Query Scheduling Research Articles

Learning-based SPARQL query performance modeling and prediction

One of the challenges of managing an RDF database is predicting performance of SPARQL queries before they are executed. Performance characteristics, such as the execution time and memory usage, can help data consumers identify unexpected long-running queries before they start and estimate the system workload for query scheduling. Extensive works address such performance prediction problem in traditional SQL queries but they are not directly applicable to SPARQL queries. In this paper, we adopt machine learning techniques to predict the performance of SPARQL queries. Our work focuses on modeling features of a SPARQL query to a vector representation. Our feature modeling method does not depend on the knowledge of underlying systems and the structure of the underlying data, but only on the nature of SPARQL queries. Then we use these features to train prediction models. We propose a two-step prediction process and consider performances in both cold and warm stages. Evaluations are performed on real world SPRAQL queries, whose execution time ranges from milliseconds to hours. The results demonstrate that the proposed approach can effectively predict SPARQL query performance and outperforms state-of-the-art approaches.

A Routing Technique for Query Scheduling and Forwarding in Unstructured Overlay Topology Networks

A Routing Technique for Query Scheduling and Forwarding in Unstructured Overlay Topology Networks

Dynamic priority scheduling of periodic queries in on-demand data dissemination systems

As an important wireless data broadcast technique, on-demand broadcast has been widely used for dynamic and large-scale data dissemination. An important class of emerging data broadcast applications requires monitoring multiple data items continuously in order to support data-driven decision making. Since wireless bandwidth is a precious shared medium, an important problem to solve is how to disseminate data to periodic queries, so that all the requests can be satisfied while the bandwidth consumption is minimized. In this paper, we first propose a new real-time scheduling algorithm called EDFS, which is a variant of the classic EDF [24] algorithm. Based on EDFS, we propose a novel on-line broadcast scheduling algorithm, called EDFS-BS. To our best knowledge, EDFS-BS is the first dynamic priority based broadcast scheduling algorithm that can be utilized to satisfy the timing constraints of periodic queries. We also propose a bandwidth utilization based schedulability test for EDFS-BS, which is used to ensure timing predictability of a periodic query set. Extensive experiments have been conducted to compare EDFS-BS versus existing solutions with comparable quality. The results show that EDFS-BS outperforms them considerably in terms of wireless bandwidth consumption and query service ratio.

LocationSpark

We present LocationSpark, a spatial data processing system built on top of Apache Spark, a widely used distributed data processing system. LocationSpark offers a rich set of spatial query operators, e.g., range search, k NN, spatio-textual operation, spatial-join, and k NN-join. To achieve high performance, LocationSpark employs various spatial indexes for in-memory data, and guarantees that immutable spatial indexes have low overhead with fault tolerance. In addition, we build two new layers over Spark, namely a query scheduler and a query executor. The query scheduler is responsible for mitigating skew in spatial queries, while the query executor selects the best plan based on the indexes and the nature of the spatial queries. Furthermore, to avoid unnecessary network communication overhead when processing overlapped spatial data, We embed an efficient spatial Bloom filter into LocationSpark's indexes. Finally, LocationSpark tracks frequently accessed spatial data, and dynamically flushes less frequently accessed data into disk. We evaluate our system on real workloads and demonstrate that it achieves an order of magnitude performance gain over a baseline framework.

Acentric Scheduling Strategy for SLA-Based Multi-Tenant Queries

The response time for multi-tenant queries is one of the most important indicators in the service level agreements (SLA). The service provider tries to optimize query scheduling strategy to finish queries of different tenants before the deadline to avoid penalty due to jeopardizing the SLA. With continuous expansion of tenants scale, peer-to-peer (P2P) structure becomes more and more popular in organizing and managing multi-tenant data. In this paper we propose an acentric scheduling approach for SLA-based multi-tenant queries according to the distribution characteristics of multi-tenant data. Our scheduling approach deploys multiple scheduling engines on the computing nodes in the cloud, where the computing node of each engine schedules its assigned queries, estimates whether these queries could be finished before the deadline, and migrates the queries that might jeopardize the SLA to another engine which can respond to it before the deadline. Since the high efficiency of the scheduling process is critical, we improve the balanced binary tree and use it to organize queries on each computing node. Using this structure, the online time complexity of the scheduling strategy is [Formula: see text]. Our extensive experiments demonstrate that our scheduling strategy is sufficient to meet the high scheduling efficiency requirement, while the penalty cost can be reduced up to [Formula: see text] compared with the benchmarking solution and with a better scalability.

Prediction and Predictability for Search Query Acceleration

A commercial web search engine shards its index among many servers, and therefore the response time of a search query is dominated by the slowest server that processes the query. Prior approaches target improving responsiveness by reducing the tail latency , or high-percentile response time, of an individual search server. They predict query execution time, and if a query is predicted to be long-running, it runs in parallel; otherwise, it runs sequentially. These approaches are, however, not accurate enough for reducing a high tail latency when responses are aggregated from many servers because this requires each server to reduce a substantially higher tail latency (e.g., the 99.99th percentile), which we call extreme tail latency. To address tighter requirements of extreme tail latency, we propose a new design space for the problem, subsuming existing work and also proposing a new solution space. Existing work makes a prediction using features available at indexing time and focuses on optimizing prediction features for accelerating tail queries. In contrast, we identify “when to predict?” as another key optimization question. This opens up a new solution of delaying a prediction by a short duration to allow many short-running queries to complete without parallelization and, at the same time, to allow the predictor to collect a set of dynamic features using runtime information. This new question expands a solution space in two meaningful ways. First, we see a significant reduction of tail latency by leveraging “dynamic” features collected at runtime that estimate query execution time with higher accuracy. Second, we can ask whether to override prediction when the “predictability” is low. We show that considering predictability accelerates the query by achieving a higher recall. With this prediction, we propose to accelerate the queries that are predicted to be long-running. In our preliminary work, we focused on parallelization as an acceleration scenario. We extend to consider heterogeneous multicore hardware for acceleration. This hardware combines processor cores with different microarchitectures such as energy-efficient little cores and high-performance big cores, and accelerating web search using this hardware has remained an open problem. We evaluate the proposed prediction framework in two scenarios: (1) query parallelization on a multicore processor and (2) query scheduling on a heterogeneous processor. Our extensive evaluation results show that, for both scenarios of query acceleration using parallelization and heterogeneous cores, the proposed framework is effective in reducing the extreme tail latency compared to a start-of-the-art predictor because of its higher recall, and it improves server throughput by more than 70% because of its improved precision.

WiSeDB

Workload management for cloud databases deals with the tasks of resource provisioning, query placement, and query scheduling in a manner that meets the application's performance goals while minimizing the cost of using cloud resources. Existing solutions have approached these three challenges in isolation while aiming to optimize a single performance metric. In this paper, we introduce WiSeDB, a learning-based framework for generating holistic workload management solutions customized to application-defined performance goals and workload characteristics. Our approach relies on supervised learning to train cost-effective decision tree models for guiding query placement, scheduling, and resource provisioning decisions. Applications can use these models for both batch and online scheduling of incoming workloads. A unique feature of our system is that it can adapt its offline model to stricter/looser performance goals with minimal re-training. This allows us to present to the application alternative workload management strategies that address the typical performance vs. cost trade-off of cloud services. Experimental results show that our approach has very low training overhead while offering low cost strategies for a variety of performance metrics and workload characteristics.

Hetero-DB: Next Generation High-Performance Database Systems by Best Utilizing Heterogeneous Computing and Storage Resources

With recent advancement on hardware technologies, new general-purpose high-performance devices have been widely adopted, such as the graphics processing unit (GPU) and solid state drive (SSD). GPU may offer an order of higher throughput for applications with massive data parallelism, compared with the multicore CPU. Moreover, new storage device SSD is also capable of offering a much higher I/O throughput and lower latency than a traditional hard disk device (HDD). These new hardware devices can significantly boost the performance of many applications; thus the database community has been actively engaging in adopting them into database systems. However, the performance benefit cannot be easily reaped if the new hardwares are improperly used. In this paper, we propose Hetero-DB, a high-performance database system by exploiting both the characteristics of the database system and the special properties of the new hardware devices in system’s design and implementation. Hetero-DB develops a GPU-aware query execution engine with GPU device memory management and query scheduling mechanism to support concurrent query execution. Furthermore, with the SSD-HDD hybrid storage system, we redesign the storage engine by organizing HDD and SSD into a two-level caching hierarchy in Hetero-DB. To best utilize the hybrid hardware devices, the semantic information that is critical for storage I/O is identified and passed to the storage manager, which has a great potential to improve the efficiency and performance. Hetero-DB aims to maximize the performance benefits of GPU and SSD, and demonstrates the effectiveness for designing next generation database systems.

Query Scheduling Techniques and Power-Latency Trade-off Model for Large-Scale Search Engines

Web search engines have to deal with a huge increase of information, demanded by high incoming query traffic. This situation has driven companies to build large, geographically distributed data centres housing thousands of servers and consuming enormous amounts of electricity. At this scale, even minor efficiency improvements may result in large financial and power savings. This thesis represents a novel contribution to the state-of-the-art of Query Scheduling and Green Information Retrieval (Green IR), by assisting large-scale data centres to build more efficient and environmentally-friendly search engines. The main contributions of this work are the following: Query Scheduling. We introduce query efficiency predictors as suitable estimators to improve Query Scheduling. We estimate the processing time of the queries waiting in each query server and we calculate an approximate time that a new query must spend in each queue. Based on this estimation, the fastest query server is selected. Green IR. Once we have developed new methods to improve the average response time of a search engine, we focus on reducing the power consumption of the whole system. This thesis proposes a mathematical model that establishes a trade-off between latency and power consumption. This model attempts to automatically adapt the number of active servers in the system based on the fluctuations of a daily query traffic flow. Queueing Theory. We prove the limitation of Queueing Theory models for estimating the latency in search engines. As a consequence, we develop our trade-off model by predicting the latency using historical data. Results show the good performance of this approach. IR evaluation. We attest that Simulation platforms are suitable for IR experimentation. We support this conclusion by establishing an exhaustive analysis of the current IR evaluation platforms. .

EM-KDE: A locality-aware job scheduling policy with distributed semantic caches

In modern query processing systems, the caching facilities are distributed and scale with the number of servers. To maximize the overall system throughput, the distributed system should balance the query loads among servers and also leverage cached results. In particular, leveraging distributed cached data is becoming more important as many systems are being built by connecting many small heterogeneous machines rather than relying on a few high-performance workstations. Although many query scheduling policies exist such as round-robin and load-monitoring, they are not sophisticated enough to both balance the load and leverage cached results. In this paper, we propose distributed query scheduling policies that take into account the dynamic contents of distributed caching infrastructure and employ statistical prediction methods into query scheduling policy.We employ the kernel density estimation derived from recent queries and the well-known exponential moving average (EMA) in order to predict the query distribution in a multi-dimensional problem space that dynamically changes. Based on the estimated query distribution, the front-end scheduler assigns incoming queries so that query workloads are balanced and cached results are reused. Our experiments show that the proposed query scheduling policy outperforms existing policies in terms of both load balancing and cache hit ratio.

Multi-dimensional multiple query scheduling with distributed semantic caching framework

It is becoming more important to leverage a large number of distributed cache memory seamlessly in modern large scale systems. Several previous studies showed that traditional scheduling policies often fail to exhibit high cache hit ratio and to achieve good system load balance with large scale distributed caching facilities. To maximize the system throughput, distributed caching facilities should balance the workloads and leverage cached data at the same time. In this work, we present a distributed job processing framework that yields high cache hit ratio while achieving balanced system load. Our framework employs a scheduling policy--DEMA that considers both cache hit ratio and system load and it supports geographically distributed multiple job schedulers. We show collaborative task scheduling and the data migration can even further improve the performance by increasing the cache hit ratio while achieving good load balance. Our experiments show that the proposed job scheduling policies outperform legacy load-based job scheduling policy in terms of job response time, load balancing, and cache hit ratio.

Branch-based centralized data collection for smart grids using wireless sensor networks.

A smart grid is one of the most important applications in smart cities. In a smart grid, a smart meter acts as a sensor node in a sensor network, and a central device collects power usage from every smart meter. This paper focuses on a centralized data collection problem of how to collect every power usage from every meter without collisions in an environment in which the time synchronization among smart meters is not guaranteed. To solve the problem, we divide a tree that a sensor network constructs into several branches. A conflict-free query schedule is generated based on the branches. Each power usage is collected according to the schedule. The proposed method has important features: shortening query processing time and avoiding collisions between a query and query responses. We evaluate this method using the ns-2 simulator. The experimental results show that this method can achieve both collision avoidance and fast query processing at the same time. The success rate of data collection at a sink node executing this method is 100%. Its running time is about 35 percent faster than that of the round-robin method, and its memory size is reduced to about 10% of that of the depth-first search method.

Application of Machine Learning Algorithms for the Query Performance Prediction

This paper analyzes the relationship between the system load/throughput and the query response time in a real Online transaction processing (OLTP) system environment. Although OLTP systems are characterized by short transactions, which normally entail high availability and consistent short response times, the need for operational reporting may jeopardize these objectives. We suggest a new approach to performance prediction for concurrent database workloads, based on the system state vector which consists of 36 attributes. There is no bias to the importance of certain attributes, but the machine learning methods are used to determine which attributes better describe the behavior of the particular database server and how to model that system. During the learning phase, the system's profile is created using multiple reference queries, which are selected to represent frequent business processes. The possibility of the accurate response time prediction may be a foundation for automated decision-making for database (DB) query scheduling. Possible applications of the proposed method include adaptive resource allocation, quality of service (QoS) management or real-time dynamic query scheduling (e.g. estimation of the optimal moment for a complex query execution).

A Query Beehive Algorithm for Data Warehouse Buffer Management and Query Scheduling

Analytical queries, like those used in data warehouses and OLAP, are generally interdependent. This is due to the fact that the database is usually modeled with a denormalized star schema or its variants, where most queries pass through a large central fact table. Such interaction has been largely exploited in query optimization techniques such as materialized views. Nevertheless, such approaches usually ignore buffer management and assume queries have a fixed order and are known in advance. We believe such assumptions are too strong and thus they need to be revisited and simplified. In this paper, we study the combination of two problems: buffer management and query scheduling, in both static and dynamic scenarios. We present an NP-hardness study of the joint problem, highlighting its complexity. We then introduce a new and highly efficient algorithm inspired by a beehive. We conduct an extensive experimental evaluation on a real DBMS showing the superiority of our algorithm compared to previous ones as well as its excellent scalability.

Concurrent analytical query processing with GPUs

In current databases, GPUs are used as dedicated accelerators to process each individual query. Sharing GPUs among concurrent queries is not supported, causing serious resource underutilization. Based on the profiling of an open-source GPU query engine running commonly used single-query data warehousing workloads, we observe that the utilization of main GPU resources is only up to 25%. The underutilization leads to low system throughput. To address the problem, this paper proposes concurrent query execution as an effective solution. To efficiently share GPUs among concurrent queries for high throughput, the major challenge is to provide software support to control and resolve resource contention incurred by the sharing. Our solution relies on GPU query scheduling and device memory swapping policies to address this challenge. We have implemented a prototype system and evaluated it intensively. The experiment results confirm the effectiveness and performance advantage of our approach. By executing multiple GPU queries concurrently, system throughput can be improved by up to 55% compared with dedicated processing.

Evaluation of DFTDS algorithm for distributed data warehouse

The distributed data warehouse is mainly based on how the data are being processed and distinguish between dynamic and physically distributed systems. Finding the relevant information from a huge database is a very laborious process and consumes more time. This conflict is addressed using a query scheduling process in the distributed data warehouse is very compact to accomplish these tasks within a few seconds. In this paper, Dynamic Fault Tolerant Dependency Scheduling (DFTDS) algorithm has been proposed to schedule the queries based on their dependency and it automatically allocates the resources by checking the status of the virtual machine based on its acknowledgment. That is, reply to client/user queries in distributed data warehouse systems. Demonstration of the proposed DFTDS algorithm shows a significant reduction in query processing time and memory utilization compared to existing algorithms.

Real-Time Query Scheduling for Wireless Sensor Networks

Recent years have seen the emergence of wireless cyber-physical systems that must support real-time queries of physical environments through wireless sensor networks. This paper proposes Real-Time Query Scheduling (RTQS), a novel approach to conflict-free transmission scheduling for real-time queries in wireless sensor networks. First, we show that there is an inherent tradeoff between latency and real-time capacity in query scheduling. We then present three new real-time schedulers. The nonpreemptive query scheduler supports high real-time capacity but cannot provide low response times to high-priority queries due to priority inversions. The preemptive query scheduler eliminates priority inversions at the cost of reduced capacity. The slack stealing query scheduler combines the benefits of the preemptive and nonpreemptive schedulers to improve the capacity while meeting the end-to-end deadlines of queries. We provide schedulability analysis for each scheduler. The analysis and advantages of our approach are validated through NS2 simulations.

Towards predicting query execution time for concurrent and dynamic database workloads

Predicting query execution time is crucial for many database management tasks including admission control, query scheduling, and progress monitoring. While a number of recent papers have explored this problem, the bulk of the existing work either considers prediction for a single query, or prediction for a static workload of concurrent queries, where by "static" we mean that the queries to be run are fixed and known. In this paper, we consider the more general problem of dynamic concurrent workloads. Unlike most previous work on query execution time prediction, our proposed framework is based on analytic modeling rather than machine learning. We first use the optimizer's cost model to estimate the I/O and CPU requirements for each pipeline of each query in isolation, and then use a combination queueing model and buffer pool model that merges the I/O and CPU requests from concurrent queries to predict running times. We compare the proposed approach with a machine-learning based approach that is a variant of previous work. Our experiments show that our analytic-model based approach can lead to competitive and often better prediction accuracy than its machine-learning based counterpart.

Distribution-based query scheduling

Query scheduling, a fundamental problem in database management systems, has recently received a renewed attention, perhaps in part due to the rise of the "database as a service" (DaaS) model for database deployment. While there has been a great deal of work investigating different scheduling algorithms, there has been comparatively little work investigating what the scheduling algorithms can or should know about the queries to be scheduled. In this work, we investigate the efficacy of using histograms describing the distribution of likely query execution times as input to the query scheduler. We propose a novel distribution-based scheduling algorithm, Shepherd, and show that Shepherd substantially outperforms state-of-the-art point-based methods through extensive experimentation with both synthetic and TPC workloads.

Deadline‐Aware Energy‐Efficient Query Scheduling in Wireless Sensor Networks with Mobile Sink

Mobile sinks are proposed to save sensor energy spent for multihop communication in transferring data to a base station (sink) in Wireless Sensor Networks. Due to relative low speed of mobile sinks, these approaches are mostly suitable for delay-tolerant applications. In this paper, we study the design of a query scheduling algorithm for query-based data gathering applications using mobile sinks. However, these kinds of applications are sensitive to delays due to specified query deadlines. Thus, the proposed scheduling algorithm aims to minimize the number of missed deadlines while keeping the level of energy consumption at the minimum.