Search for: All records

Specialized accelerators such as GPUs, TPUs, FPGAs, and custom ASICs have been increasingly deployed to train deep learning models. These accelerators exhibit heterogeneous performance behavior across model architectures. Existing schedulers for clusters of accelerators, which are used to arbitrate these expensive training resources across many users, have shown how to optimize for various multi-job, multiuser objectives, like fairness and makespan. Unfortunately, existing schedulers largely do not consider performance heterogeneity. In this paper, we propose Gavel, a heterogeneity-aware scheduler that systematically generalizes a wide range of existing scheduling policies. Gavel expresses these policies as optimization problems and then systematically transforms these problems into heterogeneity-aware versions using an abstraction we call effective throughput. Gavel then uses a round-based scheduling mechanism to ensure jobs receive their ideal allocation given the target scheduling policy. Gavel’s heterogeneity-aware policies allow a heterogeneous cluster to sustain higher input load, and improve end objectives such as makespan and average job completion time by 1.4⇥ and 3.5⇥ compared to heterogeneity-agnostic policies. 

Cloud providers offer instances with similar compute capabilities (for example, instances with different generations of GPUs like K80s, P100s, V100s) across many regions, availability zones, and on-demand and spot markets, with prices governed independently by individual supplies and demands. In this paper, using machine learning model training as an example application, we explore the potential cost reductions possible by leveraging this cross-cloud instance market. We present quantitative results on how the prices of cloud instances change with time, and how total costs can be decreased by considering this dynamic pricing market. Our preliminary experiments show that a) the optimal instance choice for a model is dependent on both the objective (e.g., cost, time, or combination) and the model’s performance characteristics, b) the cost of moving training jobs between instances is cheap, c) jobs do not need to be preempted more frequently than once a day to leverage the benefits from spot instance price variations, and d) the cost of training a model can be decreased by as much as 3.5× compared to a static policy. We also look at contexts where users specify higherlevel objectives over collections of jobs, show examples of policies for these contexts, and discuss additional challenges involved in making these cost reductions viable. 

As specialized hardware accelerators such as GPUs become increasingly popular, developers are looking for ways to target these platforms with high-level APIs. One promising approach is kernel libraries such as PyTorch or cuML, which provide interfaces that mirror CPU-only counterparts such as NumPy or Scikit-Learn. Unfortunately, these libraries are hard to develop and to adopt incrementally: they only support a subset of their CPU equivalents, only work with datasets that fit in device memory, and require developers to reason about data placement and transfers manually. To address these shortcomings, we present a new approach called offload annotations (OAs) that enables heterogeneous GPU computing in existing workloads with few or no code modifications. An annotator annotates the types and functions in a CPU library with equivalent kernel library functions and provides an offloading API to specify how the inputs and outputs of the function can be partitioned into chunks that fit in device memory and transferred between devices. A runtime then maps existing CPU functions to equivalent GPU kernels and schedules execution, data transfers and paging. In data science workloads using CPU libraries such as NumPy and Pandas, OAs enable speedups of up to 1200⇥ and a median speedup of 6.3⇥ by transparently offloading functions to a GPU using existing kernel libraries. In many cases, OAs match the performance of handwritten heterogeneous implementations. Finally, OAs can automatically page data in these workloads to scale to datasets larger than GPU memory, which would need to be done manually with most current GPU libraries. 

We present POSH, a framework that accelerates shell applications with I/O-heavy components, such as data analytics with command-line utilities. Remote storage such as networked filesystems can severely limit the performance of these applications: data makes a round trip over the network for relatively little computation at the client. Reducing the data movement by moving the code to the data can improve performance. POSH automatically optimizes unmodified I/O-intensive shell applications running over remote storage by offloading the I/O-intensive portions to proxy servers closer to the data. A proxy can run directly on a storage server, or on a machine closer to the storage layer than the client. POSH intercepts shell pipelines and uses metadata called annotations to decide where to run each command within the pipeline. We address three principal challenges that arise: an annotation language that allows POSH to understand which files a command will access, a scheduling algorithm that places commands to minimize data movement, and a system runtime to execute a distributed schedule but retain local semantics. We benchmark POSH on real shell pipelines such as image processing, network security analysis, log analysis, distributed system debugging, and git. We find that POSH provides speedups ranging from 1.6× to 15× compared to NFS, without requiring any modifications to the applications. 

Systems for ML inference are widely deployed today, but they typically optimize ML inference workloads using techniques designed for conventional data serving workloads and miss critical opportunities to leverage the statistical nature of ML. In this paper, we present WILLUMP, an optimizer for ML inference that introduces two statistically-motivated optimizations targeting ML applications whose performance bottleneck is feature computation. First, WILLUMP automatically cascades feature computation for classification queries: WILLUMP classifies most data inputs using only high-value, low-cost features selected through empirical observations of ML model performance, improving query performance by up to 5× without statistically significant accuracy loss. Second, WILLUMP accurately approximates ML top-K queries, discarding low-scoring inputs with an automatically constructed approximate model and then ranking the remainder with a more powerful model, improving query performance by up to 10× with minimal accuracy loss. WILLUMP automatically tunes these optimizations’ parameters to maximize query performance while meeting an accuracy target. Moreover, WILLUMP complements these statistical optimizations with compiler optimizations to automatically generate fast inference code for ML applications. We show that WILLUMP improves the end-to-end performance of real-world ML inference pipelines curated from major data science competitions by up to 16× without statistically significant loss of accuracy.