Azure Synapse Analytics 的内存和并发限制Memory and concurrency limits for Azure Synapse Analytics

查看分配给 Azure Synapse Analytics 中的各个性能级别和资源类的内存和并发限制。View the memory and concurrency limits allocated to the various performance levels and resource classes in Azure Synapse Analytics.

数据仓库容量设置Data warehouse capacity settings

以下各表显示了不同性能级别的数据仓库的最大容量。The following tables show the maximum capacity for the data warehouse at different performance levels. 若要更改性能级别,请参阅缩放计算 - 门户To change the performance level, see Scale compute - portal.

服务级别Service Levels

服务级别范围为 DW100c 到 DW30000c。The service levels range from DW100c to DW30000c.

性能级别Performance level 计算节点Compute nodes 每个计算节点的分布区数Distributions per Compute node 每个数据仓库的内存 (GB)Memory per data warehouse (GB)
DW100cDW100c 11 6060 6060
DW200cDW200c 11 6060 120120
DW300cDW300c 11 6060 180180
DW400cDW400c 11 6060 240240
DW500cDW500c 11 6060 300300
DW1000cDW1000c 22 3030 600600
DW1500cDW1500c 33 2020 900900
DW2000cDW2000c 44 1515 12001200
DW2500cDW2500c 55 1212 15001500
DW3000cDW3000c 66 1010 18001800
DW5000cDW5000c 1010 66 30003000
DW6000cDW6000c 1212 55 36003600
DW7500cDW7500c 1515 44 45004500
DW10000cDW10000c 2020 33 60006000
DW15000cDW15000c 3030 22 90009000
DW30000cDW30000c 6060 11 1800018000

最大服务级别为 DW30000c,包含 60 个计算节点,每个计算节点有一个分布区。The maximum service level is DW30000c, which has 60 Compute nodes and one distribution per Compute node. 例如,DW30000c 级别的 600 TB 数据仓库的每个计算节点可以处理大约 10 TB 数据。For example, a 600 TB data warehouse at DW30000c processes approximately 10 TB per Compute node.

工作负载组的并发最大值Concurrency maximums for workload groups

随着工作负载组的引入,并发槽位的概念不再适用。With the introduction of workload groups, the concept of concurrency slots no longer applies. 每个请求的资源按百分比分配,并在工作负载组定义中指定。Resources per request are allocated on a percentage basis and specified in the workload group definition. 但是,即使删除了并发槽位,每个查询也需要基于服务级别的最小资源量。However, even with the removal of concurrency slots, there are minimum amounts of resources needed per queries based on the service level. 下表定义了各服务级别协议中,每个查询所需的最小资源量以及可以实现的关联并发。The below table defined the minimum amount of resources needed per query across service levels and the associated concurrency that can be achieved.

服务级别Service Level 最大并行查询Maximum concurrent queries REQUEST_MIN_RESOURCE_GRANT_PERCENT 支持的最小百分比Min % supported for REQUEST_MIN_RESOURCE_GRANT_PERCENT
DW100cDW100c 44 25%25%
DW200cDW200c 88 12.5%12.5%
DW300cDW300c 1212 8%8%
DW400cDW400c 1616 6.25%6.25%
DW500cDW500c 2020 5%5%
DW1000cDW1000c 3232 3%3%
DW1500cDW1500c 3232 3%3%
DW2000cDW2000c 4848 2%2%
DW2500cDW2500c 4848 2%2%
DW3000cDW3000c 6464 1.5%1.5%
DW5000cDW5000c 6464 1.5%1.5%
DW6000cDW6000c 128128 0.75%0.75%
DW7500cDW7500c 128128 0.75%0.75%
DW10000cDW10000c 128128 0.75%0.75%
DW15000cDW15000c 128128 0.75%0.75%
DW30000cDW30000c 128128 0.75%0.75%

资源类的并发最大值Concurrency maximums for resource classes

为了确保每个查询都有足够的资源来有效执行,Synapse SQL 会通过向每个查询分配并发槽位来跟踪资源利用率。To ensure each query has enough resources to execute efficiently, Synapse SQL tracks resource utilization by assigning concurrency slots to each query. 系统根据重要性和并发槽位将查询放入某个队列。The system puts queries into a queue based on importance and concurrency slots. 查询在队列中等待,直到有足够的并发槽位可用。Queries wait in the queue until enough concurrency slots are available. 重要性和并发槽位确定了 CPU 优先级。Importance and concurrency slots determine CPU prioritization. 有关详细信息,请参阅分析工作负荷For more information, see Analyze your workload

静态资源类Static resource classes

下表显示了每个静态资源类的最大并发查询数和并发槽位数。The following table shows the maximum concurrent queries and concurrency slots for each static resource class.

服务级别Service Level 最大并行查询Maximum concurrent queries 可用的并发槽位数Concurrency slots available staticrc10 使用的槽位数Slots used by staticrc10 staticrc20 使用的槽位数Slots used by staticrc20 staticrc30 使用的槽位数Slots used by staticrc30 staticrc40 使用的槽位数Slots used by staticrc40 staticrc50 使用的槽位数Slots used by staticrc50 staticrc60 使用的槽位数Slots used by staticrc60 staticrc70 使用的槽位数Slots used by staticrc70 staticrc80 使用的槽位数Slots used by staticrc80
DW100cDW100c 44 44 11 22 44 44 44 44 44 44
DW200cDW200c 88 88 11 22 44 88 88 88 88 88
DW300cDW300c 1212 1212 11 22 44 88 88 88 88 88
DW400cDW400c 1616 1616 11 22 44 88 1616 1616 1616 1616
DW500cDW500c 2020 2020 11 22 44 88 1616 1616 1616 1616
DW1000cDW1000c 3232 4040 11 22 44 88 1616 3232 3232 3232
DW1500cDW1500c 3232 6060 11 22 44 88 1616 3232 3232 3232
DW2000cDW2000c 4848 8080 11 22 44 88 1616 3232 6464 6464
DW2500cDW2500c 4848 100100 11 22 44 88 1616 3232 6464 6464
DW3000cDW3000c 6464 120120 11 22 44 88 1616 3232 6464 6464
DW5000cDW5000c 6464 200200 11 22 44 88 1616 3232 6464 128128
DW6000cDW6000c 128128 240240 11 22 44 88 1616 3232 6464 128128
DW7500cDW7500c 128128 300300 11 22 44 88 1616 3232 6464 128128
DW10000cDW10000c 128128 400400 11 22 44 88 1616 3232 6464 128128
DW15000cDW15000c 128128 600600 11 22 44 88 1616 3232 6464 128128
DW30000cDW30000c 128128 12001200 11 22 44 88 1616 3232 6464 128128

动态资源类Dynamic resource classes

下表显示了每个动态资源类的最大并发查询数和并发槽位数。The following table shows the maximum concurrent queries and concurrency slots for each dynamic resource class. 动态资源类对所有服务级别的 small-medium-large-xlarge 资源类使用 3-10-22-70 内存百分比分配。Dynamic resource classes use a 3-10-22-70 memory percentage allocation for small-medium-large-xlarge resource classes across all service levels.

服务级别Service Level 最大并行查询Maximum concurrent queries 可用的并发槽位数Concurrency slots available smallrc 使用的槽数Slots used by smallrc mediumrc 使用的槽数Slots used by mediumrc largerc 使用的槽数Slots used by largerc xlargerc 使用的槽数Slots used by xlargerc
DW100cDW100c 44 44 11 11 11 22
DW200cDW200c 88 88 11 11 11 55
DW300cDW300c 1212 1212 11 11 22 88
DW400cDW400c 1616 1616 11 11 33 1111
DW500cDW500c 2020 2020 11 22 44 1414
DW1000cDW1000c 3232 4040 11 44 88 2828
DW1500cDW1500c 3232 6060 11 66 1313 4242
DW2000cDW2000c 3232 8080 22 88 1717 5656
DW2500cDW2500c 3232 100100 33 1010 2222 7070
DW3000cDW3000c 3232 120120 33 1212 2626 8484
DW5000cDW5000c 3232 200200 66 2020 4444 140140
DW6000cDW6000c 3232 240240 77 2424 5252 168168
DW7500cDW7500c 3232 300300 99 3030 6666 210210
DW10000cDW10000c 3232 400400 1212 4040 8888 280280
DW15000cDW15000c 3232 600600 1818 6060 132132 420420
DW30000cDW30000c 3232 12001200 3636 120120 264264 840840

如果没有足够的并发槽位来启动查询执行,查询将根据重要性进行排队和执行。When there are not enough concurrency slots free to start query execution, queries are queued and executed based on importance. 如果重要性相同,查询将以先进先出的方式执行。If there is equivalent importance, queries are executed on a first-in, first-out basis. 如果查询已完成并且查询数和槽位数低于限制,则 Azure SQL 数据仓库会释放排队的查询。As a queries finishes and the number of queries and slots fall below the limits, SQL Data Warehouse releases queued queries.

后续步骤Next steps

若要详细了解如何利用资源类来进一步优化工作负荷,请查看以下文章:To learn more about how to leverage resource classes to optimize your workload further please review the following articles: