Sap Se

Sap Ag

Tobias Scheuer

Norman May

Anastasia Ailamaki

Iraklis Psaroudakis

Dirk Thomsen

Thorsten Glebe

Daniel Booss

Jaeyoung Choi

Ruediger Karl

Patrick Voelker

Hyogi Jung

Robin Joy

Ivan Schreter

Werner Thesing

Andrei Suvernev

Abdelkader Sellami

Johannes Gloeckle

Mathias Gottschlag

Martin Thebes

Jan Wedel

Systems and methods include establishment of a first authenticated database session with a first database user of a database instance hosting a plurality of tenants, determination of a first tenant of the plurality of tenants based on a mapping of the first database user to the first tenant, determination of a first workload class based on the first tenant, and management of resource consumption of the first authenticated database session based on the first workload class.

Mapping of tenant-specific database sessions to workload classes

The present disclosure relates to computer-implemented methods, software, and systems for controlling the execution of concurrent threads executing tasks in a database management system. A first task is executed at a first thread from a pool of threads for execution of tasks at the database management system. It can be identified that the execution of the first task is paused and that the first task is in a sleep mode. In response to identifying that the first task is in an awake mode after the sleep mode, it can be determined whether a current number of threads in the pool that are currently processing tasks in parallel is below an allowed number of threads. It can be determined that the allowed number of threads has been reached and a waiting status can be assigned to the first task at the first thread.

Controlling concurrently executing threads

Systems and methods include establishment of a first database session with a first database user, determine first session variable values associated with the first database session, the first session variable values including a first tenant value, determine a first workload class based on the first tenant value and on zero or more of the other first session variable values, the first workload class including first parameter values specifying first resource consumption limits, and manage resource consumption of the first database session based on the first parameter values.

Workload classes for tenant-level resource isolation

Processor(s) of a sampling profiler can identify an activity of multiple activities serviced by a central processing unit (CPU). Each activity can be performed by computing thread(s) of multiple computing threads executing various subroutines of a computer program. The processor(s) can set a target representing a total number of computing threads required to work simultaneously for a maximal use of the CPU. The processor(s) can determine a number of busy computing threads that are performing the activity by using the CPU. The processor(s) can calculate a number of wasted computing threads that are not performing the activity and not using the CPU by computing a difference between the target and the number of busy threads. The processor(s) can compute a CPU time usage for the activity by multiplying time duration of the activity by a value obtained by dividing the number of wasted threads by the number of busy threads.

Computing CPU time usage of activities serviced by CPU

Aspects of the current subject matter are directed to an approach in which a parallel load operation of file ID mapping containers is accomplished at start and/or restart of a database system. Parallel load operation of file ID mapping and/or large binary object (LOB) file ID mapping is done among a plurality of scanning engines into a plurality of data buffers that are associated with each of the plurality of scanning engines. Each scanning engine operates on a certain path of a page chain of a page structure including the mapping, causing the page chain to be split among scanning engines to process maps. Contents of the data buffers are pushed to mapping engines via a queue. The mapping engines load the file ID mapping and the LOB file ID mapping into maps for in-system access.

Parallel load of mapping containers for database system start and restart operations

Techniques and solutions are provided for performing adaptive database table placement in a non-uniform memory access (NUMA) architecture. The adaptive table placement can occur in response to changing workloads on the NUMA nodes. For example, if a particular NUMA node is saturated, a database table may be moved from the memory of the saturated NUMA node to the memory of another NUMA node that is underutilized. In some cases, an entire database table is moved, while in other cases the database table is partitioned and only part of the table is moved.

Adaptive table placement in NUMA architectures

Systems and methods are described herein for system critical phase lock job inhibitors. Acquisition of a consistent change exclusive lock is initiated. A job request having a scope object is received. Execution of the job request and generation of a replacement job associated with the job request is prohibited based on the scope object indicating that the job requires consistent change access during the consistent change exclusive lock.

Savepoint critical phase lock job inhibitor

Systems and methods are described herein for job execution using system critical threads. An in-memory database system having pages loaded into a memory and having associated physical disk storage generates a safety critical thread pool. The safety critical thread pool has one or more safety critical threads associated with flushing the pages to the physical disk storage. Execution of the one or more safety critical threads is initiated within the safety critical thread pool. A job request to access data stored within the physical disk storage is received. Utilization levels of the one or more safety critical threads are monitored. Execution of the job request is permitted based on the utilization levels associated with the one or more safety critical threads.

Job execution using system critical threads

Systems and method for a task scheduler with dynamic adjustment of concurrency levels and task granularity are disclosed for improved execution of highly concurrent analytical and transactional systems. The task scheduler can avoid both over commitment and underutilization of computing resources by monitoring and controlling the number of active worker threads. The number of active worker threads can be adapted to avoid underutilization of computing resources by giving the OS control of additional worker threads processing blocked application tasks. The task scheduler can dynamically determine a number of parallel operations for a particular task based on the number of available threads. The number of available worker threads can be determined based on the average availability of worker threads in the recent history of the application. Based on the number of available worker threads, the partitionable operation can be partitioned into a number of sub operations and executed in parallel.

Task scheduling for highly concurrent analytical and transaction workloads

Scheduling jobs from an application based on a job concurrency hint. The job concurrency hint providing an indication of the number and/or size of the jobs that can be handled by the job scheduler. The scheduling of the jobs based on the job concurrency hint including selecting the number and/or size of the jobs to pass to the job scheduler for execution by a thread in a core of a processor.

Optimizing job execution in parallel processing with improved job scheduling using job currency hints

Computing threads can be sampled such that the samples hit each activity with a probability proportional to a time period of that activity. Therefore, the sampling can advantageously account for all activities, including activities that are smaller than the applicable sampling interval.

Recording CPU time for sample of computing thread based on CPU use state of activity associated with the sample

Parallel execution of parsed query based on a concurrency level

In a computing system having a multiple central processing unit (CPU) cores the task scheduler can be configured to generate one or more priority value lists of elements, with each priority value list comprising elements having the same priority value. The priority queue of a task scheduler can be populated by links to priority value lists that are arranged in order of priority. Worker threads can access an input SIAO and determine the maximum priority of any element in the input SIAO. If the input SIAO has an element with higher priority than the priority queue of the task scheduler then the worker thread can cause the task associated with that element to be processed, otherwise the worker thread can cause all of the elements of the SIAO to be put into the priority value lists linked to by the elements in the priority queue.

Scaling priority queue for task scheduling

A plurality of virtual files are instantiated in a columnar database. The columnar database includes a columnar data store that persists tables of data contained therein to the plurality of virtual files. Each virtual file is stored in physical data storage and a portion of the virtual file can be cached transiently in a persistence layer intermediate the columnar data store and the physical data storage. The virtual files each comprise a database object for storing data records. Operation are processed, in the columnar database, using the virtual files such that changes to virtual files resulting from the operations of one transaction are isolated from changes on the same virtual files resulting from the operations of other transactions. Related apparatus, systems, techniques and articles are also described.

Growing community of inventors

Bruchsal, Germany

Tobias Scheuer