An official website of the United States government
We propose Clamor, a functional cluster computing framework that adds support for fine-grained, transparent access to global variables for distributed, data-parallel tasks. Clamor targets workloads that perform sparse accesses and updates within the bulk synchronous parallel execution model, a setting where the standard technique of broadcasting global variables is highly inefficient. Clamor implements a novel dynamic replication mechanism in order to enable efficient access to popular data regions on the fly, and tracks finegrained dependencies in order to retain the lineage-based fault tolerance model of systems like Spark. Clamor can integrate with existing Rust and C++ libraries to transparently distribute programs on the cluster. We show that Clamor is competitive with Spark in simple functional workloads and can improve performance significantly compared to custom systems on workloads that sparsely access large global variables: from 5x for sparse logistic regression to over 100x on distributed geospatial queries.
more »
« less
Meteor: Optimizing spark-on-yarn for short applications
Due to its speed and ease of use, Spark has become a popular tool amongst data scientists to analyze data in various sizes. Counter-intuitively, data processing workloads in industrial companies such as Google, Facebook, and Yahoo are dominated by short-running applications, which is due to the majority of applications being mostly consisted of simple SQL-like queries (Dean, 2004, Zaharia et al, 2008). Unfortunately, the current version of Spark is not optimized for such kinds of workloads. In this paper, we propose a novel framework, called Meteor, which can dramatically improve the performance for short-running applications. We extend Spark with three additional operating modes: one-thread, one-container, and distributed. The one-thread mode executes all tasks on just one thread; the one-container mode runs these tasks in one container by multi-threading; the distributed mode allocates all tasks over the whole cluster. A new framework for submitting applications is also designed, which utilizes a fine-grained Spark performance model to decide which of the three modes is the most efficient to invoke upon a new application submission. From our extensive experiments on Amazon EC2, one-thread mode is the optimal choice when the input size is small, otherwise the distributed mode is better. Overall, Meteor is up to 2 times faster than the original Spark for short applications.
more »
« less
- Award ID(s):
- 1836881
- PAR ID:
- 10154802
- Date Published:
- Journal Name:
- Future generation computer systems
- Volume:
- 101
- ISSN:
- 0167-739X
- Page Range / eLocation ID:
- 262-271
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Apache Spark is a popular cluster computing framework for iterative analytics workloads due to its use of Resilient Distributed Datasets (RDDs) to cache data for in-memory processing. We have revealed that the performance of Spark RDD cache can be severely limited if its capacity falls short to the needs of the workloads. In this paper, we have explored different memory hybridization strategies to leverage emergent Non-Volatile Memory (NVM) devices for Spark's RDD cache. We have found that a simple layered hybridization approach does not offer an effective solution. Therefore, we have designed a flat hybridization scheme to leverage NVM for caching RDD blocks, along with several architectural optimizations such as dynamic memory allocation for block unrolling, asynchronous migration with preemption, and opportunistic eviction to disk. We have performed an extensive set of experiments to evaluate the performance of our proposed flat hybridization strategy and found it to be robust in handling different system and NVM characteristics. Our proposed approach uses DRAM for a fraction of the hybrid memory system and yet manages to keep the increase in execution time to be within 10% on average. Moreover, our opportunistic eviction of blocks to disk improves performance by up to 7.5% when utilized alongside the current mechanism.more » « less
-
Apache Spark arguably is the most prominent Big Data processing framework tackling the scalability challenge of a wide variety of modern workloads. A key to its success is caching critical data in memory, thereby eliminating wasteful computations of regenerating intermediate results. While critical to performance, caching is not automated. Instead, developers have to manually handle such a data management task via APIs, a process that is error-prone and labor-intensive, yet may still yield sub-optimal performance due to execution complexities. Existing optimizations rely on expensive profiling steps and/or application-specific cost models to enable a postmortem analysis and a manual modification to existing applications. This paper presents CACHEIT, built to take the guesswork off the users while running applications as-is. CACHEIT analyzes the program’s workflow, extracting important features such as dependencies and access patterns, using them as an oracle to detect high-value data candidates and guide the caching decisions at run time. CACHEIT liberates users from low-level memory management requirements, allowing them to focus on the business logic instead. CACHEIT is application-agnostic and requires no profiling or a cost model. A thorough evaluation with a broad range of Spark applications on real-world datasets shows that CACHEIT is effective in maintaining satisfactory performance, incurring only marginal slowdown compared to the manually well-tuned counterpartsmore » « less
-
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
-
Modern distributed machine learning (ML) training workloads benefit significantly from leveraging GPUs. However, significant contention ensues when multiple such workloads are run atop a shared cluster of GPUs. A key question is how to fairly apportion GPUs across workloads. We find that established cluster scheduling disciplines are a poor fit because of ML workloads' unique attributes: ML jobs have long-running tasks that need to be gang-scheduled, and their performance is sensitive to tasks' relative placement. We propose Themis, a new scheduling framework for ML training workloads. It's GPU allocation policy enforces that ML workloads complete in a finish-time fair manner, a new notion we introduce. To capture placement sensitivity and ensure efficiency, Themis uses a two-level scheduling architecture where ML workloads bid on available resources that are offered in an auction run by a central arbiter. Our auction design allocates GPUs to winning bids by trading off fairness for efficiency in the short term, but ensuring finish-time fairness in the long term. Our evaluation on a production trace shows that Themis can improve fairness by more than 2.25X and is ~5% to 250% more cluster efficient in comparison to state-of-the-art schedulers.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/https://doi.org/10.1016/j.future.2019.05.077
