Current serverless Function-as-a-Service (FaaS) platforms generally use simple, classic scheduling algorithms for distributing function invocations while ignoring FaaS characteristics such as rapid changes in resource utilization and the freeze-thaw life cycle. In this paper, we present FaaSRank, a function scheduler for serverless FaaS platforms based on information monitored from servers and functions. FaaSRank automatically learns scheduling policies through experience using reinforcement learning (RL) and neural networks supported by our novel Score-Rank-Select architecture. We implemented FaaSRank in Apache OpenWhisk, an open source FaaS platform, and evaluated performance against other baseline schedulers including OpenWhisk's default scheduler on two 13-node OpenWhisk clusters. For training and evaluation, we adapted real-world serverless workload traces provided by Microsoft Azure. For the duration of test workloads, FaaSRank sustained on average a lower number of inflight invocations 59.62 % and 70.43 % as measured on two clusters respectively. We also demonstrate the generalizability of FaaSRank for any workload. When trained using a composite of 50 episodes each for 10 distinct random workloads, FaaSRank reduced average function completion time by 23.05% compared to OpenWhisk's default scheduler.
more »
« less
This content will become publicly available on September 23, 2025
Dejavu: Reinforcement Learning-based Cloud Scheduling with Demonstration and Competition
As Cloud's adoption surges across industries, the limitations of its default scheduler, particularly on large scales or for jobs outside of its initial design scope, have become increasingly prominent. While the default schedulers in various cloud platforms were primarily engineered to focus on simple and predictable tasks, reinforcement learning (RL)-based schedulers are attracting attention as they can predict a larger and more diverse cloud environment. Nevertheless, there are practical constraints to the use of RL. Retraining for adaptation is necessary for each new environment, and exploration taken during each training may lead to unexpected performance degradation at runtime. To address these issues, this paper presents Dejavu which combines reinforcement learning with neural networks to learn and resolve scheduling problems more effectively. To tackle the extended training time and performance degradation by unexpected explorations, we apply pretraining using Demonstrations from existing heuristics. This guides the RL agent to explore in a safe and efficient manner. Furthermore, we design a robust reward function to push Dejavu to compete with and eventually outperform, the exploited heuristics and other baselines. The experimental results demonstrate the efficacy of Dejavu, showing remarkable improvements in key metrics. Compared to the default scheduler, it boosts resource utilization by 6 % and shortens scheduling time by 3% during the scheduling period.
more »
« less
- Award ID(s):
- 1908910
- PAR ID:
- 10553557
- Publisher / Repository:
- IEEE
- Date Published:
- ISBN:
- 979-8-3503-6399-9
- Page Range / eLocation ID:
- 469 to 478
- Subject(s) / Keyword(s):
- Training Degradation Measurement Industries Job shop scheduling Neural networks Reinforcement learning Smart systems Resource management Surges Container-based cloud Scheduling Reinforcement learning Offline RL
- Format(s):
- Medium: X
- Location:
- Seoul, Korea, Republic of
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Serverless Function-as-a-Service (FaaS) offers improved programmability for customers, yet it is not server-“less” and comes at the cost of more complex infrastructure management (e.g., resource provisioning and scheduling) for cloud providers. To maintain function service-level objectives (SLOs) and improve resource utilization efficiency, recent research has been focused on applying online learning algorithms such as reinforcement learning (RL) to manage resources. Compared to rule-based solutions with heuristics, RL-based approaches eliminate humans in the loop and avoid the painstaking generation of heuristics. Despite the initial success of applying RL, we first show in this paper that the state-of-the-art single-agent RL algorithm (S-RL) suffers up to 4.8x higher p99 function latency degradation on multi-tenant serverless FaaS platforms compared to isolated environments and is unable to converge during training. We then design and implement a scalable and incremental multi-agent RL framework based on Proximal Policy Optimization (SIMPPO). Our experiments on widely used serverless benchmarks demonstrate that in multi-tenant environments, SIMPPO enables each RL agent to efficiently converge during training and provides online function latency performance comparable to that of S-RL trained in isolation (which we refer to as the baseline for assessing RL performance) with minor degradation (<9.2%). In addition, SIMPPO reduces the p99 function latency by 4.5x compared to S-RL in multi-tenant cases.more » « less
-
Kernel task scheduling is important for application performance, adaptability to new hardware, and complex user requirements. However, developing, testing, and debugging new scheduling algorithms in Linux, the most widely used cloud operating system, is slow and difficult. We developed Enoki, a framework for high velocity development of Linux kernel schedulers. Enoki schedulers are written in safe Rust, and the system supports live upgrade of new scheduling policies into the kernel, userspace debugging, and bidirectional communication with applications. A scheduler implemented with Enoki achieved near identical performance (within 1% on average) to the default Linux scheduler CFS on a wide range of benchmarks. Enoki is also able to support a range of research schedulers, specifically the Shinjuku scheduler, a locality aware scheduler, and the Arachne core arbiter, with good performance.more » « less
-
Serverless Function-As-A-Service (FaaS) is an emerging cloud computing paradigm that frees application developers from infrastructure management tasks such as resource provisioning and scaling. To reduce the tail latency of functions and improve resource utilization, recent research has been focused on applying online learning algorithms such as reinforcement learning (RL) to manage resources. Compared to existing heuristics-based resource management approaches, RL-based approaches eliminate humans in the loop and avoid the painstaking generation of heuristics. In this paper, we show that the state-of-The-Art single-Agent RL algorithm (S-RL) suffers up to 4.6x higher function tail latency degradation on multi-Tenant serverless FaaS platforms and is unable to converge during training. We then propose and implement a customized multi-Agent RL algorithm based on Proximal Policy Optimization, i.e., multi-Agent PPO (MA-PPO). We show that in multi-Tenant environments, MA-PPO enables each agent to be trained until convergence and provides online performance comparable to S-RL in single-Tenant cases with less than 10% degradation. Besides, MA-PPO provides a 4.4x improvement in S-RL performance (in terms of function tail latency) in multi-Tenant cases.more » « less
-
With the wide adoption of deep neural network (DNN) models for various applications, enterprises, and cloud providers have built deep learning clusters and increasingly deployed specialized accelerators, such as GPUs and TPUs, for DNN training jobs. To arbitrate cluster resources among multi-user jobs, existing schedulers fall short, either lacking fine-grained heterogeneity awareness or hardly generalizable to various scheduling policies. To fill this gap, we propose a novel design of a task-level heterogeneity-aware scheduler, Hadar, based on an online optimization framework that can express other scheduling algorithms. Hadar leverages the performance traits of DNN jobs on a heterogeneous cluster, characterizes the task-level performance heterogeneity in the optimization problem, and makes scheduling decisions across both spatial and temporal dimensions. The primal-dual framework is employed, with our design of a dual subroutine, to solve the optimization problem and guide the scheduling design. Extensive trace-driven simulations with representative DNN models have been conducted to demonstrate that Hadar improves the average job completion time (JCT) by 3× over an Apache YARN-based resource manager used in production. Moreover, Hadar outperforms Gavel[1], the state-of-the-art heterogeneity-aware scheduler, by 2.5× for the average JCT, and shortens the queuing delay by 13% and improve FTF (Finish-Time-Fairness) by 1.5%.more » « less
