topics: routing control plane; transport, flow and congestion control, socket API, addressing and naming)

Security and availability (2 session)

(potential topics: isolation, access controls, NFs, fault tolerance, function warm up vs. impact on availability, role of the network in above aspects)

Business models (1 session)

(potential topics: “killer apps”, different models of serverless akin to PaaS, IaaS, etc in the Cloud world), design challenges that arise)

Systems (1 session)

(potential topics: granularity of computation; compute control plane (scheduler scheduling and executing DAGs of functions); how to ensure low latency and high throughput task execution; storage; operating systems support; packaging, … )

Monitoring and tracing (1 session)

(potential topics: cross layer; causal tracing; abstractions for gathering statistics)

Programming and formal methods (1 session)

(new programming models; what is the role of verification and formal methods -- does the simplicity of functions make these problems simpler? are network verification etc. impacted by serverless?)