Automatic Caching Research Articles

Cache Programming for Scientific Loops Using Leases

Cache management is important in exploiting locality and reducing data movement. This article studies a new type of programmable cache called the lease cache. By assigning leases, software exerts the primary control on when and how long data stays in the cache. Previous work has shown an optimal solution for an ideal lease cache. This article develops and evaluates a set of practical solutions for a physical lease cache emulated in FPGA with the full suite of PolyBench benchmarks. Compared to automatic caching, lease programming can further reduce data movement by 10% to over 60% when the data size is 16 times to 3,000 times the cache size, and the techniques in this article realize over 80% of this potential. Moreover, lease programming can reduce data movement by another 0.8% to 20% after polyhedral locality optimization.

CARL: Compiler Assigned Reference Leasing

Data movement is a common performance bottleneck, and its chief remedy is caching. Traditional cache management is transparent to the workload: data that should be kept in cache are determined by the recency information only, while the program information, i.e., future data reuses, is not communicated to the cache. This has changed in a new cache design named Lease Cache . The program control is passed to the lease cache by a compiler technique called Compiler Assigned Reference Lease (CARL). This technique collects the reuse interval distribution for each reference and uses it to compute and assign the lease value to each reference. In this article, we prove that CARL is optimal under certain statistical assumptions. Based on this optimality, we prove miss curve convexity, which is useful for optimizing shared cache, and sub-partitioning monotonicity, which simplifies lease compilation. We evaluate the potential using scientific kernels from PolyBench and show that compiler insertions of up to 34 leases in program code achieve similar or better cache utilization (in variable size cache) than the optimal fixed-size caching policy, which has been unattainable with automatic caching but now within the potential of cache programming for all tested programs and most cache sizes.

Heterogeneous Information Network-Based Content Caching in the Internet of Vehicles

With the rapid development of the Internet of Vehicles (IoV), the data in the network have become more complicated, and user demand for popular content has been growing. The focus of this paper is how to address the ever-changing mobile network environment of the IoV with the popular content caching strategy. To achieve the automatic selection, storage and delivery of popular content in the IoV, this paper modeled the network content as a heterogeneous information network (HIN). In this way, we can greatly reduce the load of the limited network with a small computational cost and give the user in the car a better experience with the automatic caching mode, in which the popular content can be cached in real time. For high-risk driving behavior, this popular content caching method is more secure and effective, which can reduce the network load, improve user satisfaction and bring a higher quality of experience (QoE) and a better quality of service (QoS) to users, as confirmed by the experimental results.

Abstract 2478: Cancer genome interpretation with CIViCpy

Abstract The promise of precision medicine–in which a cancer patient’s treatment is tailored to their specific cancer variants–requires concise, standardized, and searchable clinical interpretations. We recently developed the Clinical Interpretations for Variants in Cancer (CIViC; civicdb.org) crowd-sourced, open-access knowledgebase, providing a structured framework for evaluating genomic variants of various types (e.g., fusions, SNVs) for their therapeutic, prognostic, predisposing or diagnostic utility. CIViC is fully open-source, and has a documented API for accessing records of interpretation assertions, evidence, variants, and genes. However, this API was built around the framework for which the CIViC web tool was designed, with a focus on search and display of the interpretation of individual genomic variants in cancer. Third-party tools to analyze or access the contents of this database programmatically must therefore construct methods to leverage this API, often reimplementing similar functionality in the pursuit of common analysis tasks that are not within the scope of the CIViC web interface. To address this shortfall, we have developed CIViCpy (civicpy.org), an analysis toolkit and Software Developer Kit (SDK) for extracting and exploring the content of the CIViC knowledgebase. CIViCpy lets you pull data from CIViC using the civic module and interact with records as dynamic objects in Python. With the aid of automatic caching, CIViCpy provides fully-documented data objects for CIViC records (e.g. assertions, genes, variants) without worrying about querying CIViC more than once for the same data. Thus, repeated requests for a given CIViC record in the same session will load quickly, enabling rapid identification of records matching desired characteristics. All CIViCpy objects are directly comparable for ease of data analysis. CIViCpy also performs sensible pre-caching, e.g. requests for variants pre-load related gene and evidence records. Like CIViC, CIViCpy is fully documented, open-source, and permissively (MIT) licensed. CIViCpy is also easily and freely installable from the python package index (PyPI). In this study, we illustrate how the CIViCpy SDK is used to enable the analysis and export of CIViC entries. We highlight the use of the CIViCpy SDK to extract variants from CIViC and export them into the VCFv4.2 format through the built-in export tools, enabling their use in annotation pipelines or software that expects this common variant format. We also assess the viability of CIViCpy as a tool for advancing individualized patient care by examining how the CIViCpy interface can be used to systematically generate reports for the interpretation of observed variants in patient cancers. We evaluate this through the analysis of the 48,447 sequenced tumors of the AACR Project GENIE dataset, revealing clinical actionability at the patient level and addressing ongoing challenges in the programmatic analysis of genomic variants of cancers. Citation Format: Alex H. Wagner, Malachi Griffith, Obi L. Griffith. Cancer genome interpretation with CIViCpy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2478.

Automatic cache partitioning method for high-level synthesis

Existing algorithms can be automatically translated from software to hardware using High-Level Synthesis (HLS), allowing for quick prototyping or deployment of embedded designs. High-level software is written with a single main memory in mind, whereas hardware designs can take advantage of many parallel memories. The translation and optimization of memory usage, and the generation of resulting architectures, is important for high-performance designs. Tools provide optimizations on memory structures targeting data reuse and partitioning, but generally these are applied separately for a given object in memory. Memory access that cannot be effectively optimized is serialized to the memory, hindering any further parallelization of the surrounding generated hardware.In this work, we present an automated optimization method for creating custom cache memory architectures for HLS generated designs. Our optimization uses runtime profiling data, and is performed at a localized scope. This method combines data reuse savings and memory partitioning to further increase the potential parallelism and alleviate the serialized memory access, increasing performance. Comparisons are made against architectures without this optimization, and against other HLS caching approaches. Results are presented showing this method requires 72% of the number of execution cycles compared to a single-cache design, and 31% compared to designs with no caches.

Accelerating the image processing by the optimization strategy for deep learning algorithm DBN

In recent years, image processing especially for remote sensing technology has developed rapidly. In the field of remote sensing, the efficiency of processing remote sensing images has been a research hotspot in this field. However, the remote sensing data has some problems when processing by a distributed framework, such as Spark, and the key problems to improve execution efficiency are data skew and data reused. Therefore, in this paper, a parallel acceleration strategy based on a typical deep learning algorithm, deep belief network (DBN), is proposed to improve the execution efficiency of the DBN algorithm in Spark. First, the re-partition algorithm based on the tag set is proposed to the relief data skew problem. Second, the cache replacement algorithm on the basis of characteristics is proposed to automatic cache the frequently used resilient distributed dataset (RDD). By caching RDD, the re-computation time of frequently reused RDD is reduced, which lead to the decrease of total computation time of the job. The numerical and analysis verify the effectiveness of the strategy.

VPH-HF: A software framework for the execution of complex subject-specific physiology modelling workflows

Computational medicine more and more requires complex orchestrations of multiple modelling & simulation codes, written in different programming languages and with different computational requirements, which when validated need to be run many times on large cohorts of patients. The aim of this paper is to present a new open source software, the VPH Hypermodelling Framework (VPH-HF). The VPH-HF overcomes the limitations of most workflow execution environments by supporting both Taverna and Muscle2; the addition of Muscle2 support makes possible the execution of very complex orchestrations that include strongly-coupled models. The overhead that the VPH-HF imposes in exchange for this is small, and tends to be flat regardless of the complexity and the computational cost of the hypermodel being executed. We recommend the use of the VPH-HF to orchestrate any hypermodel with an execution time of 200 s or higher, which would confine the VPH-HF overhead to less than 10%. The VPH-HF also provide an automatic caching system over the execution of every hypomodel, which may provide considerable speed-up when the orchestration is run repeatedly over large numbers of patients or within stochastic frameworks, and the input sets are properly binned. The caching system also makes it easy to form large input set/output set databases required to develop reduced-order models, and the framework offers the possibility to dynamically replace single models in the orchestration with reduced-order versions built from cached results, an essential feature when the orchestration of multiple models produces a combinatory explosion of the computational cost.

ATR4S: toolkit with state-of-the-art automatic terms recognition methods in Scala

Automatically recognized terminology is widely used for various domain-specific texts processing tasks, such as machine translation, information retrieval or ontology construction. However, there is still no agreement on which methods are best suited for particular settings and, moreover, there is no reliable comparison of already developed methods. We believe that one of the main reasons is the lack of state-of-the-art method implementations, which are usually non-trivial to recreate—mostly, in terms of software engineering efforts. In order to address these issues, we present ATR4S, an open-source software written in Scala that comprises 13 state-of-the-art methods for automatic terminology recognition (ATR) and implements the whole pipeline from text document preprocessing, to term candidates collection, term candidate scoring, and finally, term candidate ranking. It is highly scalable, modular and configurable tool with support of automatic caching. We also compare 13 state-of-the-art methods on 7 open datasets by average precision and processing time. Experimental comparison reveals that no single method demonstrates best average precision for all datasets and that other available tools for ATR do not contain the best methods.

Mollib: a molecular and NMR data analysis software.

Mollib is a software framework for the analysis of molecular structures, properties and data with an emphasis on data collected by NMR. It uses an open source model and a plugin framework to promote community-driven development of new and enhanced features. Mollib includes tools for the automatic retrieval and caching of protein databank (PDB) structures, the hydrogenation of biomolecules, the analysis of backbone dihedral angles and hydrogen bonds, and the fitting of residual dipolar coupling (RDC) and residual anisotropic chemical shift (RACS) data. In this article, we release version 1.0 of mollib and demonstrate its application to common molecular and NMR data analyses.

Distributed proxy cache technology based on autonomic computing in smart cities

With the rapid growth of Internet services in smart cities, large network flow has caused network congestion among users, thereby deteriorating network service quality and user experience. In this study, a distributed proxy cache management platform based on autonomous decision is proposed. The platform realizes highly efficient automatic cache management by taking the advantages of autonomous perception and decision. First, the platform establishes the autonomous management framework of proxy cache on the basis of the idea of feedback loop. Next, the URL is distributed evenly among cache nodes through the hash ring storage mechanism with virtual nodes. Cache status adjustment strategies are made based on historical status sequences and the gray prediction model. A method that is used to predict content hot-rank based on domain name set is proposed to reduce loads of cache nodes. Second, the platform designs a distributed proxy cache management system by depending on the autonomous management framework of proxy cache and the cache management mechanism of autonomous decision, and the Cache Hot Spots Migrate algorithm is suggested for the dynamic migration and integration of virtual nodes on the hash ring. A system knowledge base is established through status sequence and corresponding adjustment strategies. Finally, experimental results on the hash ring, content hot-rank prediction, and adjustment of autonomous decision of cache status are analyzed, thereby confirming that the distributed proxy cache management mechanism based on autonomous decision could manage cache nodes effectively and automatically, and could improve the overall performance of the system.

Data Lifetime Enhancement for Improving QoS in NDN

Because of the further development of the Internet, the high-speed and image-related services now have occupied most of the network resources. For this reason, the Next Generation Network (NGN) will aim at optimizing ISP networks. Funded by the U.S. National Science Foundation (NSF), the Named Data Networking (NDN) project purposes to attain automatic caching of data and create a name-based routing mechanism for promoting reaction time and decreasing throughput. To present an applicable QoS mechanism for the NDN architecture, this paper proposes to classify different types of data and imports the idea of data lifetime, which means that an IP packet that has been in the network too long should be discarded, to take the place of traditional memory update methods, like the Fist In, First Out (FIFO) algorithm and the Least Recently Used (LRU) algorithm. In our proposed scheme, data must be classified at the source and different lifetime is defined for different data types in order to decrease the load of the Content Store (CS). Without additional service latency, our scheme reduces not only the cost of memory construction but also the service latency for NDN under low throughput situations. In addition, this scheme can regulate the data lifetime dynamically and optimize the length of the lifetime based on network parameters. The simulation results show that our proposed scheme decreases approximately 15% of service latency under low throughput situations.

The HaLoop approach to large-scale iterative data analysis

The growing demand for large-scale data mining and data analysis applications has led both industry and academia to design new types of highly scalable data-intensive computing platforms. MapReduce has enjoyed particular success. However, MapReduce lacks built-in support for iterative programs, which arise naturally in many applications including data mining, web ranking, graph analysis, and model fitting. This paper (This is an extended version of the VLDB 2010 paper “HaLoop: Efficient Iterative Data Processing on Large Clusters” PVLDB 3(1):285–296, 2010.) presents HaLoop, a modified version of the Hadoop MapReduce framework, that is designed to serve these applications. HaLoop allows iterative applications to be assembled from existing Hadoop programs without modification, and significantly improves their efficiency by providing inter-iteration caching mechanisms and a loop-aware scheduler to exploit these caches. HaLoop retains the fault-tolerance properties of MapReduce through automatic cache recovery and task re-execution. We evaluated HaLoop on a variety of real applications and real datasets. Compared with Hadoop, on average, HaLoop improved runtimes by a factor of 1.85 and shuffled only 4 % as much data between mappers and reducers in the applications that we tested.

Performance-Enhanced Caching Scheme for Web Clusters for Dynamic Content

In order to improve the QoS of applications, clusters of web servers are increasingly used in web services. Caching helps improve performance in web servers, but is largely exploited only for static web content. With more web applications using backend databases today, caching of dynamic content has a crucial role in web performance. This paper presents a set of cache management schemes for handling dynamic data in web clusters by sharing cached contents. These schemes use either automatic or expiry-based cache validation, and work with any type of request distribution. The techniques improve response by utilizing the caches efficiently and reducing redundant database accesses by web servers while ensuring cache consistency. The authors present caching schemes for both horizontal and vertical cluster architectures. Simulations show an appreciable performance rise in response times of queries in clustered web servers.

A novel dynamic cache size adjustment approach for better data retrieval performance over the internet

A novel model for adaptive cache size control (MACSC) at runtime is proposed in this paper. It automatically maintains the prescribed hit ratio to guarantee the minimum expected caching performance. With this capability the system delivers better client/server communication and service response. The fringe benefit is better availability of the backbone bandwidth because caching reduces the traffic for long-haul data transfer over the Internet. The MACSC adjusts the cache size quickly, optimally and without the deleterious effect of remedying a past event spuriously regarded as a present one. The MACSC is a form of adaptive cache control within the area of cache adaptivity. It is unique and differs from other previous methods because it leverages the relative data object popularity as the only parameter in the control process, in the form of the popularity ratio. For the present MACSC research scope, the relative data object popularity profile, namely, the popularity distribution, is assumed to be bell-shaped and unimodel, but allows for a gentle mode skew. The MACSC computes the cache size adjustment from the sampled life data with the help of the point-estimate approach represented by the N-equation. The preliminary results from the verification and validation experiments unanimously confirm that the MACSC is indeed effective for automatic adaptive cache size control and economizing Internet backbone bandwidth.

Language-Based Caching of Dynamically Generated HTML

Increasingly, HTML documents are dynamically generated by interactive Web services. To ensure that the client is presented with the newest versions of such documents it is customary to disable client caching causing a seemingly inevitable performance penalty. In the system, dynamic HTML documents are composed of higher-order templates that are plugged together to construct complete documents. We show how to exploit this feature to provide an automatic fine-grained caching of document templates, based on the service source code. A service transmits not the full HTML document but instead a compact JavaScript recipe for a client-side construction of the document based on a static collection of fragments that can be cached by the browser in the usual manner. We compare our approach with related techniques and demonstrate on a number of realistic benchmarks that the size of the transmitted data and the latency may be reduced significantly.

Database cache requirements calculation using near-line data access traces in a 500+ user transaction system

A pragmatic way is shown to calculate database cache sizes for a given desired level of cache performance, without use of application information. Treating applications as black boxes allows focusing on higher order structures of database accesses to classify applications. Online tracing of all database accesses allows a near-line simulation of cache performance for various cache sizes. These simulations result in typical patterns for application groups which point to areas with strong gradients in the relation between cache size and cache efficiency which can be exploited. The near-line-ness of this calculation lends itself to automatic adaptive cache size regulation.