Spanning FaaS Platform Scheduling Across Infrastructures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
FaaS infrastructures, especially private clouds, face challenges in scalability and availability due to transient issues like network congestion and hardware failures, which conventional systems cannot mitigate by scheduling functions across different infrastructures.
Innovation Solution
A spanning FaaS service platform that communicates with multiple FaaS infrastructures, retrieves metadata and capability information, and dynamically selects the best infrastructure for function execution based on defined criteria, allowing offloading to public clouds when private cloud resources are constrained.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If functions are executed on a single FaaS infrastructure (private or public cloud), then the system is simple to manage, but scalability and availability are limited due to transient problems like network congestion, excessive server load, and hardware failures
Solution Approach 1:
The patent combines multiple FaaS infrastructures (both private and public clouds) into a unified federated FaaS system. The gateway aggregates function invocation requests and distributes them across multiple backend infrastructures, allowing the system to leverage the combined resources of multiple clouds to improve availability and reliability while maintaining a relatively simple interface for developers.
Solution Approach 2:
The gateway acts as an intermediary layer between function invocations and the multiple FaaS infrastructures. It receives function execution requests, determines the appropriate target infrastructure based on current conditions, and manages the distribution of workloads. This mediator approach allows the system to improve reliability across infrastructures without requiring complex changes to individual cloud systems.
2Reliability
If private cloud resources are pre-allocated to ensure availability, then service level agreements can be met, but scalability is limited when transient problems occur
Solution Approach 1:
The system implements dynamic workload distribution where the gateway can adjust in real-time which FaaS infrastructure receives function invocations. When transient problems occur in one infrastructure, the gateway dynamically redirects requests to other available infrastructures. This dynamic adaptation allows the system to maintain service level agreements while scaling flexibly in response to changing conditions without being constrained by static pre-allocation.
Solution Approach 2:
The federated FaaS system creates a universal platform that can execute functions across multiple types of infrastructures (private clouds, public clouds) through a common interface. This multi-functionality allows the system to meet service level agreements by distributing workloads across diverse infrastructure types, improving both reliability and scalability simultaneously.
3Productivity
If functions are scheduled across multiple FaaS infrastructures dynamically, then scalability and availability are enhanced, but the complexity of managing and selecting the best infrastructure increases
Solution Approach 1:
The gateway implements self-service scheduling by automatically monitoring the status and performance of multiple FaaS infrastructures and making intelligent routing decisions without external intervention. It autonomously determines which infrastructure is best suited for each function invocation based on current conditions, thereby achieving enhanced scalability while keeping the management interface simple. The system serves itself by automatically adapting to infrastructure changes and optimizing workload distribution.
Data Source
AI summary
Techniques for scheduling and executing functions across a plurality of different Functions-as-a-Service (FaaS) infrastructures are provided. In one set of embodiments, a computer system can determine that a function has been invoked, where the computer system implements a spanning FaaS service platform that is communicatively coupled with the plurality of different FaaS infrastructures. In response, the computer system can retrieve metadata associated with the function, where the metadata includes criteria or policies indicating how the function should be scheduled for execution, and can retrieve information associated with each of the plurality of different FaaS infrastructures, where the information includes capabilities or characteristics of each FaaS infrastructure. The computer system can then select a FaaS infrastructure in the plurality of different FaaS infrastructures based on the retrieved metadata and the retrieved information and can schedule the function for execution on the selected FaaS infrastructure.