Multi-accelerator servers are increasingly being deployed in shared multi-tenant environments (such as in cloud data centers) in order to meet the demands of large-scale compute-intensive workloads. In addition, these accelerators are increasingly being inter-connected in complex topologies and workloads are exhibiting a wider variety of inter-accelerator communication patterns. However, existing allocation policies are ill-suited for these emerging use-cases. Specifically, this work identifies that multi-accelerator workloads are commonly fragmented leading to reduced bandwidth and increased latency for inter-accelerator communication. We propose Multi-Accelerator Pattern Allocation (MAPA), a graph pattern mining approach towards providing generalized allocation support for allocating multi-accelerator workloads on multi-accelerator servers. We demonstrate that MAPA is able to improve the execution time of multi-accelerator workloads and that MAPA is able to provide generalized benefits across various accelerator topologies. Finally, we demonstrate a speedup of 12.4% for 75th percentile of jobs with the worst case execution time reduced by up to 35% against baseline policy using MAPA.
While providing low latency is a fundamental requirement in deploying recommendation services, achieving high resource utility is also crucial in cost-effectively maintaining the datacenter. Co-locating multiple workers of a model is an effective way to maximize query-level parallelism and server throughput, but the interference caused by concurrent workers at shared resources can prevent server queries from meeting its SLA. Hera utilizes the heterogeneous memory requirement of multi-tenant recommendation models to intelligently determine a productive set of co-located models and its resource …
Bolted is a new architecture for a bare metal cloud with the goal of providing security-sensitive customers of a cloud the same level of security and control that they can obtain in their own private data centers. It allows tenants to elastically allocate secure resources within a cloud while being protected from other previous, current, and future tenants of the cloud. The provisioning of a new server to a tenant isolates a bare metal server, only allowing it to communicate with other tenant's servers once its critical firmware and software have been attested to the tenant. Tenants, rather th…
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 infl…
Edge Computing (EC) allows users to access computing resources at the network frontier, which paves the way for deploying delay-sensitive applications such as Mobile Augmented Reality (MAR). Under the EC paradigm, MAR users connect to the EC server, open sessions and send continuously frames to be processed. The EC server sends back virtual information to enhance the human perception of the world by merging it with the real environment. Resource allocation arises as a critical challenge when several MAR Service Providers (SPs) compete for limited resources at the edge of the network. In this p…
Participating in demand response programs is a promising tool for reducing energy costs in data centers by modulating energy consumption. Towards this end, data centers can employ a rich set of resource management knobs, such as workload shifting and dynamic server provisioning. Nonetheless, these knobs may not be readily available in a cloud data center (CDC) that serves cloud tenants/users, because workloads in CDCs are managed by tenants themselves who are typically charged based on a usage-based or flat-rate pricing and often have no incentive to cooperate with the CDC operator for demand …
Ecomap: Sustainability-Driven Optimization of Multi-Tenant DNN Execution on Edge Servers
School of Electrical, Computer and Biomedical Engineering, Southern Illinois University（南伊利诺伊大学电气、计算机与生物医学工程学院）；The University of Texas at Austin（德克萨斯大学奥斯汀分校）
In cloud computing, network Denial of Service (DoS) attacks are well studied and defenses have been implemented, but severe DoS attacks on a victim's working memory by a single hostile VM are not well understood. Memory DoS attacks are Denial of Service (or Degradation of Service) attacks caused by contention for hardware memory resources on a cloud server. Despite the strong memory isolation techniques for virtual machines (VMs) enforced by the software virtualization layer in cloud servers, the underlying hardware memory layers are still shared by the VMs and can be exploited by a clever att…
With CPU scaling slowing down in today's data centers, more functionalities are being offloaded from the CPU to auxiliary devices. One such device is the SmartNIC, which is being increasingly adopted in data centers. In today's cloud environment, VMs on the same server can each have their own network computation (or network tasks) or workflows of network tasks to offload to a SmartNIC. These network tasks can be dynamically added/removed as VMs come and go and can be shared across VMs. Such dynamism demands that a SmartNIC not only schedules and processes packets but also manages and executes …
Business Intelligence (BI) has gained a new lease of life through Cloud computing as its demand for unlimited hardware and platform resources expandability is fulfilled by the Cloud elasticity features. BI can be seamlessly deployed on the Cloud given that its multilayered model coincides with the Cloud multilayer models. It is considered by many Cloud service providers as one of the prominent applications services on public, outsourced private and outsourced community Clouds. However, in the shared domains of Cloud computing, BI is exposed to security and privacy threats by virtue of exploits…
Under the paradigm of Edge Computing (EC), a Network Operator (NO) deploys computational resources at the network edge and let third-party Service Providers (SPs) run on top of them, as tenants. Besides the clear advantages for SPs and final users thanks to the vicinity of computation nodes, a NO aims to allocate edge resources in order to increase its own utility, including bandwidth saving, operational cost reduction, QoE for its users, etc. However, while the number of third-party services competing for edge resources is expected to dramatically grow, the resources deployed cannot increase …
We revisit multi-tenant network virtualization in data centers, and make the case for tenant-specific virtual switches. In particular, tenant-specific virtual switches allow cloud providers to extend fine-grained billing (known, e.g., from serverless architectures) to the network, accounting not only for IO, but also CPU or energy. We sketch an architecture and present economical motivation and recent technological enablers. We also find that virtual switches today do not offer sufficient multi-tenancy and can introduce artificial performance bottlenecks, e.g., in load balancers. We conclude b…
Powered by advances in deep learning (DL) techniques, machine learning and artificial intelligence have achieved astonishing successes. However, the rapidly growing needs for DL also led to communication- and resource-intensive distributed training jobs for large-scale DL training, which are typically deployed over GPU clusters. To sustain the ever-increasing demand for DL training, the so-called "ring-all-reduce" (RAR) technologies have recently emerged as a favorable computing architecture to efficiently process network communication and computation load in GPU clusters. The most salient fea…
University of Washington（华盛顿大学）；Microsoft Research（微软研究院）
Distributed deep neural network (DDNN) training constitutes an increasingly important workload that frequently runs in the cloud. Larger DNN models and faster compute engines are shifting DDNN training bottlenecks from computation to communication. This paper characterizes DDNN training to precisely pinpoint these bottlenecks. We found that timely training requires high performance parameter servers (PSs) with optimized network stacks and gradient processing pipelines, as well as server and network hardware with balanced computation and communication resources. We therefore propose PHub, a hig…
Since emerging edge applications such as Internet of Things (IoT) analytics and augmented reality have tight latency constraints, hardware AI accelerators have been recently proposed to speed up deep neural network (DNN) inference run by these applications. Resource-constrained edge servers and accelerators tend to be multiplexed across multiple IoT applications, introducing the potential for performance interference between latency-sensitive workloads. In this paper, we design analytic models to capture the performance of DNN inference workloads on shared edge accelerators, such as GPU and ed…
