Data Placement Optimization Research Articles

Dynamic Performance and Power Optimization with Heterogeneous Processing-in-Memory for AI Applications on Edge Devices

The rapid advancement of artificial intelligence (AI) technology, combined with the widespread proliferation of Internet of Things (IoT) devices, has significantly expanded the scope of AI applications, from data centers to edge devices. Running AI applications on edge devices requires a careful balance between data processing performance and energy efficiency. This challenge becomes even more critical when the computational load of applications dynamically changes over time, making it difficult to maintain optimal performance and energy efficiency simultaneously. To address these challenges, we propose a novel processing-in-memory (PIM) technology that dynamically optimizes performance and power consumption in response to real-time workload variations in AI applications. Our proposed solution consists of a new PIM architecture and an operational algorithm designed to maximize its effectiveness. The PIM architecture follows a well-established structure known for effectively handling data-centric tasks in AI applications. However, unlike conventional designs, it features a heterogeneous configuration of high-performance PIM (HP-PIM) modules and low-power PIM (LP-PIM) modules. This enables the system to dynamically adjust data processing based on varying computational load, optimizing energy efficiency according to the application’s workload demands. In addition, we present a data placement optimization algorithm to fully leverage the potential of the heterogeneous PIM architecture. This algorithm predicts changes in application workloads and optimally allocates data to the HP-PIM and LP-PIM modules, improving energy efficiency. To validate and evaluate the proposed technology, we implemented the PIM architecture and developed an embedded processor that integrates this architecture. We performed FPGA prototyping of the processor, and functional verification was successfully completed. Experimental results from running applications with varying workload demands on the prototype PIM processor demonstrate that the proposed technology achieves up to 29.54% energy savings.

Automating cloud virtual machines allocation via machine learning

Automating cloud virtual machines allocation via machine learning

Cloud storage cost: a taxonomy and survey

Cloud service providers offer application providers with virtually infinite storage and computing resources, while providing cost-efficiency and various other quality of service (QoS) properties through a storage-as-a-service (StaaS) approach. Organizations also use multi-cloud or hybrid solutions by combining multiple public and/or private cloud service providers to avoid vendor lock-in, achieve high availability and performance, and optimise cost. Indeed cost is one of the important factors for organizations while adopting cloud storage; however, cloud storage providers offer complex pricing policies, including the actual storage cost and the cost related to additional services (e.g., network usage cost). In this article, we provide a detailed taxonomy of cloud storage cost and a taxonomy of other QoS elements, such as network performance, availability, and reliability. We also discuss various cost trade-offs, including storage and computation, storage and cache, and storage and network. Finally, we provide a cost comparison across different storage providers under different contexts and a set of user scenarios to demonstrate the complexity of cost structure and discuss existing literature for cloud storage selection and cost optimization. We aim that the work presented in this article will provide decision-makers and researchers focusing on cloud storage selection for data placement, cost modelling, and cost optimization with a better understanding and insights regarding the elements contributing to the storage cost and this complex problem domain.

HERMES: data placement and schema optimization for enterprise knowledge bases

HERMES: data placement and schema optimization for enterprise knowledge bases

H2RUN: An efficient vendor lock‐in solution for multi‐cloud environment using horse herd Runge Kutta based data placement optimization

AbstractVendor lock‐in is a scenario where the cost of switching to another vendor is mainly high and for this reason, the consumer is mainly struck with a single cloud vendor. The vendor lock‐in mainly implies the situation of an organization where it is locked into a service or product of a single Cloud Service Providers (CSP) to prevent the disruption of the business operation or due to an insufficient workforce. Due to its high availability, portability, and other benefits, various researchers frequently adopt the multi‐cloud strategy to prevent vendor lock‐in. However, the multi‐cloud architecture also comes with major drawbacks such as cost, multiple CSP management, and the tradeoff between cost and availability. The organizations are mainly considerate about achieving the tradeoff between the cost and availability. To overcome this problem, we have proposed an H2RUN (Horse Herd Runge Kutta) optimization algorithm to achieve a tradeoff between the cost and availability using erasure coding. The proposed H2RUN optimization algorithm selects the optimal CSP and erasure coding parameters that enhance the availability of the CSP with minimal cost. The efficiency of the proposed methodology can be evaluated using the experiments conducted on different real‐time CSPs using various metrics. Based on the results achieved, the H2RUN optimizer minimizes the cost up to 48% and availability up to 58%.

A Minimal-overhead Multi-cloud Data Placement Strategy Based on Heuristic Algorithm

Multi-cloud storage can provide better service levels than single cloud storage in terms of confidentiality, integrity, and availability at a low cost. More and more customers are storing data in multi-cloud storage systems. However, inefficient data placement is a challenge for multi-cloud storage, and current research has the problem of low algorithm efficiency. In response to this problem, a multi-cloud storage data placement optimization strategy, MPO-MO, is proposed for minimum overhead requirements. MPO-MO models availability and cost, and provides users with the lowest cost multi-cloud storage data placement strategy while meeting the specified availability constraints. A heuristic algorithm is proposed, which uses the sorting information of storage price and traffic price to select cloud storage services. According to analysis, the heuristic algorithm adopted by MPO-MO greatly reduces the optimization space compared with the traditional algorithm, and the efficiency is significantly improved. Finally, experiments prove the effectiveness of MPO-MO.

Multi-objective Optimization of Data Placement in a Storage-as-a-Service Federated Cloud

Cloud federation enables service providers to collaborate to provide better services to customers. For cloud storage services, optimizing customer object placement for a member of a federation is a real challenge. Storage, migration, and latency costs need to be considered. These costs are contradictory in some cases. In this article, we modeled object placement as a multi-objective optimization problem. The proposed model takes into account parameters related to the local infrastructure, the federated environment, customer workloads, and their SLAs. For resolving this problem, we propose CDP-NSGAII IR , a Constraint Data Placement matheuristic based on NSGAII with Injection and Repair functions. The injection function aims to enhance the solutions’ quality. It consists to calculate some solutions using an exact method then inject them into the initial population of NSGAII. The repair function ensures that the solutions obey the problem constraints and so prevents from exploring large sets of unfeasible solutions. It reduces drastically the execution time of NSGAII. Experimental results show that the injection function improves the HV of NSGAII and the exact method by up to 94% and 60%, respectively, while the repair function reduces the execution time by an average of 68%.

Towards Deadline Guaranteed Cloud Storage Services

More and more organizations move their data and workload to commercial cloud storage systems. However, the multiplexing and sharing of the resources in a cloud storage system present unpredictable data access latency to tenants, which may make online data-intensive applications unable to satisfy their deadline requirements. Thus, it is important for cloud storage systems to provide deadline guaranteed services. In this paper, to meet a current form of service level objective (SLO) that constrains the percentage of each tenant’s data access requests failing to meet its required deadline below a given threshold, we build a mathematical model to derive the upper bound of acceptable request arrival rate on each server. We then propose a Deadline Guaranteed storage service (called DGCloud ) that incorporates three basic algorithms. Its deadline-aware load balancing scheme redirects requests and creates replicas to release the excess load of each server beyond the derived upper bound. Its workload consolidation algorithm tries to maximally reduce servers while still satisfying the SLO to maximize the resource utilization. Its data placement optimization algorithm re-schedules the data placement to minimize the transmission cost of data replication. We further propose three enhancement methods to further improve the performance of DGCloud . A dynamic load balancing method allows an overloaded server to quickly offload its excess workload. A data request queue improvement method sets different priorities to the data responses in a server’s queue so that more requests can satisfy the SLO requirement. A wakeup server selection method selects a sleeping server that stores more popular data to wake up, which allows it to handle more data requests. Our trace-driven experiments in simulation and Amazon EC2 show the superior performance of DGCloud compared with previous methods in terms of deadline guarantees and system resource utilization, and the effectiveness of its individual algorithms.

An Adaptive Data Placement Architecture in Multicloud Environments

Cloud service providers (CSPs) can offer infinite storage space with cheaper maintenance cost compared to the traditional storage mode. Users tend to store their data in geographical and diverse CSPs so as to avoid vendor lock-in. Static data placement has been widely studied in recent works. However, the data access pattern is often time-varying and users may pay more cost if static placement is adopted during the data lifetime. Therefore, it is a pending problem and challenge of how to dynamically store users’ data under time-varying data access pattern. To this end, we propose ADPA, an adaptive data placement architecture that can adjust the data placement scheme based on the time-varying data access pattern and subject for minimizing the total cost and maximizing the data availability. The proposed architecture includes two main components: data retrieval frequency prediction module based on LSTM and data placement optimization module based on Q-learning. The performance of ADPA is evaluated through several experimental scenarios using NASA-HTTP workload and cloud providers information.

Architecture of a distributed storage that combines file system, memory and computation in a single layer

Storage and memory systems for modern data analytics are heavily layered, managing shared persistent data, cached data, and non-shared execution data in separate systems such as a distributed file system like HDFS, an in-memory file system like Alluxio, and a computation framework like Spark. Such layering introduces significant performance and management costs. In this paper, we propose a single system called Pangea that can manage all data—both intermediate and long-lived data, and their buffer/caching, page replacement, data placement optimization, and failure recovery—all in one monolithic distributed storage system, without any layering. We present a detailed performance evaluation of Pangea and show that its performance compares favorably with several widely used layered systems such as Spark.

Pangea

Storage and memory systems for modern data analytics are heavily layered, managing shared persistent data, cached data, and nonshared execution data in separate systems such as a distributed file system like HDFS, an in-memory file system like Alluxio, and a computation framework like Spark. Such layering introduces significant performance and management costs. In this paper we propose a single system called Pangea that can manage all data---both intermediate and long-lived data, and their buffer/caching, data placement optimization, and failure recovery---all in one monolithic distributed storage system, without any layering. We present a detailed performance evaluation of Pangea and show that its performance compares favorably with several widely used layered systems such as Spark.

Deep Learning-Based Data Storage for Low Latency in Data Center Networks

Low-latency data access is becoming an upcoming and increasingly important challenge. The proper placement of data blocks can reduce data travel among distributed storage systems, which contributes significantly to the latency reduction. However, the dominant data placement optimization has primarily relied on prior known data requests or static initial data distribution, which ignores the dynamics of clients' data access requests and networks. The learning technology can help the data center networks (DCNs) learn from historical access information and make optimal data storage decision. Consider a more practical DCNs with fat-tree topology, we utilize a deep-learning technology k-means to help store data blocks and then improve the read and write latency of the DCN, where k is the number of cores in the fat-tree. The evaluation results demonstrate that the average write and read latency of the whole system can be lowered by 33% and 45%, respectively. And the best set of parameter k is analyzed and recommended to provide guidance to the real application, which is equal to the number of cores in the DCNs.

Joint optimization of data placement and scheduling for improving user experience in edge computing

In recent years, edge computing becomes an increasingly popular alternative. Edge computing allows the computation is implemented in the edge of network, in which the data are stored in the edge of network, to improve the efficiency of data process. However, some resource management techniques in cloud or distributed system cannot better suit for edge computing. Therefore, there exist some challenges on the performance improvement of edge computing. In this paper, the main purpose is to combine the optimal placement of data blocks and the optimal scheduling of tasks to reduce the computation delay and response time for the submitted tasks and improve user experience in edge computing. In optimal placement of data blocks, the value of the data blocks considers not only the popularity of the data blocks, but the data storage capacity and replacement ratios of an edge server that will store those data blocks. Furthermore, the replacement cost for placed data blocks is regarded as an important component of data block placement. This optimal placement scheme can avoid replacing the placed data blocks repeatedly so that the bandwidth overhead is reduced. In optimal scheduling of tasks, the containers are taken as the lightweight resource unit for the services for user requests to make full use of data storage in edge servers and improve the services performance of edge servers. Finally, extensive experiments are conducted to value the performance of task scheduling strategy. The results show that the performance of the proposed task scheduling algorithm is better than that of the compared algorithms.

Distributed Joins and Data Placement for Minimal Network Traffic

Network communication is the slowest component of many operators in distributed parallel databases deployed for large-scale analytics. Whereas considerable work has focused on speeding up databases on modern hardware, communication reduction has received less attention. Existing parallel DBMSs rely on algorithms designed for disks with minor modifications for networks. A more complicated algorithm may burden the CPUs but could avoid redundant transfers of tuples across the network. We introduce track join, a new distributed join algorithm that minimizes network traffic by generating an optimal transfer schedule for each distinct join key. Track join extends the trade-off options between CPU and network. Track join explicitly detects and exploits locality, also allowing for advanced placement of tuples beyond hash partitioning on a single attribute. We propose a novel data placement algorithm based on track join that minimizes the total network cost of multiple joins across different dimensions in an analytical workload. Our evaluation shows that track join outperforms hash join on the most expensive queries of real workloads regarding both network traffic and execution time. Finally, we show that our data placement optimization approach is both robust and effective in minimizing the total network cost of joins in analytical workloads.

Characterizing emerging heterogeneous memory

Heterogeneous memory (HM, also known as hybrid memory) has become popular in emerging parallel architectures due to its programming flexibility and energy efficiency. Unlike the traditional memory subsystem, HM consists of fast and slow components. Usually, the fast memory lacks hardware support, which puts extra burdens on programmers and compilers for explicit data placement. Thus, HM provides both opportunities and challenges with programming parallel codes. It is important to understand how to utilize HM and set expectations on the benefits of HM. Prior work principally uses simulators to study HM, which lacks the analysis on a real hardware. To address this issue, this paper experiments with a real system—the TI KeyStone II—to study HM. We make three contributions. First, we develop a set of parallel benchmarks to characterize the performance and power efficiency of HM. It is the first benchmark suite with OpenMP 4.0 features that is functional on real HM architectures. Second, we build a profiling tool to provide guidance for placing data in HM. Our tool analyzes memory access patterns and provides high-level feedback at the source-code level for optimization. Third, we apply the data placement optimization to our benchmarks and evaluate the effectiveness of HM in boosting performance and saving energy.

Systematic Data Placement Optimization in Multi-Cloud Storage for Complex Requirements

Multi-cloud storage can provide better features such as availability and scalability. Current works use multiple cloud storage providers with erasure coding to achieve certain benefits including fault-tolerance improving or vendor lock-in avoiding. However, these works only use the multi-cloud storage in ad-hoc ways, and none of them considers the optimization issue in general. In fact, the key to optimize the multi-cloud storage is to effectively choose providers and erasure coding parameters. Meanwhile, the data placement should satisfy system or application developers’ requirements. As developers often demand various objectives to be optimized simultaneously, such complex requirement optimization cannot be easily fulfilled by ad-hoc ways. This paper presents Triones, a systematic model to formally formulate data placement in multi-cloud storage by using erasure coding. Firstly, Triones addresses the problem of data placement optimization by applying non-linear programming and geometric space abstraction. It could satisfy complex requirements involving multi-objective optimization. Secondly, Triones can effectively balance among different objectives in optimization and is scalable to incorporate new ones. The effectiveness of the model is proved by extensive experiments on multiple cloud storage providers in the real world. For simple requirements, Triones can achieve 50 percent access latency reduction, compared with the model in $\mu$ LibCloud. For complex requirements, Triones can improve fault-tolerance level by 2 $\times$ and reduce access latency and vendor lock-in level by 30 $\sim$ 70 percent and 49.85 percent respectively with about 19.19 percent more cost, compared with the model only optimizing cost in Scalia.

Partitioning of web applications for hybrid cloud deployment

Abstract Hybrid cloud deployment offers flexibility in trade-offs between the cost-savings/scalability of the public cloud and control over data resources provided at a private premise. However, this flexibility comes at the expense of complexity in distributing a system over these two locations. For multi-tier web applications, this challenge manifests itself primarily in the partitioning of application- and database-tiers. While there is existing research that focuses on either application-tier or data-tier partitioning, we show that optimized partitioning of web applications benefits from both tiers being considered simultaneously. We present our research on a new cross-tier partitioning approach to help developers make effective trade-offs between performance and cost in a hybrid cloud deployment. The general approach primarily benefits from two technical improvements to integer-programming based application partitioning. First, an asymmetric cost-model for optimizing data transfer in environments where ingress and egress data-transfer have differing costs, such as in many infrastructure as a service platforms. Second, a new encoding of database query plans as integer programs, to enable simultaneous optimization of code and data placement in a hybrid cloud environment. In two case studies the approach results in up to 54% reduction in monetary costs compared to a premise only deployment and 56% improvement in response time compared to a naive partitioning where the application-tier is deployed in the public cloud and the data-tier is on private infrastructure.

ANTELOPE: A Semantic-Aware Data Cube Scheme for Cloud Data Center Networks

Today’s cloud data centers contain more than millions of servers and offer high bandwidth. A fundamental problem is how to significantly improve the large-scale system’s scalability to interconnect a large number of servers and meanwhile support various online services in cloud computing. One way is to deal with the challenge of potential mismatching between the network architecture and the data placement. To address this challenge, we present ANTELOPE, a scalable distributed data-centric scheme in cloud data centers, in which we systematically take into account both the property of network architecture and the optimization of data placement. The basic idea behind ANTELOPE is to leverage precomputation based data cube to support online cloud services. Since the construction of data cube suffers from the high costs of full materialization, we use a semantic-aware partial materialization solution to significantly reduce the operation and space overheads. Extensive experiments on real system implementations demonstrate the efficacy and efficiency of our proposed scheme.

XRootd, disk-based, caching proxy for optimization of data access, data placement and data replication

Following the success of the XRootd-based US CMS data federation, the AAA project investigated extensions of the federation architecture by developing two sample implementations of an XRootd, disk-based, caching proxy. The first one simply starts fetching a whole file as soon as a file open request is received and is suitable when completely random file access is expected or it is already known that a whole file be read. The second implementation supports on-demand downloading of partial files. Extensions to the Hadoop Distributed File System have been developed to allow for an immediate fallback to network access when local HDFS storage fails to provide the requested block. Both cache implementations are in pre-production testing at UCSD.

THEMIS Component Dependence Metadata in Adaptive Parallel Applications

This paper describes THEMIS, a programming model and run-time library being designed to support cross-component performance optimization through explicit manipulation of the computation's iteration space at run-time. Each is augmented with component dependence metadata, which characterizes the constraints on its execution order, data distribution and memory access order. We show how this supports dynamic adaptation of each to exploit the available resources, the context in which its operands are generated, and results are used, and the evolution of the problem instance. Using a computational fluid dynamics visualization example as motivation, we show how dependence metadata provides a framework in which a number of interesting optimizations become possible. Examples include data placement optimization, loop fusion, tiling, memoization, checkpointing and incrementalization.