اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدPerseus: Characterizing Performance and Cost of Multi-Tenant Serving for CNN Models
67
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
Deep learning models are increasingly used for end-user applications, supporting both novel features such as facial recognition, and traditional features, e.g. web search. To accommodate high inference throughput, it is common to host a single pre-trained Convolutional Neural Network (CNN) in dedicated cloud-based servers with hardware accelerators such as Graphics Processing Units (GPUs). However, GPUs can be orders of magnitude more expensive than traditional Central Processing Unit (CPU) servers. These resources could also be under-utilized facing dynamic workloads, which may result in inflated serving costs. One potential way to alleviate this problem is by allowing hosted models to share the underlying resources, which we refer to as multi-tenant inference serving. One of the key challenges is maximizing the resource efficiency for multi-tenant serving given hardware with diverse characteristics, models with unique response time Service Level Agreement (SLA), and dynamic inference workloads. In this paper, we present Perseus, a measurement framework that provides the basis for understanding the performance and cost trade-offs of multi-tenant model serving. We implemented Perseus in Python atop a popular cloud inference server called Nvidia TensorRT Inference Server. Leveraging Perseus, we evaluated the inference throughput and cost for serving various models and demonstrated that multi-tenant model serving led to up to 12% cost reduction.
قيم البحث
اقرأ أيضاً
Recurrent Neural Network (RNN) applications form a major class of AI-powered, low-latency data center workloads. Most execution models for RNN acceleration break computation graphs into BLAS kernels, which lead to significant inter-kernel data moveme
nt and resource underutilization. We show that by supporting more general loop constructs that capture design parameters in accelerators, it is possible to improve resource utilization using cross-kernel optimization without sacrificing programmability. Such abstraction level enables a design space search that can lead to efficient usage of on-chip resources on a spatial architecture across a range of problem sizes. We evaluate our optimization strategy on such abstraction with DeepBench using a configurable spatial accelerator. We demonstrate that this implementation provides a geometric speedup of 30x in performance, 1.6x in area, and 2x in power efficiency compared to a Tesla V100 GPU, and a geometric speedup of 2x compared to Microsoft Brainwave implementation on a Stratix 10 FPGA.
Deep learning (DL) is becoming increasingly popular in several application domains and has made several new application features involving computer vision, speech recognition and synthesis, self-driving automobiles, drug design, etc. feasible and acc
urate. As a result, large scale on-premise and cloud-hosted deep learning platforms have become essential infrastructure in many organizations. These systems accept, schedule, manage and execute DL training jobs at scale. This paper describes the design, implementation and our experiences with FfDL, a DL platform used at IBM. We describe how our design balances dependability with scalability, elasticity, flexibility and efficiency. We examine FfDL qualitatively through a retrospective look at the lessons learned from building, operating, and supporting FfDL; and quantitatively through a detailed empirical evaluation of FfDL, including the overheads introduced by the platform for various deep learning models, the load and performance observed in a real case study using FfDL within our organization, the frequency of various faults observed including unanticipated faults, and experiments demonstrating the benefits of various scheduling policies. FfDL has been open-sourced.
Cloud GPU servers have become the de facto way for deep learning practitioners to train complex models on large-scale datasets. However, it is challenging to determine the appropriate cluster configuration---e.g., server type and number---for differe
nt training workloads while balancing the trade-offs in training time, cost, and model accuracy. Adding to the complexity is the potential to reduce the monetary cost by using cheaper, but revocable, transient GPU servers. In this work, we analyze distributed training performance under diverse cluster configurations using CM-DARE, a cloud-based measurement and training framework. Our empirical datasets include measurements from three GPU types, six geographic regions, twenty convolutional neural networks, and thousands of Google Cloud servers. We also demonstrate the feasibility of predicting training speed and overhead using regression-based models. Finally, we discuss potential use cases of our performance modeling such as detecting and mitigating performance bottlenecks.
Container technologies have been evolving rapidly in the cloud-native era. Kubernetes, as a production-grade container orchestration platform, has been proven to be successful at managing containerized applications in on-premises datacenters. However
, Kubernetes lacks sufficient multi-tenant supports by design, meaning in cloud environments, dedicated clusters are required to serve multiple users, i.e., tenants. This limitation significantly diminishes the benefits of cloud computing, and makes it difficult to build multi-tenant software as a service (SaaS) products using Kubernetes. In this paper, we propose Virtual-Cluster, a new multi-tenant framework that extends Kubernetes with adequate multi-tenant supports. Basically, VirtualCluster provides both control plane and data plane isolations while sharing the underlying compute resources among tenants. The new framework preserves the API compatibility by avoiding modifying the Kubernetes core components. Hence, it can be easily integrated with existing Kubernetes use cases. Our experimental results show that the overheads introduced by VirtualCluster, in terms of latency and throughput, is moderate.
Systems for processing big data---e.g., Hadoop, Spark, and massively parallel databases---need to run workloads on behalf of multiple tenants simultaneously. The abundant disk-based storage in these systems is usually complemented by a smaller, but m
uch faster, {em cache}. Cache is a precious resource: Tenants who get to use cache can see two orders of magnitude performance improvement. Cache is also a limited and hence shared resource: Unlike a resource like a CPU core which can be used by only one tenant at a time, a cached data item can be accessed by multiple tenants at the same time. Cache, therefore, has to be shared by a multi-tenancy-aware policy across tenants, each having a unique set of priorities and workload characteristics. In this paper, we develop cache allocation strategies that speed up the overall workload while being {em fair} to each tenant. We build a novel fairness model targeted at the shared resource setting that incorporates not only the more standard concepts of Pareto-efficiency and sharing incentive, but also define envy freeness via the notion of {em core} from cooperative game theory. Our cache management platform, ROBUS, uses randomization over small time batches, and we develop a proportionally fair allocation mechanism that satisfies the core property in expectation. We show that this algorithm and related fair algorithms can be approximated to arbitrary precision in polynomial time. We evaluate these algorithms on a ROBUS prototype implemented on Spark with RDD store used as cache. Our evaluation on a synthetically generated industry-standard workload shows that our algorithms provide a speedup close to performance optimal algorithms while guaranteeing fairness across tenants.
الأسئلة المقترحة
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
