[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC):2025-09-03。"],[[["\u003cp\u003eYou can adjust the CPU resources for columnar engine maintenance tasks by setting the \u003ccode\u003egoogle_job_scheduler.maintenance_cpu_percentage\u003c/code\u003e flag, with a default of 20% and a minimum of 1 vCPU usage.\u003c/p\u003e\n"],["\u003cp\u003eTo accelerate columnar maintenance, increase the number of worker processes for background jobs using the \u003ccode\u003egoogle_job_scheduler.max_parallel_workers_per_job\u003c/code\u003e flag, with a default of 2 worker processes.\u003c/p\u003e\n"],["\u003cp\u003eThe \u003ccode\u003egoogle_columnar_engine.auto_columnarization_schedule\u003c/code\u003e flag can be used to set the frequency of auto-columnarization, based upon anticipated scan-heavy workloads, without requiring a database restart.\u003c/p\u003e\n"],["\u003cp\u003eIf manual control of the columnar engine population is desired, you can disable auto-columnarization by setting the \u003ccode\u003egoogle_columnar_engine.enable_auto_columnarization\u003c/code\u003e flag to \u003ccode\u003eOFF\u003c/code\u003e.\u003c/p\u003e\n"],["\u003cp\u003eThe \u003ccode\u003egoogle_columnar_engine.population_heap_size_in_mb\u003c/code\u003e flag allows you to tune the memory allocated for data conversion to columnar format, with allowed values ranging from 200 MB to 1000 MB and a default of 250 MB.\u003c/p\u003e\n"]]],[],null,["# Tune the columnar engine for AlloyDB for PostgreSQL\n\nThis document describes how to tune the columnar engine for an AlloyDB instance. This document assumes that you are familiar with PostgreSQL.\n\nAdjust the CPU resources for maintenance tasks\n----------------------------------------------\n\nThe following example shows you how to configure the columnar engine to use 50% of the CPU cores: \n\n google_job_scheduler.maintenance_cpu_percentage = 50\n\nColumnar engine uses up to 20% of CPU cores for background maintenance jobs, such as population, refresh, and recommendation. To adjust the CPU resources to allow for maintenance tasks, set the `google_job_scheduler.maintenance_cpu_percentage` flag to a value from `20` to `100`. The default value is `20`. The lowest CPU utilization for background maintenance tasks is 1 vCPU.\n\nAdd workers to accelerate columnar maintenance\n----------------------------------------------\n\nThe following example shows you how to set four worker processes: \n\n google_job_scheduler.max_parallel_workers_per_job = 4\n\nThe columnar engine uses up to two worker processes for population and refresh background jobs. To increase the speed of these operations, set the following flag to increase the number of worker processes for background jobs. The default setting is 2 worker processes. These processes run on limited CPU resources as specified in the `google_job_scheduler.maintenance_cpu_percentage` flag. This flag doesn't require a database restart.\n\nChange the auto-columnarization schedule\n----------------------------------------\n\nThe following example shows you how to set the auto-columnarization schedule to six hours: \n\n google_columnar_engine.auto_columnarization_schedule = 'EVERY 6 HOURS'\n\nThe auto-columnarization schedule can be set in the `google_columnar_engine.auto_columnarization_schedule` flag based upon anticipated scan-heavy workloads. Specify one or more hours for this flag. Changing the value of this flag doesn't require a database restart.\n\nManually populate the columnar engine\n-------------------------------------\n\nThe following example shows you how to manually populate the columnar engine: \n\n google_columnar_engine.enable_auto_columnarization = OFF\n\nIf you prefer to only manually populate the columnar engine, set the `google_columnar_engine.enable_auto_columnarization` flag to `OFF`. The default value of this flag is `ON`. Changing the value of this flag doesn't require a database restart.\nSee [Manage Content Manually](/alloydb/omni/docs/columnar-engine/manage-content-manually) for more details.\n\nTuning the columnar engine population heap size\n-----------------------------------------------\n\nThe following example shows you how to configure the `google_columnar_engine.population_heap_size_in_mb` flag: \n\n population_heap_size_in_mb = 250\n\nTo convert data to columnar format, the columnar engine allocates the amount of memory in megabytes specified by the `google_columnar_engine.population_heap_size_in_mb` flag. Once the conversion is complete, this memory is immediately deallocated, typically within seconds.\n\nThe allowed values for the `population_heap_size_in_mb` setting range from a minimum of 200 MB to a maximum of 1000 MB.\n\nThe default value of 250 MB for the `population_heap_size_in_mb` flag works well for most typical database usage. However, you might face situations where you need to increase this value."]]