Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new usertf-Darshan: Understanding Fine-grained I/O Performance in Machine Learning Workloads
173
0
0.0
(
0
)
Added by
Steven W. D. Chien
Publication date
2020
fields
Informatics Engineering
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
Machine Learning applications on HPC systems have been gaining popularity in recent years. The upcoming large scale systems will offer tremendous parallelism for training through GPUs. However, another heavy aspect of Machine Learning is I/O, and this can potentially be a performance bottleneck. TensorFlow, one of the most popular Deep-Learning platforms, now offers a new profiler interface and allows instrumentation of TensorFlow operations. However, the current profiler only enables analysis at the TensorFlow platform level and does not provide system-level information. In this paper, we extend TensorFlow Profiler and introduce tf-Darshan, both a profiler and tracer, that performs instrumentation through Darshan. We use the same Darshan shared instrumentation library and implement a runtime attachment without using a system preload. We can extract Darshan profiling data structures during TensorFlow execution to enable analysis through the TensorFlow profiler. We visualize the performance results through TensorBoard, the web-based TensorFlow visualization tool. At the same time, we do not alter Darshans existing implementation. We illustrate tf-Darshan by performing two case studies on ImageNet image and Malware classification. We show that by guiding optimization using data from tf-Darshan, we increase POSIX I/O bandwidth by up to 19% by selecting data for staging on fast tier storage. We also show that Darshan has the potential of being used as a runtime library for profiling and providing information for future optimization.
rate research
Read More
Recent trend towards increasing large machine learning models require both training and inference tasks to be distributed. Considering the huge cost of training these models, it is imperative to unlock optimizations in computation and communication to obtain best performance. However, current logical separation between computation and communication kernels in deep learning frameworks misses the optimization opportunities across such barrier. Breaking this abstraction with a holistic consideration can provide many optimizations to provide performance improvements in distributed workloads. Manually applying these optimizations needs modifications in underlying computation and communication libraries for each scenario, which is time consuming and error-prone. Therefore, we present CoCoNeT, with a DSL to express a program with both computation and communication. CoCoNeT contains several machine learning aware transformations to optimize a program and a compiler to generate high performance kernels. Providing both computation and communication as first class constructs allows users to work on a high-level abstraction and apply powerful optimizations, such as fusion or overlapping of communication and computation. CoCoNeT enables us to optimize data-, model-and pipeline-parallel workloads in large language models with only a few lines of code. Experiments show CoCoNeT significantly outperforms state-of-the-art distributed machine learning implementations.
We describe TF-Replicator, a framework for distributed machine learning designed for DeepMind researchers and implemented as an abstraction over TensorFlow. TF-Replicator simplifies writing data-parallel and model-parallel research code. The same models can be effortlessly deployed to different cluster architectures (i.e. one or many machines containing CPUs, GPUs or TPU accelerators) using synchronous or asynchronous training regimes. To demonstrate the generality and scalability of TF-Replicator, we implement and benchmark three very different models: (1) A ResNet-50 for ImageNet classification, (2) a SN-GAN for class-conditional ImageNet image generation, and (3) a D4PG reinforcement learning agent for continuous control. Our results show strong scalability performance without demanding any distributed systems expertise of the user. The TF-Replicator programming model will be open-sourced as part of TensorFlow 2.0 (see https://github.com/tensorflow/community/pull/25).
Applying machine learning techniques to the quickly growing data in science and industry requires highly-scalable algorithms. Large datasets are most commonly processed data parallel distributed across many nodes. Each nodes contribution to the overall gradient is summed using a global allreduce. This allreduce is the single communication and thus scalability bottleneck for most machine learning workloads. We observe that frequently, many gradient values are (close to) zero, leading to sparse of sparsifyable communications. To exploit this insight, we analyze, design, and implement a set of communication-efficient protocols for sparse input data, in conjunction with efficient machine learning algorithms which can leverage these primitives. Our communication protocols generalize standard collective operations, by allowing processes to contribute arbitrary sparse input data vectors. Our generic communication library, SparCML, extends MPI to support additional features, such as non-blocking (asynchronous) operations and low-precision data representations. As such, SparCML and its techniques will form the basis of future highly-scalable machine learning frameworks.
Scientific computing sometimes involves computation on sensitive data. Depending on the data and the execution environment, the HPC (high-performance computing) user or data provider may require confidentiality and/or integrity guarantees. To study the applicability of hardware-based trusted execution environments (TEEs) to enable secure scientific computing, we deeply analyze the performance impact of AMD SEV and Intel SGX for diverse HPC benchmarks including traditional scientific computing, machine learning, graph analytics, and emerging scientific computing workloads. We observe three main findings: 1) SEV requires careful memory placement on large scale NUMA machines (1$times$$-$3.4$times$ slowdown without and 1$times$$-$1.15$times$ slowdown with NUMA aware placement), 2) virtualization$-$a prerequisite for SEV$-$results in performance degradation for workloads with irregular memory accesses and large working sets (1$times$$-$4$times$ slowdown compared to native execution for graph applications) and 3) SGX is inappropriate for HPC given its limited secure memory size and inflexible programming model (1.2$times$$-$126$times$ slowdown over unsecure execution). Finally, we discuss forthcoming new TEE designs and their potential impact on scientific computing.
Big data applications and analytics are employed in many sectors for a variety of goals: improving customers satisfaction, predicting market behavior or improving processes in public health. These applications consist of complex software stacks that are often run on cloud systems. Predicting execution times is important for estimating the cost of cloud services and for effectively managing the underlying resources at runtime. Machine Learning (ML), providing black box solutions to model the relationship between application performance and system configuration without requiring in-detail knowledge of the system, has become a popular way of predicting the performance of big data applications. We investigate the cost-benefits of using supervised ML models for predicting the performance of applications on Spark, one of todays most widely used frameworks for big data analysis. We compare our approach with textit{Ernest} (an ML-based technique proposed in the literature by the Spark inventors) on a range of scenarios, application workloads, and cloud system configurations. Our experiments show that Ernest can accurately estimate the performance of very regular applications, but it fails when applications exhibit more irregular patterns and/or when extrapolating on bigger data set sizes. Results show that our models match or exceed Ernests performance, sometimes enabling us to reduce the prediction error from 126-187% to only 5-19%.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
