Adaptive Cache Research Articles

Memory intensive AND/OR search for combinatorial optimization in graphical models

In this paper we explore the impact of caching during search in the context of the recent framework of AND/OR search in graphical models. Specifically, we extend the depth-first AND/OR Branch-and-Bound tree search algorithm to explore an AND/OR search graph by equipping it with an adaptive caching scheme similar to good and no-good recording. Furthermore, we present best-first search algorithms for traversing the same underlying AND/OR search graph and compare both algorithms empirically. We focus on two common optimization problems in graphical models: finding the Most Probable Explanation (MPE) in belief networks and solving Weighted CSPs (WCSP). In an extensive empirical evaluation we demonstrate conclusively the superiority of the memory intensive AND/OR search algorithms on a variety of benchmarks.

Process-Variation-Aware Adaptive Cache Architecture and Management

Fabricating circuits that employ ever-smaller transistors leads to dramatic variations in critical process parameters. This in turn results in large variations in execution/access latencies of different hardware components. This situation is even more severe for memory components due to minimum-sized transistors used in their design. Current design methodologies that are tuned for the worst case scenarios are becoming increasingly pessimistic from the performance angle, and thus, may not be a viable option at all for future designs. This paper makes two contributions targeting on-chip data caches. First, it presents an adaptive cache management policy based on nonuniform cache access. Second, it proposes a latency compensation approach that employs several circuit-level techniques to change the access latency of select cache lines based on the criticalities of the load instructions that access them. Our experiments reveal that both these techniques can recover significant amount of the lost performance due to worst case designs.

Persistence and Communication State Transfer in an Asynchronous Pipe Mechanism

Wide-area distributed systems offer new opportunities for executing large-scale scientific applications. On these systems, communication mechanisms have to deal with dynamic resource availability and the potential for resource and network failures. Connectivity losses can affect the execution of workflow applications, which require reliable data transport between components. We present the design and implementation of p-channels, an asynchronous and fault-tolerant pipe mechanism suitable for coupling workflow components. Fault-tolerant communication is made possible by persistence, through adaptive caching of pipe segments while providing direct data streaming. We present the distributed algorithm for implementing: (a) caching of pipe data segments; (b) asynchronous read operation; and (c) communication state transfer to handle dynamic process joins and leaves.

High Performance Approach for Server Side SOAP Processing

Based on XML, Web services inherit not only the advantages of XML, but its relatively poor performance, which makes it a poor choice for many high-performance applications. In this paper, we propose a new approach to improve Web services performance. Focusing on avoiding traditional XML parsing and Java reflection at runtime, this paper develops a service-specific SOAP Processor to accelerate execution. Moreover, SOAP Processor embeds several cache implementations and uses a novel adaptive cache mechanism, which can choose an optimized cache implementation dynamically in the runtime. Through our experiments in this paper, we observed that our approach can achieve a huge performance gain by incorporating the SOAP Processor into the SOAP engine.

A Caching Scheme for Session Setup in IMS Network

In IP Multimedia Subsystem (IMS), the session setup delay is a critical value for Quality of Service (QoS). The existing approaches to improve this metric depend on optimization of Session Initiation Protocol (SIP) message transmitting and signaling flows. Unfortunately, some service features are missing considered al-though they have been used widely in traditional 2G networks. This paper proposes a novel session setup scheme based on caching, upon the investigation of the performance of IMS session establishment. This mechanism involves cache based local routing policy and an adaptive caching algorithm, which can decrease call setup delay effectively as cached information in the terminating Serving-Call Session Control Function (S-CSCF) hit. The analytical model is deduced, as well as the delay and cost ratio functions are presented based on the model. Moreover, the analytical model is validated through the performance simulation in which the performance of the proposed novel method is evaluated against the basic session setup mechanism in terms of cost and delay.

COACS: A Cooperative and Adaptive Caching System for MANETs

This paper introduces a cooperation-based database caching system for Mobile Ad Hoc Networks (MANETs). The heart of the system is the nodes that cache submitted queries. The queries are used as indices to data cached in nodes that previously requested them. We discuss how the system is formed and how requested data is found if cached, or retrieved from the external database and then cached. Analysis is performed and expressions are derived for the different parameters, including upper and lower bounds for the number of query caching nodes as well as the average load they experience, generated network traffic, node bandwidth consumption, and other performance-related measures. Simulations with the ns-2 software were used to study the performance of the system in terms of average delay and hit ratio, and to compare it with the performance of two other caching schemes for MANETs, namely CachePath and CacheData. The results demonstrate the effectiveness of the proposed system in terms of achieved hit ratio and low delay.

An Adaptive Caching Strategy for m-Learning Based on SCORM Sequencing and Navigation

One of the main problems encountered in the usage of mobile devices as a learning platform is the presence of an impermanent network environment due to insufficient coverage or link failure in wireless communication. On the other hand, a persistent connection is usually offered by cellular phones using a telecommunication protocol, but with a relatively weak computing power and very limited network bandwidth which makes m-learning a time-consuming process. Moreover, learning contents are currently composed of various multimedia resources that induce long latency to display on handheld devices such as smartphones with GPRS. Recently, a lot of m-learning systems and contents have conformed to the Sharable Content Object Reference Model (SCORM) since it was introduced by ADL in the late 90s. The Sequencing and Navigation (S&N) specification is an important part of SCORM. S&N is defined to prescribe the intended student learning sequence by instructors. In this paper, we propose an adaptive course caching strategy based on the S&N specification in an m-learning environment. The system automatically switches to the corresponding course caching strategies, namely, the virtual memory management (VMM) mode and caching on disk (COD) mode, according to the current networking capability. The proposed mechanism is implemented on an m-learning system--Pocket SCORM--which received the 2005 Brandon Hall Excellence in Learning Award in the USA. Our simulation and experiments demonstrate that the proposed course caching strategy ultimately reduces the latency during the learning process and decreases the requests for Internet reconnection.

High Speed 3D Tomography on CPU, GPU, and FPGA

AbstractBack-projection (BP) is a costly computational step in tomography image reconstruction such as positron emission tomography (PET). To reduce the computation time, this paper presents a pipelined, prefetch, and parallelized architecture for PET BP (3PA-PET). The key feature of this architecture is its original memory access strategy, masking the high latency of the external memory. Indeed, the pattern of the memory references to the data acquired hinders the processing unit. The memory access bottleneck is overcome by an efficient use of the intrinsic temporal and spatial locality of the BP algorithm. A loop reordering allows an efficient use of general purpose processor's caches, for software implementation, as well as the 3D predictive and adaptive cache (3D-AP cache), when considering hardware implementations. Parallel hardware pipelines are also efficient thanks to a hierarchical 3D-AP cache: each pipeline performs a memory reference in about one clock cycle to reach a computational throughput close to 100%. The 3PA-PET architecture is prototyped on a system on programmable chip (SoPC) to validate the system and to measure its expected performances. Time performances are compared with a desktop PC, a workstation, and a graphic processor unit (GPU).

RACE: A Robust Adaptive Caching Strategy for Buffer Cache

Although many block replacement algorithms for buffer caches have been proposed to address the well-known drawbacks of the LRU algorithm, they are not robust and cannot maintain a consistent performance improvement over all workloads. This paper proposes a novel and simple replacement scheme, called the Robust Adaptive buffer Cache management schemE (RACE), which differentiates the locality of I/O streams by actively detecting access patterns that are inherently exhibited in two correlated spaces, that is, the discrete block space of program contexts from which I/O requests are issued and the continuous block space within files to which I/O requests are addressed. This scheme combines the global I/O regularities of an application and the local I/O regularities of individual files that are accessed in that application to accurately estimate the locality strength, which is crucial in deciding which blocks are to be replaced upon a cache miss. Through comprehensive simulations on 10 real-application traces, RACE is shown to have higher hit ratios than LRU and all other state-of-the-art cache management schemes studied in this paper.

Semantic-Aware and QoS-Aware Image Caching in Ad Hoc Networks

Semantic caching was originally used for structural data objects such as 2D location data and cannot be directly applied to mobile image data access. First, traditional semantic caching relies on exact match and therefore is not suitable for approximate and similarity-based queries. Second, the semantic description of cached data is defined on query context instead of data content, which leads to inefficient use of cache storage. Third, the semantic description of cached data does not reflect the popularity of the data, making it difficult to conduct popularity-driven content analysis and prediction. To facilitate content-based image retrieval in wireless ad hoc networks, we propose a semantic-aware image caching (SAIC) scheme in this paper. The proposed scheme can efficiently utilize the cache space and significantly reduce the cost of image retrieval. The proposed SAIC scheme is based on several innovative ideas: 1) multilevel representation of the semantic contents, 2) association-based and Bayesian probability-based content prediction, 3) constraint-based representation method showing the semantic similarity between images, 4) nonflooding query processing, and 5) adaptive cache consistency maintenance. The proposed model is introduced, and through extensive simulation, its behavior has been compared against two state-of-the-art ad hoc caching schemes as advanced in the literature.

Improving Web Server Performance with Adaptive Proxy Caching in Soft Real-time Mobile Applications

This paper introduces an adaptive cache proxy to improve the performance of web access in soft real-time applications. It consists of client proxies and cooperative proxy servers with a server-side pushing schema. The large amount of heterogeneous data will be stored in the proxy servers and delivered to clients through computer networks to reduce the response time and network traffic. The adaptive proxy pre-fetches and replaces heterogeneous data dynamically in consideration of networks cost, data size, data change rate, etc. The simulation results show that the modified LUV algorithm has better performance in terms of hit rate, byte hit rate, and delay saving rate. With the cooperative proxy caching, it is shown that the performance of the proxy caching system is more predictable even if the proxies need to deal with a variety of data. The modified adaptive TTL algorithm has better performance in terms of the combination of temporal coherency and system overheads.

Adaptive Stack Cache with Fast Address Generation

Adaptive Stack Cache with Fast Address Generation

Caching video objects: layers vs versions?

Because Internet access rates are highly heterogeneous, many video content providers today make available different versions of the videos, with each version encoded at a different rate. Multiple video versions, however, require more server storage and may also dramatically impact cache performance in a traditional cache or in a CDN server. An alternative to versions is layered encoding, which can also provide multiple quality levels. Layered encoding requires less server storage capacity and may be more suitable for caching; but it typically increases transmission bandwidth due to encoding overhead. In this paper we compare video streaming of multiple versions with that of multiple layers in a caching environment. We examine caching and distribution strategies that use both versions and layers. We consider two cases: the request distribution for the videos is known a priori; and adaptive caching, for which the request distribution is unknown. Our analytical and simulation results indicate that mixed distribution/caching strategies provide the best overall performance.

A Coherence-Replacement Protocol For Web Proxy Cache Systems

As World Wide Web usage has grown dramatically in recent years, so has the recognition that web caches (especially proxy caches) will have an important role in reducing server loads, client request latencies, and network traffic. In this paper, we propose an adaptive cache coherence-replacement scheme for web proxy cache systems that is based on several criteria about the system and applications, with the objective of optimizing the distributed cache system performance. Our coherence-replacement scheme assigns a replacement priority value to each cache block according to a set of criteria for deciding which block to remove. The goal is to provide an effective utilization of the distributed cache memory and a good application performance.

Microarchitecture-level leakage reduction with data retention

In this paper, we study microarchitecture-level leakage energy reduction by power gating. We consider the virtual power/ground rails clamp (VRC) and multithreshold CMOS (MTCMOS) techniques and apply VRC to memory-based units for data retention and MTCMOS to the other units. We propose a systematic methodology for leakage reduction at the microarchitecture level, in which profiling of idle period distribution and ideal power gating analysis are used to select a target component for realistic power gating. We show that the ideal leakage energy reduction can be up to 30% of the total energy for the modern high-performance very long instruction word processors we study and that the secondary level (L2) cache contributes most to the reduction. We further improve the existing adaptive cache decay method for leakage reduction by using VRC for data retention and name it VRC decay . Applied to L2 cache, the VRC decay, on average, increases performance by 5.6% and reduces system energy by 24.1%, compared to the adaptive cache decay without data retention.

SMA: A Self-Monitored Adaptive Cache Warm-Up Scheme for Microprocessor Simulation

This paper presents an adaptive technique for warming up caches in sampled microprocessor simulation. The simulator monitors the warm-up process of the caches and decides when the caches are warmed up based on simple heuristics. This mechanism allows the warm up length to be adaptive to cache sizes and benchmark variability characteristics. With only half or one-third of the average warm-up length of previous methods, the proposed Self Monitored Adaptive (SMA) warm-up technique achieves CPI results very similar to previous methods. On average SMA exhibits only 0.2% warmup error in CPI. For simulating small caches, the SMA technique can reduce the warm-up overhead by an order of magnitude compared to previous techniques. Finally, SMA gives the user some indicator of warm-up error at the end of the cycle-accurate simulation that helps the user to gauge the accuracy of the warm-up.

ACASH: an adaptive web caching method based on the heterogeneity of web object and reference characteristics

The use of cache to store and process web objects has increased, and studies on effective web caching have been actively conducted. A processing unit of web caching contains web objects, and changes in the heterogeneity of these web objects and their reference characteristics become a major factor of decreased performance in the existing method. This study proposes ACASH, a new web caching method. ACASH manages a cache scope by dividing it based on the heterogeneity of web objects. Also, it adaptively reflects changes in the reference characteristics of the object according to the time elapsed. In addition, it is shown in an experimental model, which captures the heterogeneity of the object that the proposed method performs better than the existing method.

Adaptive schemes for distributed web caching

In distributed web caching architectures, institutional proxies take advantage of their neighbors’ contents in order to reduce the number of requests forwarded to the server. Intuitively, the maximum benefit from this cooperation is expected when the proxies that exhibit similar requests are grouped together. The current practice is to follow a static and manual configuration of neighbors. Such an approach has a number of drawbacks: (i) static allocation may not determine the best neighbors, especially if global knowledge of the participating proxies is not available, (ii) a manual allocation places significant administrative burden, (iii) static schemes are insensitive to changes in access patterns, and (iv) they cannot deal with the introduction of new, potentially useful, proxies. In this paper, we propose a set of algorithms that allow proxies to independently explore the network for better neighbors and continuously update their configuration in an adaptive fashion. The simulation experiments illustrate that dynamic neighbor reconfiguration leads to significantly higher hit ratios compared to the static approach. Although some researchers in the past have recognized the need for adaptive caching, to the best of our knowledge this is the first study to propose concrete algorithms and evaluate their efficacy.

E-MACSC: A novel dynamic cache tuning technique to reduce information retrieval roundtrip time over the Internet

The novel technique proposed in this paper for dynamic cache size tuning is an enhancement of the previous MACSC ( Model for Adaptive Cache Size Control) approach. Similar to its MACSC predecessor the Enhanced MACSC (E-MACSC) technique consistently maintains the given cache hit ratio. The focus of the research is presently on supporting small caching systems of limited recyclable memory resources. The MACSC tunes the cache size adaptively with the instantaneous popularity ratio, which is computed statistically by the point-estimate (PE) method on the fly. It is difficult to harness the PE convergence time, because the following are unpredictable: (a) the number of data samples needed by the PE process to achieve convergence and (b) the inter-arrival times among these data samples. In the E-MACSC framework this unpredictability problem is resolved by replacing PE with the M 3RT mechanism, which is a realization of the Convergence Algorithm (CA). Therefore the E-MACSC is also called the MACSC(M 3RT) as compared to the original PE-based MACSC or MACSC(PE). The CA is an IEPM ( Internet End-to-End Performance Measurement) technique that measures the mean of a waveform quickly and accurately. The CA prediction accuracy, however, differs from other IEPM techniques, because it is independent of the type of waveform/distribution. This independence arises from the fact that CA is based on the Central Limit Theorem. The E-MACSC approach provides several benefits as follows: (a) it maintains the prescribed hit ratio efficaciously, (b) it lessens cache size oscillation, and (c) it uses a fixed number of data samples and this makes its computation time more predictable. The E-MACSC is unique because of the following reasons: (a) it utilizes the relative popularity of the data objects as the sole control parameter and (b) it tunes the cache size adaptively by direct data measurement with the CA support. The relative popularity profile of data objects is called popularity distribution (PD) in the E-MACSC context. Any change in the PD's standard deviation indicates a shift of user preference for particular data objects. Monitoring and leveraging this change is the basis for E-MACSC to find a meaningful popularity ratio for deciding how the cache size should be tuned in a dynamic manner.

Design and performance studies of an adaptive cache retrieval scheme in a mobile computing environment

In a mobile computing system, as users move to a new service area, the new server is usually considered to take over the execution of running programs for mobile users from the previous server so as to reduce the communication overhead of a mobile system. This procedure is referred to as service handoff. Note that when service handoff occurs, the new server will lose its advantage for cache access. To remedy this, we explore in this paper several cache retrieval schemes to improve the efficiency of cache retrieval. In particular, we analyze the impact of using a coordinator buffer to improve the overall performance of cache retrieval. Moreover, in light of the properties of transactions (i.e., temporal locality of data access among transactions), we devise a Dynamic and Adaptive cache Retrieval scheme (DAR) that can utilize proper cache methods according to some specific criteria to deal with the service handoff situation in a mobile computing environment. Performance of these cache retrieval schemes is analyzed and a system simulator is developed to validate our results. We devise a systematic procedure for determining the optimal operating points of DAR. Our experimental results show that by adaptively adopting the advantages of different cache retrieval methods, DAR significantly outperforms other schemes and is particularly effective for a mobile computing environment.