More Like this

1. Expanding an HPC Cluster to Support the Computational Demands of Digital Pathology

   https://doi.org/10.1109/SPMB.2018.8615614  

   Campbell, C. ; Mecca, N. ; Duong, T. ; Obeid, I. ; Picone, J. ( December 2018 , IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, Iyad ; Selesnick, Ivan ; Picone, Joseph (Ed.)

   The goal of this work was to design a low-cost computing facility that can support the development of an open source digital pathology corpus containing 1M images [1]. A single image from a clinical-grade digital pathology scanner can range in size from hundreds of megabytes to five gigabytes. A 1M image database requires over a petabyte (PB) of disk space. To do meaningful work in this problem space requires a significant allocation of computing resources. The improvements and expansions to our HPC (highperformance computing) cluster, known as Neuronix [2], required to support working with digital pathology fall into two broad categories: computation and storage. To handle the increased computational burden and increase job throughput, we are using Slurm [3] as our scheduler and resource manager. For storage, we have designed and implemented a multi-layer filesystem architecture to distribute a filesystem across multiple machines. These enhancements, which are entirely based on open source software, have extended the capabilities of our cluster and increased its cost-effectiveness. Slurm has numerous features that allow it to generalize to a number of different scenarios. Among the most notable is its support for GPU (graphics processing unit) scheduling. GPUs can offer a tremendous performance increase in machine learning applications [4] and Slurm’s built-in mechanisms for handling them was a key factor in making this choice. Slurm has a general resource (GRES) mechanism that can be used to configure and enable support for resources beyond the ones provided by the traditional HPC scheduler (e.g. memory, wall-clock time), and GPUs are among the GRES types that can be supported by Slurm [5]. In addition to being able to track resources, Slurm does strict enforcement of resource allocation. This becomes very important as the computational demands of the jobs increase, so that they have all the resources they need, and that they don’t take resources from other jobs. It is a common practice among GPU-enabled frameworks to query the CUDA runtime library/drivers and iterate over the list of GPUs, attempting to establish a context on all of them. Slurm is able to affect the hardware discovery process of these jobs, which enables a number of these jobs to run alongside each other, even if the GPUs are in exclusive-process mode. To store large quantities of digital pathology slides, we developed a robust, extensible distributed storage solution. We utilized a number of open source tools to create a single filesystem, which can be mounted by any machine on the network. At the lowest layer of abstraction are the hard drives, which were split into 4 60-disk chassis, using 8TB drives. To support these disks, we have two server units, each equipped with Intel Xeon CPUs and 128GB of RAM. At the filesystem level, we have implemented a multi-layer solution that: (1) connects the disks together into a single filesystem/mountpoint using the ZFS (Zettabyte File System) [6], and (2) connects filesystems on multiple machines together to form a single mountpoint using Gluster [7]. ZFS, initially developed by Sun Microsystems, provides disk-level awareness and a filesystem which takes advantage of that awareness to provide fault tolerance. At the filesystem level, ZFS protects against data corruption and the infamous RAID write-hole bug by implementing a journaling scheme (the ZFS intent log, or ZIL) and copy-on-write functionality. Each machine (1 controller + 2 disk chassis) has its own separate ZFS filesystem. Gluster, essentially a meta-filesystem, takes each of these, and provides the means to connect them together over the network and using distributed (similar to RAID 0 but without striping individual files), and mirrored (similar to RAID 1) configurations [8]. By implementing these improvements, it has been possible to expand the storage and computational power of the Neuronix cluster arbitrarily to support the most computationally-intensive endeavors by scaling horizontally. We have greatly improved the scalability of the cluster while maintaining its excellent price/performance ratio [1]. 

   more » « less
2. Issues in the Reproducibility of Deep Learning Results

   https://doi.org/10.1109/SPMB47826.2019.9037840  

   Jean-Paul, S. ; Elseify, T. ; Obeid, I. ; Picone, J. ( December 2019 , IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, I. ; Selesnik, I. ; Picone, J. (Ed.)

   The Neuronix high-performance computing cluster allows us to conduct extensive machine learning experiments on big data [1]. This heterogeneous cluster uses innovative scheduling technology, Slurm [2], that manages a network of CPUs and graphics processing units (GPUs). The GPU farm consists of a variety of processors ranging from low-end consumer grade devices such as the Nvidia GTX 970 to higher-end devices such as the GeForce RTX 2080. These GPUs are essential to our research since they allow extremely compute-intensive deep learning tasks to be executed on massive data resources such as the TUH EEG Corpus [2]. We use TensorFlow [3] as the core machine learning library for our deep learning systems, and routinely employ multiple GPUs to accelerate the training process. Reproducible results are essential to machine learning research. Reproducibility in this context means the ability to replicate an existing experiment – performance metrics such as error rates should be identical and floating-point calculations should match closely. Three examples of ways we typically expect an experiment to be replicable are: (1) The same job run on the same processor should produce the same results each time it is run. (2) A job run on a CPU and GPU should produce identical results. (3) A job should produce comparable results if the data is presented in a different order. System optimization requires an ability to directly compare error rates for algorithms evaluated under comparable operating conditions. However, it is a difficult task to exactly reproduce the results for large, complex deep learning systems that often require more than a trillion calculations per experiment [5]. This is a fairly well-known issue and one we will explore in this poster. Researchers must be able to replicate results on a specific data set to establish the integrity of an implementation. They can then use that implementation as a baseline for comparison purposes. A lack of reproducibility makes it very difficult to debug algorithms and validate changes to the system. Equally important, since many results in deep learning research are dependent on the order in which the system is exposed to the data, the specific processors used, and even the order in which those processors are accessed, it becomes a challenging problem to compare two algorithms since each system must be individually optimized for a specific data set or processor. This is extremely time-consuming for algorithm research in which a single run often taxes a computing environment to its limits. Well-known techniques such as cross-validation [5,6] can be used to mitigate these effects, but this is also computationally expensive. These issues are further compounded by the fact that most deep learning algorithms are susceptible to the way computational noise propagates through the system. GPUs are particularly notorious for this because, in a clustered environment, it becomes more difficult to control which processors are used at various points in time. Another equally frustrating issue is that upgrades to the deep learning package, such as the transition from TensorFlow v1.9 to v1.13, can also result in large fluctuations in error rates when re-running the same experiment. Since TensorFlow is constantly updating functions to support GPU use, maintaining an historical archive of experimental results that can be used to calibrate algorithm research is quite a challenge. This makes it very difficult to optimize the system or select the best configurations. The overall impact of all of these issues described above is significant as error rates can fluctuate by as much as 25% due to these types of computational issues. Cross-validation is one technique used to mitigate this, but that is expensive since you need to do multiple runs over the data, which further taxes a computing infrastructure already running at max capacity. GPUs are preferred when training a large network since these systems train at least two orders of magnitude faster than CPUs [7]. Large-scale experiments are simply not feasible without using GPUs. However, there is a tradeoff to gain this performance. Since all our GPUs use the NVIDIA CUDA® Deep Neural Network library (cuDNN) [8], a GPU-accelerated library of primitives for deep neural networks, it adds an element of randomness into the experiment. When a GPU is used to train a network in TensorFlow, it automatically searches for a cuDNN implementation. NVIDIA’s cuDNN implementation provides algorithms that increase the performance and help the model train quicker, but they are non-deterministic algorithms [9,10]. Since our networks have many complex layers, there is no easy way to avoid this randomness. Instead of comparing each epoch, we compare the average performance of the experiment because it gives us a hint of how our model is performing per experiment, and if the changes we make are efficient. In this poster, we will discuss a variety of issues related to reproducibility and introduce ways we mitigate these effects. For example, TensorFlow uses a random number generator (RNG) which is not seeded by default. TensorFlow determines the initialization point and how certain functions execute using the RNG. The solution for this is seeding all the necessary components before training the model. This forces TensorFlow to use the same initialization point and sets how certain layers work (e.g., dropout layers). However, seeding all the RNGs will not guarantee a controlled experiment. Other variables can affect the outcome of the experiment such as training using GPUs, allowing multi-threading on CPUs, using certain layers, etc. To mitigate our problems with reproducibility, we first make sure that the data is processed in the same order during training. Therefore, we save the data from the last experiment and to make sure the newer experiment follows the same order. If we allow the data to be shuffled, it can affect the performance due to how the model was exposed to the data. We also specify the float data type to be 32-bit since Python defaults to 64-bit. We try to avoid using 64-bit precision because the numbers produced by a GPU can vary significantly depending on the GPU architecture [11-13]. Controlling precision somewhat reduces differences due to computational noise even though technically it increases the amount of computational noise. We are currently developing more advanced techniques for preserving the efficiency of our training process while also maintaining the ability to reproduce models. In our poster presentation we will demonstrate these issues using some novel visualization tools, present several examples of the extent to which these issues influence research results on electroencephalography (EEG) and digital pathology experiments and introduce new ways to manage such computational issues. 

   more » « less
3. A Co-Scheduling Framework for DNN Models on Mobile and Edge Devices with Heterogeneous Hardware

   https://doi.org/10.1109/TMC.2021.3107424  

   Xu, Zhiyuan ; Yang, Dejun ; Yin, Chengxiang ; Tang, Jian ; Wang, Yanzhi ; Xue, Guoliang ( August 2021 , IEEE Transactions on Mobile Computing)

   With the emergence of more and more powerful chipsets and hardware and the rise of Artificial Intelligence of Things (AIoT), there is a growing trend for bringing Deep Neural Network (DNN) models to empower mobile and edge devices with intelligence such that they can support attractive AI applications on the edge in a real-time or near real-time manner. To leverage heterogeneous computational resources (such as CPU, GPU, DSP, etc) to effectively and efficiently support concurrent inference of multiple DNN models on a mobile or edge device, we propose a novel online Co-Scheduling framework based on deep REinforcement Learning (DRL), which we call COSREL. COSREL has the following desirable features: 1) it achieves significant speedup over commonly-used methods by efficiently utilizing all the computational resources on heterogeneous hardware; 2) it leverages emerging Deep Reinforcement Learning (DRL) to make dynamic and wise online scheduling decisions based on system runtime state; 3) it is capable of making a good tradeoff among inference latency, throughput and energy efficiency; and 4) it makes no changes to given DNN models, thus preserves their accuracies. To validate and evaluate COSREL, we conduct extensive experiments on an off-the-shelf Android smartphone with widely-used DNN models to compare it with three commonly-used baselines. Our experimental results show that 1) COSREL consistently and significantly outperforms all the baselines in terms of both throughput and latency; and 2) COSREL is generally superior to all the baselines in terms of energy efficiency. 

   more » « less
4. Automating Deep Neural Network Model Selection for Edge Inference

   https://doi.org/10.1109/CogMI48466.2019.00035  

   Lu, Bingqian ; Yang, Jianyi ; Chen, Lydia Y. ; Ren, Shaolei ( December 2019 , IEEE First International Conference on Cognitive Machine Intelligence (CogMI))

   The ever increasing size of deep neural network (DNN) models once implied that they were only limited to cloud data centers for runtime inference. Nonetheless, the recent plethora of DNN model compression techniques have successfully overcome this limit, turning into a reality that DNN-based inference can be run on numerous resource-constrained edge devices including mobile phones, drones, robots, medical devices, wearables, Internet of Things devices, among many others. Naturally, edge devices are highly heterogeneous in terms of hardware specification and usage scenarios. On the other hand, compressed DNN models are so diverse that they exhibit different tradeoffs in a multi-dimension space, and not a single model can achieve optimality in terms of all important metrics such as accuracy, latency and energy consumption. Consequently, how to automatically select a compressed DNN model for an edge device to run inference with optimal quality of experience (QoE) arises as a new challenge. The state-of-the-art approaches either choose a common model for all/most devices, which is optimal for a small fraction of edge devices at best, or apply device-specific DNN model compression, which is not scalable. In this paper, by leveraging the predictive power of machine learning and keeping end users in the loop, we envision an automated device-level DNN model selection engine for QoE-optimal edge inference. To concretize our vision, we formulate the DNN model selection problem into a contextual multi-armed bandit framework, where features of edge devices and DNN models are contexts and pre-trained DNN models are arms selected online based on the history of actions and users' QoE feedback. We develop an efficient online learning algorithm to balance exploration and exploitation. Our preliminary simulation results validate our algorithm and highlight the potential of machine learning for automating DNN model selection to achieve QoE-optimal edge inference. 

   more » « less
5. Sojourn Time Minimization of Successful Jobs

   https://doi.org/10.1145/3561074.3561083  

   Yao, Yuan ; Paolieri, Marco ; Golubchik, Leana ( August 2022 , ACM SIGMETRICS Performance Evaluation Review)

   Due to a growing interest in deep learning applications [5], compute-intensive and long-running (hours to days) training jobs have become a significant component of datacenter workloads. A large fraction of these jobs is often exploratory, with the goal of determining the best model structure (e.g., the number of layers and channels in a convolutional neural network), hyperparameters (e.g., the learning rate), and data augmentation strategies for the target application. Notably, training jobs are often terminated early if their learning metrics (e.g., training and validation accuracy) are not converging, with only a few completing successfully. For this motivating application, we consider the problem of scheduling a set of jobs that can be terminated at predetermined checkpoints with known probabilities estimated from historical data. We prove that, in order to minimize the time to complete the first K successful jobs on a single server, optimal scheduling does not require preemption (even when preemption overhead is negligible) and provide an optimal policy; advantages of this policy are quantified through simulation. Related Work. While job scheduling has been investigated extensively in many scenarios (see [6] and [2] for a survey of recent result), most policies require that the cost of waiting times of each job be known at scheduling time; in contrast, in our setting the scheduler does not know which job will be the K-th successful job, and sojourn times of subsequent jobs do not contribute to the target metric. For example, [4, 3] minimize makespan (i.e., the time to complete all jobs) for known execution times and waiting time costs; similarly, Gittins index [1] and SR rank [7] minimize expected sojourn time of all jobs, i.e., both successfully completed jobs and jobs terminated early. Unfortunately, scheduling policies not distinguishing between these two types of jobs may favor jobs where the next stage is short and leads to early termination with high probability, which is an undesirable outcome in our applications of interest. 

   more » « less