Serverless computing, or Function-as-a-Service (FaaS), enables a new way of building and scaling applications by allowing users to deploy fine-grained functions while providing fully-managed resource provisioning and auto-scaling. Custom FaaS container support is gaining traction as it enables better control over OSes, versioning, and tooling for modernizing FaaS applications. However, providing rapid container provisioning introduces non-trivial challenges for FaaS providers, since container provisioning is costly, and real-world FaaS workloads exhibit highly dynamic patterns.
In this paper, we design FaaSNet, a highly-scalable middleware system for accelerating FaaS container provisioning. FaaSNet is driven by the workload and infrastructure requirements of the FaaS platform at one of the world's largest cloud providers, Alibaba Cloud Function Compute. FaaSNet enables scalable container provisioning via a lightweight, adaptive function tree (FT) structure. FaaSNet uses an I/O efficient, on-demand fetching mechanism to further reduce provisioning costs at scale. We implement and integrate FaaSNet in Alibaba Cloud Function Compute. Evaluation results show that FaaSNet: (1) finishes provisioning 2,500 function containers on 1,000 virtual machines in 8.3 seconds, (2) scales 13.4× and 16.3× faster than Alibaba Cloud's current FaaS platform and a state-of-the-art P2P container registry (Kraken), respectively, and (3) sustains a bursty workload using 75.2% less time than an optimized baseline.
more »
« less
Wasserstein Adversarial Transformer for Cloud Workload Prediction
Predictive VM (Virtual Machine) auto-scaling is a promising technique to optimize cloud applications’ operating costs and performance. Understanding the job arrival rate is crucial for accurately predicting future changes in cloud workloads and proactively provisioning and de-provisioning VMs for hosting the applications. However, developing a model that accurately predicts cloud workload changes is extremely challenging due to the dynamic nature of cloud workloads. Long- Short-Term-Memory (LSTM) models have been developed for cloud workload prediction. Unfortunately, the state-of-the-art LSTM model leverages recurrences to predict, which naturally adds complexity and increases the inference overhead as input sequences grow longer. To develop a cloud workload prediction model with high accuracy and low inference overhead, this work presents a novel time-series forecasting model called WGAN-gp Transformer, inspired by the Transformer network and improved Wasserstein-GANs. The proposed method adopts a Transformer network as a generator and a multi-layer perceptron as a critic. The extensive evaluations with real-world workload traces show WGAN- gp Transformer achieves 5× faster inference time with up to 5.1% higher prediction accuracy against the state-of-the-art. We also apply WGAN-gp Transformer to auto-scaling mechanisms on Google cloud platforms, and the WGAN-gp Transformer-based auto-scaling mechanism outperforms the LSTM-based mechanism by significantly reducing VM over-provisioning and under-provisioning rates.
more »
« less
- Award ID(s):
- 1943046
- NSF-PAR ID:
- 10392506
- Date Published:
- Journal Name:
- Proceedings of the AAAI Conference on Artificial Intelligence
- Volume:
- 36
- Issue:
- 11
- ISSN:
- 2159-5399
- Page Range / eLocation ID:
- 12433 to 12439
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The advances of Machine Learning (ML) have sparked a growing demand of ML-as-a-Service: developers train ML models and publish them in the cloud as online services to provide low-latency inference at scale. The key challenge of ML model serving is to meet the response-time Service-Level Objectives (SLOs) of inference workloads while minimizing the serving cost. In this paper, we tackle the dual challenge of SLO compliance and cost effectiveness with MArk (Model Ark), a general-purpose inference serving system built in Amazon Web Services (AWS). MArk employs three design choices tailor-made for inference workload. First, MArk dynamically batches requests and opportunistically serves them using expensive hardware accelerators (e.g., GPU) for improved performance-cost ratio. Second, instead of relying on feedback control scaling or over-provisioning to serve dynamic workload, which can be too slow or too expensive for inference serving, MArk employs predictive autoscaling to hide the provisioning latency at low cost. Third, given the stateless nature of inference serving, MArk exploits the flexible, yet costly serverless instances to cover the occasional load spikes that are hard to predict. We evaluated the performance of MArk using several state-of-the-art ML models trained in popular frameworks including TensorFlow, MXNet, and Keras. Compared with the premier industrial ML serving platform SageMaker, MArk reduces the serving cost up to 7.8× while achieving even better latency performance.more » « less
-
null (Ed.)Predicting workload behavior during execution is essential for dynamic resource optimization of processor systems. Early studies used simple prediction algorithms such as a history tables. More recently, researchers have applied advanced machine learning regression techniques. Workload prediction can be cast as a time series forecasting problem. Time series forecasting is an active research area with recent advances that have not been studied in the context of workload prediction. In this paper, we first perform a comparative study of representative time series forecasting techniques to predict the dynamic workload of applications running on a CPU. We adapt state-of-the-art matrix profile and dynamic linear models (DLMs) not previously applied to workload prediction and compare them against traditional SVM and LSTM models that have been popular for handling non-stationary data. We find that all time series forecasting models struggle to predict abrupt workload changes. These changes occur because workloads go through phases, where prior work has studied workload phase detection, classification and prediction. We propose a novel approach that combines time series forecasting with phase prediction. We process each phase as a separate time series and train one forecasting model per phase. At runtime, forecasts from phase-specific models are selected and combined based on the predicted phase behavior. We apply our approach to forecasting of SPEC workloads running on a state-of-the-art Intel machine. Our results show that an LSTM-based phase-aware predictor can forecast workload CPI with less than 8% mean absolute error while reducing CPI error by more than 12% on average compared to a non-phase-aware approach.more » « less
-
Serverless computing is gaining popularity for machine learning (ML) serving workload due to its autonomous resource scaling, easy to use and pay-per-use cost model. Existing serverless platforms work well for image-based ML inference, where requests are homogeneous in service demands. That said, recent advances in natural language processing could not fully benefit from existing serverless platforms as their requests are intrinsically heterogeneous. Batching requests for processing can significantly increase ML serving efficiency while reducing monetary cost, thanks to the pay-per-use pricing model adopted by serverless platforms. Yet, batching heterogeneous ML requests leads to additional computation overhead as small requests need to be "padded" to the same size as large requests within the same batch. Reaching effective batching decisions (i.e., which requests should be batched together and why) is non-trivial: the padding overhead coupled with the serverless auto-scaling forms a complex optimization problem. To address this, we develop Multi-Buffer Serving (MBS), a framework that optimizes the batching of heterogeneous ML inference serving requests to minimize their monetary cost while meeting their service level objectives (SLOs). The core of MBS is a performance and cost estimator driven by analytical models supercharged by a Bayesian optimizer. MBS is prototyped and evaluated on AWS using bursty workloads. Experimental results show that MBS preserves SLOs while outperforming the state-of-the-art by up to 8 x in terms of cost savings while minimizing the padding overhead by up to 37 x with 3 x less number of serverless function invocations.more » « less
-
Traditionally, HPC workloads have been deployed in bare-metal clusters; but the advances in virtualization have led the pathway for these workloads to be deployed in virtualized clusters. However, HPC cluster administrators/providers still face challenges in terms of resource elasticity and virtual machine (VM) provisioning at large-scale, due to the lack of coordination between a traditional HPC scheduler and the VM hypervisor (resource management layer). This lack of interaction leads to low cluster utilization and job completion throughput. Furthermore, the VM provisioning delays directly impact the overall performance of jobs in the cluster. Hence, there is a need for effectively provisioning virtualized HPC clusters, which can best-utilize the physical hardware with minimal provisioning overheads.Towards this, we propose Multiverse, a VM provisioning framework, which can dynamically spawn VMs for incoming jobs in a virtualized HPC cluster, by integrating the HPC scheduler along with VM resource manager. We have implemented this framework on the Slurm scheduler along with the vSphere VM resource manager. In order to reduce the VM provisioning overheads, we use instant cloning which shares both the disk and memory with the parent VM, when compared to full VM cloning which has to boot-up a new VM from scratch. Measurements with real-world HPC workloads demonstrate that, instant cloning is 2.5× faster than full cloning in terms of VM provisioning time. Further, it improves resource utilization by up to 40%, and cluster throughput by up to 1.5×, when compared to full clone for bursty job arrival scenarios.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1609/aaai.v36i11.21509
