Follow the Sun through the Clouds

Shen, Zhiming; Jia, Qin; Sela, Gur-Eyal; Rainero, Ben; Song, Weijia; Renesse, Robbert van; Weatherspoon, Hakim

doi:10.1145/2987550.2987561

Cited by 27 publications

(3 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One key characteristic of server workload data is periodicity. This periodicity is seen in many server workloads [5,30,40], including serverless workloads [20,29]. Identifying the meaningful periodicities will inform scheduling and autoscaling design choices such as the architecture of the model used for forecasting these traces, how often to invoke it to make a prediction, and how long into the future it should predict.…”

Section: Are Functions Invocation Counts Periodic?mentioning

confidence: 99%

How Does It Function? Characterizing Long-term Trends in Production Serverless Workloads

Joosen,

Hassan,

Asenov

et al. 2023

Proceedings of the 2023 ACM Symposium on Cloud Computing

View full text Add to dashboard Cite

This paper releases and analyzes two new Huawei cloud serverless traces. The traces span a period of over 7 months with over 1.4 trillion function invocations combined. The first trace is derived from Huawei's internal workloads and contains detailed per-second statistics for 200 functions running across multiple Huawei cloud data centers. The second trace is a representative workload from Huawei's public FaaS platform. This trace contains per-minute arrival rates for over 5000 functions running in a single Huawei data center. We present the internals of a production FaaS platform by characterizing resource consumption, cold-start times, programming languages used, periodicity, per-second versus per-minute burstiness, correlations, and popularity. Our findings show that there is considerable diversity in how serverless functions behave: requests vary by up to 9 orders of magnitude across functions, with some functions executed over 1 billion times per day; scheduling time, execution time and cold-start distributions vary across 2 to 4 orders of magnitude and have very long tails; and function invocation counts demonstrate strong periodicity for many individual functions and on an aggregate level. Our analysis also highlights the need for further research in estimating resource reservations and time-series prediction to account for the huge diversity in how serverless functions behave.

show abstract

Section: Are Functions Invocation Counts Periodic?mentioning

confidence: 99%

How Does It Function? Characterizing Long-term Trends in Production Serverless Workloads

Joosen,

Hassan,

Asenov

et al. 2023

Proceedings of the 2023 ACM Symposium on Cloud Computing

View full text Add to dashboard Cite

show abstract

“…In an NFS-based and iSCSI-based solution the WAN VM migration experiences increased latencies, low bandwidth, and high internet cost in accessing a shared disk image if the shared storage is located in a different data centre or region. To address this issue [31] proposed Supercloud using nested virtualisation with a geo-replicated image file storage that maintains the trade-off between performance and cost. They designed an image storage that tries to propagate only data which is frequently accessed and it proactively transmits data before migration is triggered.…”

Section: User-centric Inter-cloud Migrationmentioning

confidence: 99%

Using Docker Swarm with a User-Centric Decision-Making Framework for Cloud Application Migration

Barlaskar

Kilpatrick

Spence

et al. 2018

Cloud Computing and Service Science

View full text Add to dashboard Cite

“…First, we should use historical data of computing availability as well as weather forecasts to estimate how much computing resources will be available in each home and then use this knowledge as a basis for scheduling. Moreover, the system would need to have a mechanism to move a job to a new home if it does not have available resources now or in forecasted future [47,27]. Naturally, unnecessary moving of jobs is also wasteful for resource use.…”

Section: Control and Sensory Systemsmentioning

confidence: 99%

Sunburn—using excess energy of small-scale production for distributed computing

Nurminen

Niemi²,

Strandman

et al. 2017

Energy Efficiency

View full text Add to dashboard Cite

The Sunburn solar computer system is based on the idea of consuming the excess electricity of photovoltaic energy systems for useful computing. In this way, the proposed approach allows us to move a part of computing capacity from dedicated data centers to individual solar energy systems and thereby reduce the energy consumption of the data centers. In principle, the proposed mechanism converts the excess electrical energy into valuable computation. The results of computation or data processing are easier to store and transfer than produced electricity. In small scale systems, the benefit from storing or selling the excess electricity is small, whereas the benefit from selling the generated data is potentially larger, as we demonstrate in this study. Finally, the technical feasibility of the solution is illustrated by constructing and evaluating a prototype implementation using excess solar energy for distributed BOINC computing. Keywords distributed computing • solar energy • green computing • energy storage • scientific computing • distributed data center Address(es) of author(s) should be given

show abstract

Follow the Sun through the Clouds

Cited by 27 publications

References 31 publications

How Does It Function? Characterizing Long-term Trends in Production Serverless Workloads

How Does It Function? Characterizing Long-term Trends in Production Serverless Workloads

Using Docker Swarm with a User-Centric Decision-Making Framework for Cloud Application Migration

Sunburn—using excess energy of small-scale production for distributed computing

Contact Info

Product

Resources

About