Interactive AI-powered services require low-latency evaluation of deep neural network (DNN) models—aka ""real-time AI"". The growing demand for computationally expensive, state-of-the-art DNNs, coupled with diminishing performance gains of general-purpose architectures, has fueled an explosion of specialized Neural Processing Units (NPUs). NPUs for interactive services should satisfy two requirements: (1) execution of DNN models with low latency, high throughput, and high efficiency, and (2) flexibility to accommodate evolving state-of-the-art models (e.g., RNNs, CNNs, MLPs) without costly silicon updates. This paper describes the NPU architecture for Project Brainwave, a production-scale system for real-time AI. The Brainwave NPU achieves more than an order of magnitude improvement in latency and throughput over state-of-the-art GPUs on large RNNs at a batch size of 1. The NPU attains this performance using a single-threaded SIMD ISA paired with a distributed microarchitecture capable of dispatching over 7M operations from a single instruction. The spatially distributed microarchitecture, scaled up to 96,000 multiply-accumulate units, is supported by hierarchical instruction decoders and schedulers coupled with thousands of independently addressable high-bandwidth on-chip memories, and can transparently exploit many levels of fine-grain SIMD parallelism. When targeting an FPGA, microarchitectural parameters such as native datapaths and numerical precision can be "synthesis specialized" to models at compile time, enabling atypically high FPGA performance competitive with hardened NPUs. When running on an Intel Stratix 10 280 FPGA, the Brainwave NPU achieves performance ranging from ten to over thirty-five teraflops, with no batching, on large, memory-intensive RNNs.

A configurable cloud-scale DNN processor for real-time AI

Cloud services have recently started undergoing a major shift from monolithic applications, to graphs of hundreds or thousands of loosely-coupled microservices. Microservices fundamentally change a lot of assumptions current cloud systems are designed with, and present both opportunities and challenges when optimizing for quality of service (QoS) and cloud utilization. In this paper we explore the implications microservices have across the cloud system stack. We first present DeathStarBench, a novel, open-source benchmark suite built with microservices that is representative of large end-to-end services, modular and extensible. DeathStarBench includes a social network, a media service, an e-commerce site, a banking system, and IoT applications for coordination control of UAV swarms. We then use DeathStarBench to study the architectural characteristics of microservices, their implications in networking and operating systems, their challenges with respect to cluster management, and their trade-offs in terms of application design and programming frameworks. Finally, we explore the tail at scale effects of microservices in real deployments with hundreds of users, and highlight the increased pressure they put on performance predictability.

https://dl.acm.org/doi/pdf/10.1145/3297858.3304013

An Open-Source Benchmark Suite for Microservices and Their Hardware-Software Implications for Cloud & Edge Systems

Memristor crossbars are circuits capable of performing analog matrix-vector multiplications, overcoming the fundamental energy efficiency limitations of digital logic. They have been shown to be effective in special-purpose accelerators for a limited set of neural network applications. We present the Programmable Ultra-efficient Memristor-based Accelerator (PUMA) which enhances memristor crossbars with general purpose execution units to enable the acceleration of a wide variety of Machine Learning (ML) inference workloads. PUMA's microarchitecture techniques exposed through a specialized Instruction Set Architecture (ISA) retain the efficiency of in-memory computing and analog circuitry, without compromising programmability. We also present the PUMA compiler which translates high-level code to PUMA ISA. The compiler partitions the computational graph and optimizes instruction scheduling and register allocation to generate code for large and complex workloads to run on thousands of spatial cores. We have developed a detailed architecture simulator that incorporates the functionality, timing, and power models of PUMA's components to evaluate performance and energy consumption. A PUMA accelerator running at 1 GHz can reach area and power efficiency of 577 GOPS/s/mm 2 and 837~GOPS/s/W, respectively. Our evaluation of diverse ML applications from image recognition, machine translation, and language modelling (5M-800M synapses) shows that PUMA achieves up to 2,446× energy and 66× latency improvement for inference compared to state-of-the-art GPUs. Compared to an application-specific memristor-based accelerator, PUMA incurs small energy overheads at similar inference latency and added programmability.

PUMA: A Programmable Ultra-efficient Memristor-based Accelerator for Machine Learning Inference

The widespread application of deep learning has changed the landscape of computation in data centers. In particular, personalized recommendation for content ranking is now largely accomplished using deep neural networks. However, despite their importance and the amount of compute cycles they consume, relatively little research attention has been devoted to recommendation systems. To facilitate research and advance the understanding of these workloads, this paper presents a set of real-world, production-scale DNNs for personalized recommendation coupled with relevant performance metrics for evaluation. In addition to releasing a set of open-source workloads, we conduct in-depth analysis that underpins future system design and optimization for at-scale recommendation: Inference latency varies by 60% across three Intel server generations, batching and co-location of inference jobs can drastically improve latency-bounded throughput, and diversity across recommendation models leads to different optimization strategies.

The Architectural Implications of Facebook's DNN-Based Personalized Recommendation

Performance unpredictability is a major roadblock towards cloud adoption, and has performance, cost, and revenue ramifications. Predictable performance is even more critical as cloud services transition from monolithic designs to microservices. Detecting QoS violations after they occur in systems with microservices results in long recovery times, as hotspots propagate and amplify across dependent services. We present Seer, an online cloud performance debugging system that leverages deep learning and the massive amount of tracing data cloud systems collect to learn spatial and temporal patterns that translate to QoS violations. Seer combines lightweight distributed RPC-level tracing, with detailed low-level hardware monitoring to signal an upcoming QoS violation, and diagnose the source of unpredictable performance. Once an imminent QoS violation is detected, Seer notifies the cluster manager to take action to avoid performance degradation altogether. We evaluate Seer both in local clusters, and in large-scale deployments of end-to-end applications built with microservices with hundreds of users. We show that Seer correctly anticipates QoS violations 91% of the time, and avoids the QoS violation to begin with in 84% of cases. Finally, we show that Seer can identify application-level design bugs, and provide insights on how to better architect microservices to achieve predictable performance.

https://dl.acm.org/doi/pdf/10.1145/3297858.3304004

Seer: Leveraging Big Data to Navigate the Complexity of Performance Debugging in Cloud Microservices

To meet the computational demands required of deep learning, cloud operators are turning toward specialized hardware for improved efficiency and performance. Project Brainwave, Microsofts principal infrastructure for AI serving in real time, accelerates deep neural network (DNN) inferencing in major services such as Bings intelligent search features and Azure. Exploiting distributed model parallelism and pinning over low-latency hardware microservices, Project Brainwave serves state-of-the-art, pre-trained DNN models with high efficiencies at low batch sizes. A high-performance, precision-adaptable FPGA soft processor is at the heart of the system, achieving up to 39.5 teraflops (Tflops) of effective performance at Batch 1 on a state-of-the-art Intel Stratix 10 FPGA.

/pdf/serving-dnns-in-real-time-at-datacenter-scale-with-project-2zqss674x5.pdf

Serving DNNs in Real Time at Datacenter Scale with Project Brainwave

The thermochemical decomposition of woody biomass has been widely identified as a promising route to produce renewable biofuels. More recently, the use of molten salts in combination with pyrolysis has gathered increased interest. The molten salts may act as a solvent, a heat transfer medium, and possibly also a catalyst. In this study, we report experimental studies on a process to convert woody biomass to a liquid hydrocarbon product with a very low oxygen content using molten salt pyrolysis (350–450 °C and atmospheric pressure) followed by subsequent catalytic conversions of the liquids obtained by pyrolysis. Pyrolysis of woody biomass in molten salt (ZnCl2/NaCl/KCl with a molar composition of 60:20:20) resulted in a liquid yield of 46 wt % at a temperature of 450 °C and a molten salt/biomass ratio of 10:1 (mass). The liquids are highly enriched in furfural (13 wt %) and acetic acid (14 wt %). To reduce complexity and experimental issues related to the production of sufficient amounts of pyrolysis oils for further catalytic upgrading, model studies were performed to convert both compounds to hydrocarbons using a three-step catalytic approach, viz., (i) ketonization of acetic acid to acetone, (ii) cross-aldol condensation between acetone and furfural to C8–C13 products, followed by (iii) a two-stage catalytic hydrotreatment of the latter to liquid hydrocarbons. Ketonization of acetic acid to acetone was studied in a continuous setup over a ceria–zirconia-based catalyst at 250 °C. The catalyst showed no signs of deactivation over a period of 230 h while also achieving high selectivity toward acetone. Furfural was shown to have a negative effect on the catalyst performance, and as such, a separation step is required after pyrolysis to obtain an acetic-acid-enriched fraction. The cross-aldol condensation reaction between acetone and furfural was studied in a batch using a commercial Mg/Al hydrotalcite as the catalyst. Furfural was quantitatively converted with over 90% molar selectivity toward condensed products with a carbon number between C8 and C13. The two-stage hydrotreatment of the condensed product consisted of a stabilization step using a Ni-based Picula catalyst and a further deep hydrotreatment over a NiMo catalyst, in both batch setups. The final product with a residual 1.5 wt % O is rich in (cyclo)alkanes and aromatic hydrocarbons. The overall carbon yield for the four-step approach, from pinewood biomass to middle distillates, is 21%, assuming that separation of furfural and acetic acid after the pyrolysis step can be performed without losses.

https://pubs.acs.org/doi/pdf/10.1021/acs.energyfuels.2c02044

Balaji Sridharan

Papers

Serving DNNs in Real Time at Datacenter Scale with Project Brainwave

Novel Route to Produce Hydrocarbons from Woody Biomass Using Molten Salts

Pyrolysis of LignoBoost lignin in ZnCl2-KCl-NaCl molten salt media: Insights into process-pyrolysis oil yield and composition relations