skip to main content
Explore Scholarly Publications and Datasets in the NSF-PAR

Title: Scheduling Challenges for Variable Capacity Resources
Datacenter scheduling research often assumes resources as a constant quantity, but increasingly external factors shape capacity dynamically, and beyond the control of an operator. Based on emerging examples, we define a new, open research challenge: the variable capacity resource scheduling problem. The objective here is effective resource utilization despite sudden, perhaps large, changes in the available resources. We define the problem, key dimensions of resource capacity variation, and give specific examples that arise from the natural world (carbon- content, power price, datacenter cooling, and more). Key dimensions of the resource capacity variation include dynamic range, frequency, and structure. With these dimensions, an empirical trace can be character- ized, abstracting it from the many possible important real-world generators of variation. Resource capacity variation can arise from many causes including weather, market prices, renewable energy, carbon emission targets, and internal dynamic power management constraints. We give examples of three dif- ferent sources of variable capacity. Finally, we show variable resource capacity presents new scheduling challenges. We show how variation can cause significant performance degra- dation in existing schedulers, with up to 60% goodput reduction. Further, initial results also show intelligent scheduling techniques can be helpful. These insights show the promise and opportunity for future scheduling more » studies on resource volatility. « less
Authors:
;
Editors:
Cirne, Walfredo; Rodrigo, Gonzalo P.; Klusáček, Dalibor
Award ID(s):
1832230
Publication Date:
NSF-PAR ID:
10293164
Journal Name:
Proceedings of JSSPP21
Volume:
12985
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ( , Workshop on Job Scheduling for Parallel Processing (JSSPP))
    Cirne, Walfredo ; Rodrigo, Gonzalo P. ; Klusáček, Dalibor (Ed.)
    Datacenter scheduling research often assumes resources as a constant quantity, but increasingly external factors shape capacity dynamically, and beyond the control of an operator. Based on emerging examples, we define a new, open research challenge: the variable capacity resource scheduling problem. The objective here is effective resource utilization despite sudden, perhaps large, changes in the available resources. We define the problem, key dimensions of resource capacity variation, and give specific examples that arise from the natural world (carboncontent, power price, datacenter cooling, and more). Key dimensions of the resource capacity variation include dynamic range, frequency, and structure. With these dimensions,more »an empirical trace can be characterized, abstracting it from the many possible important real-world generators of variation. Resource capacity variation can arise from many causes including weather, market prices, renewable energy, carbon emission targets, and internal dynamic power management constraints. We give examples of three different sources of variable capacity. Finally, we show variable resource capacity presents new scheduling challenges. We show how variation can cause significant performance degradation in existing schedulers, with up to 60% goodput reduction. Further, initial results also show intelligent scheduling techniques can be helpful. These insights show the promise and opportunity for future scheduling studies on resource volatility.« less
  2. ; ; ; ( , 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]more »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.« less
  3. ; ; ; ; ( , IEEE International Conference on Cluster Computing (CLUSTER))
    Compute heterogeneity is increasingly gaining prominence in modern datacenters due to the addition of accelerators like GPUs and FPGAs. We observe that datacenter schedulers are agnostic of these emerging accelerators, especially their resource utilization footprints, and thus, not well equipped to dynamically provision them based on the application needs. We observe that the state-of-the-art datacenter schedulers fail to provide fine-grained resource guarantees for latency-sensitive tasks that are GPU-bound. Specifically for GPUs, this results in resource fragmentation and interference leading to poor utilization of allocated GPU resources. Furthermore, GPUs exhibit highly linear energy efficiency with respect to utilization and hence proactivemore »management of these resources is essential to keep the operational costs low while ensuring the end-to-end Quality of Service (QoS) in case of user-facing queries.Towards addressing the GPU orchestration problem, we build Knots, a GPU-aware resource orchestration layer and integrate it with the Kubernetes container orchestrator to build Kube- Knots. Kube-Knots can dynamically harvest spare compute cycles through dynamic container orchestration enabling co-location of latency-critical and batch workloads together while improving the overall resource utilization. We design and evaluate two GPU-based scheduling techniques to schedule datacenter-scale workloads through Kube-Knots on a ten node GPU cluster. Our proposed Correlation Based Prediction (CBP) and Peak Prediction (PP) schemes together improves both average and 99 th percentile cluster-wide GPU utilization by up to 80% in case of HPC workloads. In addition, CBP+PP improves the average job completion times (JCT) of deep learning workloads by up to 36% when compared to state-of-the-art schedulers. This leads to 33% cluster-wide energy savings on an average for three different workloads compared to state-of-the-art GPU-agnostic schedulers. Further, the proposed PP scheduler guarantees the end-to-end QoS for latency-critical queries by reducing QoS violations by up to 53% when compared to state-of-the-art GPU schedulers.« less
  4. ; ( , HPDC '21)
    Deep learning (DL) is a popular technique for building models from large quantities of data such as pictures, videos, messages generated from edges devices at rapid pace all over the world. It is often infeasible to migrate large quantities of data from the edges to centralized data center(s) over WANs for training due to privacy, cost, and performance reasons. At the same time, training large DL models on edge devices is infeasible due to their limited resources. An attractive alternative for DL training distributed data is to use micro-clouds---small-scale clouds deployed near edge devices in multiple locations. However, micro-clouds presentmore »the challenges of both computation and network resource heterogeneity as well as dynamism. In this paper, we introduce DLion, a new and generic decentralized distributed DL system designed to address the key challenges in micro-cloud environments, in order to reduce overall training time and improve model accuracy. We present three key techniques in DLion: (1) Weighted dynamic batching to maximize data parallelism for dealing with heterogeneous and dynamic compute capacity, (2) Per-link prioritized gradient exchange to reduce communication overhead for model updates based on available network capacity, and (3) Direct knowledge transfer to improve model accuracy by merging the best performing model parameters. We build a prototype of DLion on top of TensorFlow and show that DLion achieves up to 4.2X speedup in an Amazon GPU cluster, and up to 2X speed up and 26% higher model accuracy in a CPU cluster over four state-of-the-art distributed DL systems.« less
  5. ; ( , Transportation Science)
    Although urban transit systems (UTS) often have fixed vehicle capacity and relatively constant departure headways, they may need to accommodate dramatically fluctuating passenger demands over space and time, resulting in either excessive passenger waiting or vehicle capacity and energy waste. Therefore, on the one hand, optimal operations of UTS rely on accurate modeling of passenger queuing dynamics, which is particularly complex on a multistop transit corridor. On the other hand, capacities of transit vehicles can be made variable and adaptive to time-variant passenger demand so as to minimize energy waste, especially with the emergence of modular vehicle technologies. This papermore »investigates operations of a multistop transit corridor in which vehicles may have different capacities across dispatches. We specify skewed time coordinates to simplify the problem structure while incorporating traffic congestion. Then, we propose a mixed integer linear programming model that determines the optimal dynamic headways and vehicle capacities over the study time horizon to minimize the overall system cost for the transit corridor. In particular, the proposed model considers a realistic multistop first-in, first-out (MSFIFO) rule that gives the same boarding priority to passengers arriving at a station in the same time interval yet with different destinations. A customized dynamic programming (DP) algorithm is proposed to solve this model efficiently. To circumvent the rapid increase of the state space of a typical DP algorithm, we analyze the theoretical properties of the investigated problem and identify upper and lower bounds to a feasible solution. The bounds largely reduce the state space during the DP iterations and greatly improve the efficiency of the proposed DP algorithm. The state dimensions are also reduced with the MSFIFO rule such that all queues with different destinations at the same origin can be tracked with a single boarding position state variable at each stage. A hypothetical example and a real-world case study show that the proposed DP algorithm greatly outperforms a state-of-the-art commercial solver (Gurobi) in both solution quality and time.« less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email