Efficient Data Placement Research Articles

A Learning-based Data Placement Framework for Low Latency in Data Center Networks

Abstract: The challenge of achieving low-latency data services has become increasingly crucial for data center applications. In modern distributed storage systems, efficient data placement plays a vital role in minimizing data movement delays, thereby significantly reducing service latency. While previous solutions for data placement have often relied on pre-assumed data request distributions or traced analysis, the dynamic nature of network conditions and evolving user request patterns poses a complex online decision-making challenge. Traditional static model-based approaches are limited in their ability to cope with such dynamic system dynamics. To address these challenges and consider both data movement and analytical latency, we introduce a novel framework named DataBot+. This framework leverages reinforcement learning techniques to autonomously learn optimal data placement policies. DataBot+ employs neural networks that are trained using a modified version of Q-learning. The input to these networks comprises real-time data flow measurements, while the output is a value function that estimates the imminent future latency. To ensure timely decision-making, DataBot+ is designed with a two-fold asynchronous architecture, consisting of production and training components. This separation guarantees that the training process does not introduce additional overhead to the handling of data flows. Our approach is validated through extensive evaluations using real-world traces, which convincingly demonstrate the efficacy of our proposed design.

Multi-Task Learning for Electricity Price Forecasting and Resource Management in Cloud Based Industrial IoT Systems

Cloud computing has gained immense popularity in the logistics industry. This innovative technology optimizes computing operations by eliminating the requirement for physical equipment for calculations. Instead, specialized companies provide cloud-based computing services, relying heavily on computers and servers that consume substantial amounts of energy. Hence, ensuring the availability of affordable and dependable electricity is paramount for the efficient design and management of these logistics services. Cloud centers, which are power-intensive, face the challenge of reducing their energy consumption due to escalating power costs. To address this issue, efficient data placement and node management strategies are commonly employed in logistics operations. An AlexNet model has been designed to optimize storage relocation and predict power prices. The outcome of this initiative has resulted in a considerable reduction in energy consumption at data centres. The model uses Dwarf Mongoose Optimization Algorithm (DMOA) to produce an optimal solution for the AlexNet and increase its performance with a real-world dataset from IESO in Ontario, Canada. 75% of the available data was used for training to assure the model’s precision, with the remaining 25% allocated to testing purposes. The model forecasts power prices with an MAE of 2.22% and an MSE of 6.33%, resulting in an average reduction of 22.21% in electricity expenses. Our proposed method has an accuracy of 97% compared to 11 benchmark algorithms, including CNN, DenseNet, and SVM having an accuracy of 89%, 88%, and 82%, respectively.

Enhancing Scalability through Extended Hybrid Switching with Mixture Path Deployment (EHS-MPD): A Novel Framework

In today's digital landscape, the exponential growth of data has created significant challenges for archiving systems. Existing solutions often struggle to cope with the ever-increasing data volumes, leading to scalability limitations, reduced efficiency, and inflexibility. To address these challenges, this research paper introduces a novel approach that extends hybrid switching with a mixture path deployment framework. By combining the benefits of hybrid switching techniques and the adaptability of the mixture path deployment framework, our proposed approach aims to enhance scalability, efficiency, and flexibility in archiving systems while minimizing the risks of data loss and system failures. This paper provides a comprehensive analysis of the design, implementation, and evaluation of our approach, highlighting its key features, benefits, and potential impact on archiving systems. We review existing archiving approaches, identifying their limitations and shortcomings. Then, we delve into the concept of hybrid switching, exploring its advantages in optimizing network performance. Additionally, we introduce the mixture path deployment framework, emphasizing its benefits in load balancing, fault tolerance, and performance improvement. The proposed approach's architecture is presented, showcasing the integration of hybrid switching techniques and the mixture path deployment framework. Strategies for efficient data placement and management are discussed, including dynamic data placement algorithms and metadata management techniques. The design and implementation of the extended archiving system are described, outlining the interactions among its components. Overall, our novel approach to archiving high scalability through extending hybrid switching with a mixture path deployment framework offers a promising solution to address the limitations of existing archiving systems. It enables improved performance, flexibility, and fault tolerance, paving the way for efficient and scalable archiving solutions in the era of big data.

Unified Holistic Memory Management Supporting Multiple Big Data Processing Frameworks over Hybrid Memories

To process real-world datasets, modern data-parallel systems often require extremely large amounts of memory, which are both costly and energy inefficient. Emerging non-volatile memory (NVM) technologies offer high capacity compared to DRAM and low energy compared to SSDs. Hence, NVMs have the potential to fundamentally change the dichotomy between DRAM and durable storage in Big Data processing. However, most Big Data applications are written in managed languages and executed on top of a managed runtime that already performs various dimensions of memory management. Supporting hybrid physical memories adds a new dimension, creating unique challenges in data replacement. This article proposes Panthera, a semantics-aware, fully automated memory management technique for Big Data processing over hybrid memories. Panthera analyzes user programs on a Big Data system to infer their coarse-grained access patterns, which are then passed to the Panthera runtime for efficient data placement and migration. For Big Data applications, the coarse-grained data division information is accurate enough to guide the GC for data layout, which hardly incurs overhead in data monitoring and moving. We implemented Panthera in OpenJDK and Apache Spark. Based on Big Data applications’ memory access pattern, we also implemented a new profiling-guided optimization strategy, which is transparent to applications. With this optimization, our extensive evaluation demonstrates that Panthera reduces energy by 32–53% at less than 1% time overhead on average. To show Panthera’s applicability, we extend it to QuickCached, a pure Java implementation of Memcached. Our evaluation results show that Panthera reduces energy by 28.7% at 5.2% time overhead on average.

Efficient XML data placement schemes over multiple mobile wireless broadcast channels

Efficient XML data placement schemes over multiple mobile wireless broadcast channels

A Novel Data Placement and Retrieval Service for Cooperative Edge Clouds

Mobile edge computing is a new paradigm in which the computing and storage resources are placed at the edge of the Internet. Data placement and retrieval are fundamental services of mobile edge computing when a network of edge clouds collaboratively provide data services. However existing methods such as distributed hash tables (DHTs) are not enough to achieve efficient data placement and retrieval services for cooperative edge clouds. This paper presents GRED, a novel data placement and retrieval service for mobile edge computing, which is efficient in not only the load balance but also routing path lengths and forwarding table sizes. GRED utilizes the programmable switches to support a virtual-space based DHT with only one overlay hop. Data location can be easily implemented on top of the GRED. We implement GRED in a P4 prototype, which provides a simple and efficient solution. Results from theoretical analysis, simulations, and experiments show that GRED can efficiently balance the load of edge clouds, and can fast answer data queries due to its low routing stretch.

Counterintuitive Characteristics of Optimal Distributed LRU Caching Over Unreliable Channels

Least-recently-used (LRU) caching and its variants have conventionally been used as a fundamental and critical method to ensure fast and efficient data access in computer and communication systems. Emerging data-intensive applications over unreliable channels, e.g., mobile edge computing and wireless content delivery networks, have imposed new challenges in optimizing LRU caching in environments prone to failures. Most existing studies focus on reliable channels, e.g., on wired Web servers and within data centers, which have already yielded good insights and successful algorithms. Surprisingly, we show that these insights do not necessarily hold true for unreliable channels. We consider a single-hop multi-cache distributed system with data items being dispatched by random hashing. The objective is to design efficient cache organization and data placement that minimize the miss probability. The former allocates the total memory space to each of the involved caches. The latter decides data routing and replication strategies. Analytically, we characterize the asymptotic miss probabilities for unreliable LRU caches, and optimize the system design. Remarkably, these results sometimes are counterintuitive, differing from the ones obtained for reliable caches. We discover an interesting phenomenon: allocating the cache space unequally can achieve a better performance, even when channel reliability levels are equal. In addition, we prove that splitting the total cache space into separate LRU caches can achieve a lower asymptotic miss probability than organizing the total space in a single LRU cache. These results provide new and even counterintuitive insights that motivate novel designs for caching systems over unreliable channels.

Fuzzy Theory-Based Data Placement for Scientific Workflows in Hybrid Cloud Environments

In hybrid cloud environments, reasonable data placement strategies are critical to the efficient execution of scientific workflows. Due to various loads, bandwidth fluctuations, and network congestions between different data centers as well as the dynamics of hybrid cloud environments, the data transmission time is uncertain. Thus, it poses huge challenges to the efficient data placement for scientific workflows. However, most of the traditional solutions for data placement focus on deterministic cloud environments, which lead to the excessive data transmission time of scientific workflows. To address this problem, we propose an adaptive discrete particle swarm optimization algorithm based on the fuzzy theory and genetic algorithm operators (DPSO-FGA) to minimize the fuzzy data transmission time of scientific workflows. The DPSO-FGA can rationally place the scientific workflow data while meeting the requirements of data privacy and the capacity limitations of data centers. Simulation results show that the DPSO-FGA can effectively reduce the fuzzy data transmission time of scientific workflows in hybrid cloud environments.

Artificial Intelligence for Wireless Caching: Schemes, Performance, and Challenges

Wireless data traffic is growing unprecedentedly and it may impede network performance by consuming an ever-greater amount of bandwidth. With the advancement in technology there exist profound techniques having potentials to improve performance of wireless networks. Artificial Intelligence (AI) is one such evolving technology that enables systems to take intelligent decisions. AI can be incorporated in wireless networks for performing an optimal data caching based on accurate predictions of users' data requests and data popularity profile. AI-based data caching is a promising candidate to effectively harness the issues of rising backhaul data traffic of future wireless networks such as duplicate data transmission and data access delay. In this paper, we provide a systematic survey of state-of-the-art intelligent data caching approaches based on learning mechanism to optimize data caching. First we give an overview of traditional caching approaches and their limitations. Then, after rendering brief introduction of several AI techniques, we introduce state-of-the-art learning approaches in cache-enabled wireless networks. We unfold significant research efforts utilizing AI for efficient data placement for optimizing network performance in terms of cache hit rate, throughput, and offloading etc. Finally, we highlight existing challenges and research directions of AI-based data caching.

Decoupling NDN caches via CCndnS: Design, analysis, and application

In-network caching is considered to be a vital part of the Internet for future applications (e.g., Internet of Things). One proposal that has attracted interest in recent years, Named Data Networking (NDN), aims to facilitate in-network caching by locating content by name. However, the efficiency of in-network caching has been questioned by experts. Data correlation among caches builds strong dependencies between caches at the edge and in the core. That dependency makes analyzing network performance difficult.This paper proposes CCndnS (Content Caching strategy for NDN with Skip), a caching policy to break the dependencies among caches, thus facilitating the design of an efficient data placement algorithm. Specifically, each cache – regardless of its location in the network – should receive an independent set of requests; otherwise, only misses from downstream caches make their way to the upstream caches, i.e. a filtering effect that induces a correlation among the caches.CCndnS breaks a file into smaller segments and spreads them in the path between requester and publisher in a way that the head of the file (the first segment) should be cached at the edge router close to the requester and the tail far from the requester and towards the content provider. Requests for a segment skip searching caches in its path, to search only the cache with the segment of interest. That reduces the number of futile checks on caches, and thus the delay from memory accesses. This mechanism also decouples the caches, so there is a simple analytical model for cache performance in the network.We illustrate an application of the model to enforce a Service Level Agreement (SLA) between a content provider and the caching system proposed in this paper. The model can be used for cache provisioning for two purposes: (1) To specify the cache size to be reserved for specific contents to reach some desired performance. For instance, if the client of an SLA requires a 50% cache hit for its content at each router, the model can be used to determine the cache size that needs to be reserved to reach the 50% hit rate. (2) To calculate the effect of such reservations on other contents that use the routers covered by the SLA.The design, analysis, and application are tested with extensive simulations.

Electricity Price Forecasting for Cloud Computing Using an Enhanced Machine Learning Model

Cloud computing is rapidly taking over the information technology industry because it makes computing a lot easier without worries of buying the physical hardware needed for computations, rather, these services are hosted by companies with provide the cloud services. These companies contain a lot of computers and servers whose main source of power is electricity, hence, design and maintenance of these companies is dependent on the availability of steady and cheap electrical power supply. Cloud centers are energy-hungry. With recent spikes in electricity prices, one of the main challenges in designing and maintenance of such centers is to minimize electricity consumption of data centers and save energy. Efficient data placement and node scheduling to offload or move storage are some of the main approaches to solve these problems. In this article, we propose an Extreme Gradient Boosting (XGBoost) model to offload or move storage, predict electricity price, and as a result reduce energy consumption costs in data centers. The performance of this method is evaluated on a real-world dataset provided by the Independent Electricity System Operator (IESO) in Ontario, Canada, to offload data storage in data centers and efficiently decrease energy consumption. The data is split into 70% training and 30% testing. We have trained our proposed model on the data and validate our model on the testing data. The results indicate that our model can predict electricity prices with a mean squared error (MSE) of 15.66 and mean absolute error (MAE) of 3.74% respectively, which can result in 25.32% cut in electricity costs. The accuracy of our proposed technique is 91% while the accuracy of benchmark algorithms RF and SVR is 89% and 88%, respectively.

Graph-based model and algorithm for minimising big data movement in a cloud environment

In this paper, we discuss load balancing and data placement strategies in cloud environments. The main goal in data placement strategies is to improve the overall performance through the reduction of data movements among the participating datacentres. Load balancing and efficient data placement on cloud systems are critical problems that are difficult to simultaneously cope with. In this context, we propose a threshold-based load balancing algorithm, which first balances the load between datacentres, and afterwards minimises the overhead of data exchanges. It is divided into three phases. First, the dependencies between the datasets are identified. Second, the load threshold of each datacentre is estimated based on the processing speed and the storage capacity. Third, the load balancing between the datacentres is managed through the threshold parameters. Our experimental results show that our approach can efficiently reduce the frequency of data movement and keep a good load balancing between the datacentres.

Special Issue on Infrastructures and Algorithms for Scalable Computing

We are happy to present this special issue of the scientific journal Scalable Computing: Practice and Experience. In this special issue on Infrastructures and Algorithms for Scalable Computing (Volume 19, No 3 June 2018), we have selected four papers out of submitted nine, which gone through a peer review according to the journal policy. All papers represent novel results in the fields of distributed algorithms and infrastructures for scalable computing.
 The first paper presents present a novel approach for efficient data placement, which improves the performance of workflow execution in distributed datacenters. The greedy heuristic algorithm, which is based on a network flow optimization framework, minimizes the total storage cost, including efforts to move and store the data from different source locations and dependencies. The second paper evaluated the significance of different clustering techniques viz. k-means, Hierarchical Agglomerative Clustering and Markov Clustering in groupingawaredata placement for data-intensive applications with interest locality. The evaluation in Azure reported that Markov Clustering-based data placement strategy improves the local map execution and reduces the execution time compared to Hadoops Default Data Placement Strategy and other evaluated clustering techniques. This is more emphasized for data-intensive applications that have interest locality. The third paper presents an experimental evaluation of the openMP thread-mapping strategies in different hardware environments (IntelXeon Phi coprocessor and hybrid CPU-MIC platforms). The paper shows the optimal choice of thread affinity, the number of threads and the execution mode that can provide optimal performance of the LU factorization. In the fourth paper, the authors study the amount of memory occupied by sparse matrices split up into same-size blocks. The paper considers and statistically evaluates four popular storage formats and combinations among them. The conclusion is that block-based storage formats may significantly reduce memory footprints of sparse matrices arising from a wide range of application domains.
 We use this opportunity to thank all contributors to this Special Issue: all authors who submitted the results of their latest research and all reviewers for their valuable comments and suggestions for improvement. We would like to express our special gratitude for the Editor-in-Chief, Professor Dana Petcu, for her constant support during the whole process of this Special Issue.

Exact and Heuristic Data Workflow Placement Algorithms for Big Data Computing in Cloud Datacenters

Several big data-driven applications are currently carried out in collaboration using distributed infrastructure. These data-driven applications usually deal with experiments at massive scale. Data generated by such experiments are huge and stored at multiple geographic locations for reuse. Workflow systems, composed of jobs using collaborative task-based models, present new dependency and data exchange needs. This gives rise to new issues when selecting distributed data and storage resources so that the execution of applications is on time, and resource usage-cost-efficient. In this paper, we present an efficient data placement approach to improve the performance of workflow processing in distributed data centres. The proposed approach involves two types of data: splittable and unsplittable intermediate data. Moreover, we place intermediate data by considering not only their source location but also their dependencies. The main objective is to minimise the total storage cost, including the effort for transferring, storing, and moving that data according to the applications needs. We first propose an exact algorithm which takes into account the intra-job dependencies, and we show that the optimal fractional intermediate data placement problem is NP-hard. To solve the problem of unsplittable intermediate data placement, we propose a greedy heuristic algorithm based on a network flow optimisation framework. The experimental results show that the performance of our approach is very promising. We also show that even with divergent conditions, the cost ratio of the heuristic approach is close to the optimal solution.

Graph-based model and algorithm for minimising big data movement in a cloud environment

In this paper, we discuss load balancing and data placement strategies in cloud environments. The main goal in data placement strategies is to improve the overall performance through the reduction of data movements among the participating datacentres. Load balancing and efficient data placement on cloud systems are critical problems that are difficult to simultaneously cope with. In this context, we propose a threshold-based load balancing algorithm, which first balances the load between datacentres, and afterwards minimises the overhead of data exchanges. It is divided into three phases. First, the dependencies between the datasets are identified. Second, the load threshold of each datacentre is estimated based on the processing speed and the storage capacity. Third, the load balancing between the datacentres is managed through the threshold parameters. Our experimental results show that our approach can efficiently reduce the frequency of data movement and keep a good load balancing between the datacentres.

Cloudkit

CloudKit is Apple's cloud backend service and application development framework that provides strongly-consistent storage for structured data and makes it easy to synchronize data across user devices or share it among multiple users. Launched more than 3 years ago, CloudKit forms the foundation for more than 50 Apple apps, including many of our most important and popular applications such as Photos, iCloud Drive, Notes, Keynote, and News, as well as many third-party apps. To deliver this at large scale, CloudKit explicitly leverages multi-tenancy at the application level as well as at the user level to guide efficient data placement and distribution. By using CloudKit application developers are free to focus on delivering the application front-end and logic while relying on CloudKit for scale, consistency, durability and security. CloudKit manages petabytes of data and handles hundreds of millions of users around the world on a daily basis.

Byzantine fault-tolerant and semantic-driven consensus protocol

To provide fault tolerance, modern distributed storage systems use specialized network topologies and consensus protocols that create high overheads. The main disadvantage of existing specialized topologies is a difficulty to implement an efficient data placement that takes into account locality of the data. In scientific problems very often it is necessary to provide the semantic proximity of the data at the nodes of the distributed system. Core drawback of modern consensus protocols is that most of them use messaging through broadcasting, which is impossible for large amounts of data and storage nodes. With more stringent requirements for a fault tolerance, e.g. requirement for resistance to Byzantine errors [1], ensuring of data localization becomes nearly impossible. The purpose of this paper is to implement a distributed consensus protocol that preserves data localization and is resistant to Byzantine errors. Theoretical tasks: a) consider how to implement distributed services based on finite state machines; b) consider how to ensure Byzantine fault tolerance; c) consider methods for organizing the semantic proximity of data. Practical tasks: a) develop ways to jointly ensure the Byzantine fault tolerance and locality of data; b) study the shortcomings of the joint provision of Byzantine fault tolerance and locality of data; c) develop a protocol of Byzantine consensus using semantic links between data; d) analyze the performance of the developed protocol within the framework of applied tasks. The object of the study is a distributed data storage system with the requirement for Byzantine fault tolerance. The subject of this study is the provision of Byzantine fault tolerance in conjunction with the use of semantic links between the data. The subject of this study has never been the subject of a special scientific study. Separate scientific foundations of this problematic set in the present study were developed in separate works on Byzantine fault tolerance, local hash methods and dimensional reduction.

Efficient location-aware data placement for data-intensive applications in geo-distributed scientific data centers

Recent developments in cloud computing and big data have spurred the emergence of data-intensive applications for which massive scientific datasets are stored in globally distributed scientific data centers that have a high frequency of data access by scientists worldwide. Multiple associated data items distributed in different scientific data centers may be requested for one data processing task, and data placement decisions must respect the storage capacity limits of the scientific data centers. Therefore, the optimization of data access cost in the placement of data items in globally distributed scientific data centers has become an increasingly important goal. Existing data placement approaches for geo-distributed data items are insufficient because they either cannot cope with the cost incurred by the associated data access, or they overlook storage capacity limitations, which are a very practical constraint of scientific data centers. In this paper, inspired by applications in the field of high energy physics, we propose an integer-programming-based data placement model that addresses the above challenges as a Non-deterministic Polynomial-time (NP)-hard problem. In addition we use a Lagrangian relaxation based heuristics algorithm to obtain ideal data placement solutions. Our simulation results demonstrate that our algorithm is effective and significantly reduces overall data access cost.

Analyzing Enterprise Storage Workloads With Graph Modeling and Clustering

Utilizing graph analysis models and algorithms to exploit complex interactions over a network of entities is emerging as an attractive network analytic technology. In this paper, we show that traditional column or row-based trace analysis may not be effective in deriving deep insights hidden in the storage traces collected over complex storage applications, such as complex spatial and temporal patterns, hotspots and their movement patterns. We propose a novel graph analytics framework, GraphLens, for mining and analyzing real world storage traces with three unique features. First, we model storage traces as heterogeneous trace graphs in order to capture multiple complex and heterogeneous factors, such as diverse spatial/temporal access information and their relationships, into a unified analytic framework. Second, we employ and develop an innovative graph clustering method that employs two levels of clustering abstractions on storage trace analysis. We discover interesting spatial access patterns and identify important temporal correlations among spatial access patterns. This enables us to better characterize important hotspots and understand hotspot movement patterns. Third, at each level of abstraction, we design a unified weighted similarity measure through an iterative dynamic weight learning algorithm. With an optimal weight assignment scheme, we can efficiently combine the correlation information for each type of storage access patterns, such as random versus sequential, read versus write, to identify interesting spatial/temporal correlations hidden in the traces. Some optimization techniques on matrix computation are proposed to further improve the efficiency of our clustering algorithm on large trace datasets. Extensive evaluation on real storage traces shows GraphLens can provide broad and deep trace analysis for better storage strategy planning and efficient data placement guidance. GraphLens can be applied to both a single PC with multiple disks and a distributed network across a cluster of compute nodes to offer a few opportunities for optimization of storage performance.

Exploiting CMS data popularity to model the evolution of data management for Run-2 and beyond

During the LHC Run-1 data taking, all experiments collected large data volumes from proton-proton and heavy-ion collisions. The collisions data, together with massive volumes of simulated data, were replicated in multiple copies, transferred among various Tier levels, transformed/slimmed in format/content. These data were then accessed (both locally and remotely) by large groups of distributed analysis communities exploiting the WorldWide LHC Computing Grid infrastructure and services. While efficient data placement strategies - together with optimal data redistribution and deletions on demand - have become the core of static versus dynamic data management projects, little effort has so far been invested in understanding the detailed data-access patterns which surfaced in Run-1. These patterns, if understood, can be used as input to simulation of computing models at the LHC, to optimise existing systems by tuning their behaviour, and to explore next-generation CPU/storage/network co-scheduling solutions. This is of great importance, given that the scale of the computing problem will increase far faster than the resources available to the experiments, for Run-2 and beyond. Studying data-access patterns involves the validation of the quality of the monitoring data collected on the “popularity of each dataset, the analysis of the frequency and pattern of accesses to different datasets by analysis end-users, the exploration of different views of the popularity data (by physics activity, by region, by data type), the study of the evolution of Run-1 data exploitation over time, the evaluation of the impact of different data placement and distribution choices on the available network and storage resources and their impact on the computing operations. This work presents some insights from studies on the popularity data from the CMS experiment. We present the properties of a range of physics analysis activities as seen by the data popularity, and make recommendations for how to tune the initial distribution of data in anticipation of how it will be used in Run-2 and beyond.