Dynamic Simulation Of Database Runtime Environment

With the popularity of cloud environments,more and more applications are migrated to the cloud.The diversity of application programs that share resources in cloud platforms,as well as the dynamic behavior of workload,leads to the complexity and uncertainty of the database runtime environment.In a multi-tenant scenario,the performance of the current user’s workload may be affected by concurrent transactions or queries from other tenants that are also running in the database.In order to more objectively and accurately reflect the stability and reliability of the database operation,simulating the external and internal runtime environment of the database is an important and urgent task to support effective benchmarking of the database.The existing simulation work for the external runtime environment of databases creates complex environments by injecting faults and simulating resource consumption.However,these tools have limitations in terms of versatility,comprehensiveness,efficiency,and accuracy of simulation results.Moreover,most of them are incapable of simulating dynamic runtime environments.For the simulation of the internal runtime environment of databases,most of the existing work applies workload pressure to the database.Currently,there is no work that dynamically simulates the internal runtime environment of databases from the perspective of system resource consumption.This paper designs a dynamic simulation tool for the database runtime environment,which can simulate the changes in the environment of hardware resources and software resources from multiple dimensions.It achieves a versatile,comprehensive,accurate,and efficient simulation of the external and internal runtime environment of the database,and is suitable for simulating composite environments and continuously changing dynamic environments.Finally,through experiments,the superiority and usability of this tool in custom scenarios are demonstrated.The main work of this paper is as follows:(1)Simulation of the external runtime environment of the database: This paper defines the workloads that simulate the external environment of databases in a cloud environment with fine granularity and high efficiency,in order to simulate the hardware and software resource consumption of the database’s external system.This module can simulate the status of system resource consumption by applications in other containers in the runtime environment.It can also simulate extreme scenarios where the resources of the container hosting the database are fully occupied or even destroyed.(2)Simulation of the internal runtime environment of the database: This paper defines flexible and controllable workloads for internal hardware and software resource consumption of a database,specifically targeting multi-tenant scenarios.It provides a comprehensive simulation of the runtime environment where the resources required by the user tenant to execute workloads are preempted by the workloads of other tenants.(3)Composite environment Modeling: This paper utilizes a linear model to capture the potential correlations among different resource simulation workloads.It also presents an agile approach to adjust the actual inputs of these workloads,which results in a simulated environment that closely resembles the target environment.This approach enhances the accuracy of complex runtime environment simulations.(4)Dynamic environment simulation: This paper presents a scheme for modeling dynamic environments and proposes a fast switching algorithm based on the dichotomous iterative method to handle frequent changes in the target environment.The proposed approach aims to enhance the efficiency of dynamic environment simulation.(5)Environmental workload description language: In order to cooperate with the use of environmental simulation tool,this tool also provides a set of expressive and userfriendly environmental workload description language to assist users in writing the environmental workload.This language is convenient for users to customize the type of workload,concurrency situation and dynamic changes in workload intensity.This tool can facilitate users to efficiently implement custom environment workloads and simulate the target environment with fine granularity and high accuracy.As a result,the difficulty of database benchmarking is lowed down,and benchmarking efficiency is improved.Additionally,it enhances the validity and referenceability of benchmarking results while fully supporting the robustness evaluation of database systems.