Toward Sustainable Serverless Computing

Abstract

Although serverless computing generally involves executing short-lived "functions", the increasing migration to this computing paradigm requires careful consideration of energy and power requirements. Serverless computing is also viewed as an economically-driven computational approach, often influenced by the cost of computation, as users are charged for per-sub-second use of computational resources rather than the coarse-grained charging that is common with virtual machines and containers. To ensure that the startup times of serverless functions do not discourage their use, resource providers need to keep these functions hot, often by passing in synthetic data. We describe the real power consumption characteristics of serverless, based on execution traces reported in the literature, and describe potential strategies (some adopted from existing VM and container-based approaches) that can be used to reduce the energy overheads of serverless execution. Our analysis is, purposefully, biased towards the use of machine learning workloads as: (i) such workloads are increasingly being used widely across different applications; (ii) functions that implement machine learning algorithms can range in complexity from long-running (deep learning) vs. short-running (inference only), enabling us to consider serverless across a variety of possible execution behaviours. The general findings are also easily translatable to other domains.

AI for Next Generation Computing: Emerging Trends and Future Directions

“…• Sustainability: An increasing amount of data generated on the edge is being submitted to the cloud for ML training/inference; the overall transmission energy is around 30% of the total energy requirements of datacenters globally and rising fast. Research is required to ensure the upcoming serverless paradigm is sustainable with a focus on power-off techniques, increased computational density via smart workload consolidation, submitting the kilobytes-long function to the data vs., submitting terrabytes of data to the function, and effectively combining serverless edge resources for multitenant clients-e.g., by sharing artificial neural network layers with an acceptable tradeoff in accuracy [148].…”

DIFFUSE: A DIstributed and decentralized platForm enabling Function composition in Serverless Environments

“…The increasing demand for ICT services has driven cloud providers to further differentiate their offerings, lowering the barrier of entry to services to better match the emerging market needs [1,2]. As a result, we see an increasing trend toward solutions that guarantee a faster time to market by making the cloud easier to program and more economically viable [3].…”

“…At the same time, the opaqueness of functions makes it more challenging to implement efficient mechanisms for process coordination, such as broadcast, aggregation and shuffling, which are common communication primitives in distributed systems. This aspect is particularly relevant when considering machine learning and big data analytics workloads [3]. Despite these open issues, major hyperscalers have embraced the paradigm and alredy provide managed FaaS offerings like e.g., Microsoft Azure, Google Cloud Platform, Amazon AWS, etc., and many open-source projects are under active development as well [4,5].…”

Demo:Smart Waste Disposal with Edge-Cloud Continuum Architecture

“…Recent progress in greener cloud computing models has already led to visions and proposals such as sustainable serverless computing that favours certain combinations of edge and cloud computing, in particular for smart or intelligent computations that require machine inference of knowledge based on sensed data [7]. Serverless computing is often realised through a function-based architecture, with functions running at the edge, in the cloud or within message brokers.…”