Understand the challenges and potentials of serverless computing for realtime networked multimedia

冰山软件是一家总部位于不列颠哥伦比亚省萨里市的加拿大软件公司，为基于互联网的实时通信需求提供先进的解决方案。云计算在支持冰冻山的服务方面发挥着至关重要的作用，这些服务可以扩展到极大数量的并发用户。受流行的基础设施即服务(IaaS)云的运营和定价模式的制约，用户不可避免地需要自行预订和管理云资源，并为空闲的服务器时间付费。此外，这些重量级云服务器实例无法灵活扩容以满足工作负载。无服务器计算作为一种新的执行模式，是为了克服传统云计算模式的不足而应运而生的。它将服务器配置和管理的任务从开发人员转移到平台，并以更精细的时间表向用户收费，实现了长期承诺的即付即用定价策略。在这个项目中，一个由SFU和冰山的研究人员和工程师组成的多学科团队将合作，了解将无服务器计算应用于网络多媒体服务的潜力和挑战。我们将对领先的无服务器计算平台进行广泛的测量，并量化功能资源分配(例如内存大小)、功能实现方案(例如调用外部API)、分配的工作负载(例如帧数量)和平台选择对典型网络多媒体应用的影响。然后，我们将开始开发新的解决方案，将无服务器计算集成到冰山的产品中。

Frozen Mountain Software, a Canadian software company based on Surrey, BC, provides advanced solutions for Internet-based realtime communication needs. Cloud computing plays an essential role in supporting Frozen Moutain's services, which handle scaling to extremely large numbers of simultaneous users. Constrained by the operation and pricing model of the prevalent Infrastructure as a Service (IaaS) cloud, users inevitably need to reserve and manage cloud resources by themselves and pay for the idle server time. Further, these heavy-weight cloud server instances cannot scale up or down flexibly to meet the workloads. As a new execution model, serverless computing is emerging to overcome the weaknesses of traditional cloud computing models. It offloads the tasks of server provisioning and management from developers to platforms, and charges users at a finer-grained timescale, achieving the long-promised "pay-as-you-go" pricing strategy.In this project, a multi-disciplinary team involving researchers and engineers from both SFU and Frozen Mountain will work together to understand the potentials and challenges on applying serverless computing for networked multimedia services. We will conduct extensive measurements on leading serverless computing platforms and quantify the impact of function resource allocation (e.g., memory size), function implementation schemes (e.g., calling external APIs), allocated workload (e.g., the number of frames), and platform selection on typical networked multimedia applications. We will then start to develop novel solutions that integrate serverless computing into Frozen Mountain's products.