The performance of modern Big Data frameworks, e.g. Spark, depends greatly on high-speed storage and shuffling, which impose a significant memory burden on production data centers. In many production situations, the persistence and shuffling intensive applications can suffer a major performance loss due to lack of memory. Thus, the common practice is usually to over-allocate the memory assigned to the data workers for production applications, which in turn reduces overall resource utilization. One efficient way to address the dilemma between the performance and cost efficiency of Big Data applications is through data center computing resource disaggregation. This paper proposes and implements a system that incorporates the Spark Big Data framework with a novel in-memory distributed file system to achieve memory disaggregation for data persistence and shuffling. We address the challenge of optimizing performance at affordable cost by co-designing the proposed in-memory distributed file system with large-volume DIMM-based persistent memory (PMEM) and RDMA technology. The disaggregation design allows each part of the system to be scaled independently, which is particularly suitable for cloud deployments. The proposed system is evaluated in a production-level cluster using real enterprise-level Spark production applications. The results of an empirical evaluation show that the system can achieve up to a 3.5- fold performance improvement for shuffle-intensive applications with the same amount of memory, compared to the default Spark setup. Moreover, by leveraging PMEM, we demonstrate that our system can effectively increase the memory capacity of the computing cluster with affordable cost, with a reasonable execution time overhead with respect to using local DRAM only.
more »
« less
Exploring the Potential of Elastic Computing Clusters in Geo-Distributed Data Centers with Fast Fabric Interconnection
Large-scale enterprise computing systems are growing rapidly, to address the increasing demand for data processing; however, in many cases, the computing resources in a single data center may not be sufficient for critical data-centric workloads, and important factors, such as space limitations, power availability, or company policies, limit the possibilities of expanding the data center's resources. In this paper, we explore the potential of harvesting spare computing resources across geo-distributed data centers with fast fabric interconnection for real-world enterprise applications. We specifically characterize the computing resource utilization of four large-scale production data centers, and we show how to efficiently combine local storage and computing clusters with remote and elastic computation resources. The primary challenge is incorporating the available remote computing resources efficiently. To achieve this goal, we propose leveraging the capabilities of Kubernetes-based elastic computing clusters to utilize the spare computing resources across geo-distributed data centers for Big Data applications. We also provide an experimental performance evaluation based on real-use case scenarios via an empirical execution and a simulation, which shows that the proposed system can accelerate Big Data services by employing existing computing resources more efficiently across geo-distributed data centers.
more »
« less
- NSF-PAR ID:
- 10158249
- Date Published:
- Journal Name:
- 2019 IEEE 21st International Conference on High Performance Computing and Communications; IEEE 17th International Conference on Smart City; IEEE 5th International Conference on Data Science and Systems (HPCC/SmartCity/DSS)
- Page Range / eLocation ID:
- 937 to 944
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Edge computing has enabled a large set of emerging edge applications by exploiting data proximity and offloading computation-intensive workloads to nearby edge servers. However, supporting edge application users at scale poses challenges due to limited point-of-presence edge sites and constrained elasticity. In this paper, we introduce a densely-distributed edge resource model that leverages capacity-constrained volunteer edge nodes to support elastic computation offloading. Our model also enables the use of geo-distributed edge nodes to further support elasticity. Collectively, these features raise the issue of edge selection. We present a distributed edge selection approach that relies on client-centric views of available edge nodes to optimize average end-to-end latency, with considerations of system heterogeneity, resource contention and node churn. Elasticity is achieved by fine-grained performance probing, dynamic load balancing, and proactive multi-edge node connections per client. Evaluations are conducted in both real-world volunteer environments and emulated platforms to show how a common edge application, namely AR-based cognitive assistance, can benefit from our approach and deliver low-latency responses to distributed users at scale.more » « less
-
This paper studies how to provision edge computing and network resources for complex microservice-based applications (MSAs) in face of uncertain and dynamic geo-distributed demands. The complex inter-dependencies between distributed microservice components make load balancing for MSAs extremely challenging, and the dynamic geo-distributed demands exacerbate load imbalance and consequently congestion and performance loss. In this paper, we develop an edge resource provisioning model that accurately captures the inter-dependencies between microservices and their impact on load balancing across both computation and communication resources. We also propose a robust formulation that employs explicit risk estimation and optimization to hedge against potential worst-case load fluctuations, with controlled robustness-resource trade-off. Utilizing a data-driven approach, we provide a solution that provides risk estimation with measurement data of past load geo-distributions. Simulations with real-world datasets have validated that our solution provides the important robustness crucially needed in MSAs, and performs superiorly compared to baselines that neglect either network or inter-dependency constraints.more » « less
-
Experience shows that on today's high performance systems the utilization of different acceleration cards in conjunction with a high utilization of all other parts of the system is difficult. Future architectures, like exascale clusters, are expected to aggravate this issue as the number of cores are expected to increase and memory hierarchies are expected to become deeper. One big aspect for distributed applications is to guarantee high utilization of all available resources, including local or remote acceleration cards on a cluster while fully using all the available CPU resources and the integration of the GPU work into the overall programming model. For the integration of CUDA code we extended HPX, a general purpose C++ run time system for parallel and distributed applications of any scale, and enabled asynchronous data transfers from and to the GPU device and the asynchronous invocation of CUDA kernels on this data. Both operations are well integrated into the general programming model of HPX which allows to seamlessly overlap any GPU operation with work on the main cores. Any user defined CUDA kernel can be launched on any (local or remote) GPU device available to the distributed application. We present asynchronous implementations for the data transfers and kernel launches for CUDA code as part of a HPX asynchronous execution graph. Using this approach we can combine all remotely and locally available acceleration cards on a cluster to utilize its full performance capabilities. Overhead measurements show, that the integration of the asynchronous operations (data transfer + launches of the kernels) as part of the HPX execution graph imposes no additional computational overhead and significantly eases orchestrating coordinated and concurrent work on the main cores and the used GPU devices.more » « less
-
We consider a large-scale service system where incoming tasks have to be instantaneously dispatched to one out of many parallel server pools. The user-perceived performance degrades with the number of concurrent tasks and the dispatcher aims at maximizing the overall quality of service by balancing the load through a simple threshold policy. We demonstrate that such a policy is optimal on the fluid and diffusion scales, while only involving a small communication overhead, which is crucial for large-scale deployments. In order to set the threshold optimally, it is important, however, to learn the load of the system, which may be unknown. For that purpose, we design a control rule for tuning the threshold in an online manner. We derive conditions that guarantee that this adaptive threshold settles at the optimal value, along with estimates for the time until this happens. In addition, we provide numerical experiments that support the theoretical results and further indicate that our policy copes effectively with time-varying demand patterns. Summary of Contribution: Data centers and cloud computing platforms are the digital factories of the world, and managing resources and workloads in these systems involves operations research challenges of an unprecedented scale. Due to the massive size, complex dynamics, and wide range of time scales, the design and implementation of optimal resource-allocation strategies is prohibitively demanding from a computation and communication perspective. These resource-allocation strategies are essential for certain interactive applications, for which the available computing resources need to be distributed optimally among users in order to provide the best overall experienced performance. This is the subject of the present article, which considers the problem of distributing tasks among the various server pools of a large-scale service system, with the objective of optimizing the overall quality of service provided to users. A solution to this load-balancing problem cannot rely on maintaining complete state information at the gateway of the system, since this is computationally unfeasible, due to the magnitude and complexity of modern data centers and cloud computing platforms. Therefore, we examine a computationally light load-balancing algorithm that is yet asymptotically optimal in a regime where the size of the system approaches infinity. The analysis is based on a Markovian stochastic model, which is studied through fluid and diffusion limits in the aforementioned large-scale regime. The article analyzes the load-balancing algorithm theoretically and provides numerical experiments that support and extend the theoretical results.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/HPCC/SmartCity/DSS.2019.00135
