Search

Deep Neural Network (DNN) inference on serverless functions is gaining prominence due to its potential for substantial budget savings. Existing works on serverless DNN inference solely optimize batching requests from one application with a single Service Level Objective (SLO) on CPU functions. However, production serverless DNN inference traces indicate that the request arrival rate of applications is surprisingly low, which inevitably causes a long batching time and SLO violations. Hence, there is an urgent need for batching multiple DNN inference requests with diverse SLOs (i.e., multi-SLO DNN inference) in serverless platforms. Moreover, the potential performance and cost benefits of deploying heterogeneous (i.e., CPU and GPU) functions for DNN inference have received scant attention. In this paper, we present HarmonyBatch, a cost-efficient resource provisioning framework designed to achieve predictable performance for multi-SLO DNN inference with heterogeneous serverless functions. Specifically, we construct an analytical performance and cost model of DNN inference on both CPU and GPU functions, by explicitly considering the GPU time-slicing scheduling mechanism and request arrival rate distribution. Based on such a model, we devise a two-stage merging strategy in HarmonyBatch to judiciously batch the multi-SLO DNN inference requests into application groups. It aims to minimize the budget of function provisioning for each application group while guaranteeing diverse performance SLOs of inference applications. We have implemented a prototype of HarmonyBatch on Alibaba Cloud Function Compute. Extensive prototype experiments with representative DNN inference workloads demonstrate that HarmonyBatch can provide predictable performance to serverless DNN inference workloads while reducing the monetary cost by up to 82.9% compared to the state-of-the-art methods., Comment: 10 pages, 14 figures, accepted by IWQOS24

GPUs are essential to accelerating the latency-sensitive deep neural network (DNN) inference workloads in cloud datacenters. To fully utilize GPU resources, spatial sharing of GPUs among co-located DNN inference workloads becomes increasingly compelling. However, GPU sharing inevitably brings severe performance interference among co-located inference workloads, as motivated by an empirical measurement study of DNN inference on EC2 GPU instances. While existing works on guaranteeing inference performance service level objectives (SLOs) focus on either temporal sharing of GPUs or reactive GPU resource scaling and inference migration techniques, how to proactively mitigate such severe performance interference has received comparatively little attention. In this paper, we propose iGniter, an interference-aware GPU resource provisioning framework for cost-efficiently achieving predictable DNN inference in the cloud. iGniter is comprised of two key components: (1) a lightweight DNN inference performance model, which leverages the system and workload metrics that are practically accessible to capture the performance interference; (2) A cost-efficient GPU resource provisioning strategy that jointly optimizes the GPU resource allocation and adaptive batching based on our inference performance model, with the aim of achieving predictable performance of DNN inference workloads. We implement a prototype of iGniter based on the NVIDIA Triton inference server hosted on EC2 GPU instances. Extensive prototype experiments on four representative DNN models and datasets demonstrate that iGniter can guarantee the performance SLOs of DNN inference workloads with practically acceptable runtime overhead, while saving the monetary cost by up to 25% in comparison to the state-of-the-art GPU resource provisioning strategies., Comment: 16 pages, 21 figures, submitted to IEEE Transactions on Parallel and Distributed Systems

Based on a simple tight-binding model, we propose a general theory of light-induced magnetic phase transition (MPT) in antiferromagnets based on the general conclusion that the bandgap of antiferromagnetic (AFM) phase is usually larger than that of ferromagnetic (FM) one in a given system. Light-induced electronic excitation prefers to stabilize the FM state over the AFM one, and once the critical photocarrier concentration ({\alpha}_c) is reached, an MPT from AFM phase to FM phase takes place. This theory has been confirmed by performing first-principles calculations on a series of two-dimensional (2D) van der Waals (vdW) antiferromagnets and a linear relationship between {\alpha}_c and the intrinsic material parameters is obtained. Importantly, our conclusion is still valid even considering the strong exciton effects during photoexcitation. Our general theory provides new ideas to realize reversible read-write operations for future memory devices., Comment: 14 pages, 3 figures

Two-dimensional (2D) ferromagnetic charge density wave (CDW), an exotic quantum state for exploring the intertwining effect between correlated charge and spin orders in 2D limit, has not been discovered in the experiments yet. Here, we propose a feasible strategy to realize 2D ferromagnetic CDWs under external fields, which is demonstrated in monolayer VSe$_2$ using first-principles calculations. Under external tensile strain, two novel ferromagnetic CDWs ($\sqrt{3}$$\times$$\sqrt{3}$ and 2$\times2\sqrt{3}$ CDWs) can be generated, accompanied by distinguishable lattice reconstructions of magnetic V atoms. Remarkably, because the driving forces for generating these two ferromagnetic CDWs are strongly spin-dependent, fundamentally different from that in conventional CDWs, the $\sqrt{3}$$\times$$\sqrt{3}$ and 2$\times2\sqrt{3}$ CDWs can exhibit two dramatically different half-metallic phases under a large strain range, along with either a flat band or a Dirac cone around Fermi level. Our proposed strategy and material demonstration may open a door to generate and manipulate correlation effect between collective charge and spin orders via external fields., Comment: 6 pages,4 figures

The intertwining of charge order, superconductivity and band topology has promoted the AV$_3$Sb$_5$ (A=K, Rb, Cs) family of materials to the center of attention in condensed matter physics. Underlying those mysterious macroscopic properties such as giant anomalous Hall conductivity (AHC) and chiral charge density wave is their nontrivial band topology. While there have been numerous experimental and theoretical works investigating the nontrivial band structure and especially the van Hove singularities, the exact topological phase of this family remains to be clarified. In this work, we identify CsV$_3$Sb$_5$ as a Dirac nodal line semimetal based on the observation of multiple Dirac nodal lines and loops close to the Fermi level. Combining photoemission spectroscopy and density functional theory, we identify two groups of Dirac nodal lines along $k_z$ direction and one group of Dirac nodal loops in the A-H-L plane. These nodal loops are located at the Fermi level within the instrumental resolution limit. Importantly, our first-principle analyses indicate that these nodal loops may be a crucial source of the mysterious giant AHC observed. Our results not only provide a clear picture to categorize the band structure topology of this family of materials, but also suggest the dominant role of topological nodal loops in shaping their transport behavior.

Quantum confinement is a restriction on the motion of electrons in a material to specific region, resulting in discrete energy levels rather than continuous energy bands. In certain materials quantum confinement could dramatically reshape the electronic structure and properties of the surface with respect to the bulk. Here, in the recently discovered kagome superconductor CsV$_3$Sb$_5$ (A=K, Rb, Cs) family of materials, we unveil the dominant role of quantum confinement in determining their surface electronic structure. Combining angle-resolved photoemission spectroscopy (ARPES) measurement and density-functional theory simulation, we report the observations of two-dimensional quantum well states due to the confinement of bulk electron pocket and Dirac cone to the nearly isolated surface layer. The theoretical calculations on the slab model also suggest that the ARPES observed spectra are almost entirely contributed by the top two layers. Our results not only explain the disagreement of band structures between the recent experiments and calculations, but also suggest an equally important role played by quantum confinement, together with strong correlation and band topology, in shaping the electronic properties of this family of materials.

We present a principled approach for designing stochastic Newton methods for solving finite sum optimization problems. Our approach has two steps. First, we re-write the stationarity conditions as a system of nonlinear equations that associates each data point to a new row. Second, we apply a Subsampled Newton Raphson method to solve this system of nonlinear equations. Using our approach, we develop a new Stochastic Average Newton (SAN) method, which is incremental by design, in that it requires only a single data point per iteration. It is also cheap to implement when solving regularized generalized linear models, with a cost per iteration of the order of the number of the parameters. We show through extensive numerical experiments that SAN requires no knowledge about the problem, neither parameter tuning, while remaining competitive as compared to classical variance reduced gradient methods (e.g. SAG and SVRG), incremental Newton and quasi-Newton methods (e.g. SNM, IQN)., Comment: Accepted at AISTATS 2022