为 Azure 数据资源管理器群集选择正确的计算 SKUSelect the correct compute SKU for your Azure Data Explorer cluster

为某个不断变化的工作负荷创建新群集或优化群集时,Azure 数据资源管理器会提供多个虚拟机 (VM) SKU 供你选择。When you create a new cluster or optimize a cluster for a changing workload, Azure Data Explorer offers multiple virtual machine (VM) SKUs to choose from. 这些计算 SKU 经过精选,可为任何工作负载提供最佳性价比。These compute SKUs have been carefully chosen to give you the most optimal cost for any workload.

数据管理群集的大小和 VM SKU 完全由 Azure 数据资源管理器服务进行管理。The size and VM SKU of the data-management cluster are fully managed by the Azure Data Explorer service. 它们由引擎的 VM 大小和引入工作负荷等因素决定。They're determined by such factors as the engine's VM size and the ingestion workload.

随时可通过纵向扩展群集来更改引擎群集的计算 SKU。You can change the compute SKU for the engine cluster at any time by scaling up the cluster. 最好是从适合初始方案的最小 SKU 大小开始。It's best to start with the smallest SKU size that fits the initial scenario. 请注意,使用新的 SKU 重新创建群集时,纵向扩展群集会导致最长 30 分钟的停机。Keep in mind that scaling up the cluster results in a downtime of up to 30 minutes while the cluster is re-created with the new SKU.

提示

计算预留实例 (RI) 适用于 Azure 数据资源管理器群集。Compute Reserved Instances (RI) is applicable to the Azure Data Explorer cluster.

本文介绍各个计算 SKU 选项,并提供技术详细信息来帮助你做出最佳选择。This article describes various compute SKU options and provides the technical details that can help you make the best choice.

选择群集类型Select a cluster type

Azure 数据资源管理器提供两种类型的群集:Azure Data Explorer offers two types of clusters:

  • 生产:生产群集包含用于引擎和数据管理群集的两个节点,根据 Azure 数据资源管理器的 SLA 运行。Production: Production clusters contain two nodes for engine and data-management clusters and are operated under the Azure Data Explorer SLA.

  • 开发/测试(无 SLA) :开发/测试群集为引擎和数据管理群集提供单个节点。Dev/Test (no SLA): Dev/Test clusters have a single node for the engine and data-management cluster. 此群集类型是成本最低的配置,因为它的实例计数较小,且不收取引擎标记费用。This cluster type is the lowest cost configuration because of its low instance count and no engine markup charge. 此群集配置不附带 SLA,因为它不提供冗余。There's no SLA for this cluster configuration, because it lacks redundancy.

计算 SKU 类型Compute SKU types

对于不同类型的工作负载,Azure 数据资源管理器群集支持多种 SKU。Azure Data Explorer cluster supports a variety of SKUs for different types of workloads. 每个 SKU 均提供不同的 SSD 和 CPU 比率,以帮助客户正确地调整其部署规模,并为其企业分析工作负载构建成本最佳的解决方案。Each SKU offers a distinct SSD and CPU ratio to help customers correctly size their deployment and build cost optimal solutions for their enterprise analytical workload.

计算优化Compute optimized

  • 提供较高的核心与缓存比率。Provides a high core to cache ratio.
  • 适用于中小型数据大小的高查询率。Suited for a high rate of queries over small to moderate data sizes.
  • 可实现低延迟 I/O 的本地 SSD。Local SSD for low latency I/O.

计算密集Heavy compute

  • 核心与缓存比率高得多的 AMD SKU。AMD SKUs that offer a much higher core to cache ratio.
  • 可实现低延迟 I/O 的本地 SSD。Local SSD for low latency I/O.

存储优化Storage optimized

  • 适用于每引擎节点介于 1 TB 到 4 TB 的更大存储。Option for larger storage ranging from 1 TB to 4 TB per engine node.
  • 适用于需要存储大量数据,而要求执行更少计算查询的工作负载。Suited for workloads that require storing a large volume of data with less intensive compute query requirements.
  • 某些 SKU 使用附加到引擎节点的高级存储(托管磁盘)而不是本地 SSD 来实现热数据存储。Certain SKUs use premium storage (managed disk) attached to the engine node instead of local SSD for hot data storage.

独立计算Isolated compute

用于运行需要服务器实例级隔离的工作负载的理想 SKU。Ideal SKU for running workloads that require server instance-level isolation.

选择并优化计算 SKUSelect and optimize your compute SKU

在群集创建过程中选择计算 SKUSelect your compute SKU during cluster creation

创建 Azure 数据资源管理器群集时,请根据计划的工作负荷选择最佳的 VM SKU。When you create an Azure Data Explorer cluster, select the optimal VM SKU for the planned workload.

以下属性也可帮助你作出 SKU 选择:The following attributes can also help you make SKU selection:

属性Attribute 详细信息Details
可用性Availability 并非所有 SKU 都可在所有区域中使用Not all SKUs are available in all regions
每核心每 GB 缓存成本Cost per GB cache per core 成本高,计算和密集计算均经过优化。High cost with compute and heavy compute optimized. 成本低,具有优化了存储的 SKULow cost with storage optimized SKUs
预留实例 (RI) 定价Reserved Instances (RI) pricing RI 折扣因区域和 SKU 而异RI discount varies by region and by SKU

备注

对于 Azure 数据资源管理器群集,与存储和网络相比,计算成本是群集成本中最重要的部分。For Azure Data Explorer cluster, compute cost is the most significant part of cluster cost as compared to storage and networking.

优化群集计算 SKUOptimize your cluster compute SKU

若要优化群集计算 SKU,请配置垂直缩放To optimize your cluster compute SKU, configure vertical scaling.

有各种计算 SKU 选项可供选择,因此可根据方案的性能和热缓存要求来优化成本。With various compute SKU options to choose from, you can optimize costs for the performance and hot-cache requirements for your scenario.

  • 如果需要为高查询量实现最佳性能,理想的 SKU 是计算优化版本。If you need the most optimal performance for a high query volume, the ideal SKU should be compute-optimized.
  • 如果需要以较低的查询负载查询大量数据,则存储优化的 SKU 可帮助降低成本,同时仍可提供出色的性能。If you need to query large volumes of data with relatively lower query load, the storage-optimized SKU can help reduce costs and still provide excellent performance.

由于小型 SKU 的每群集实例数受限制,因此最好是使用 RAM 较大的大型 VM。Because the number of instances per cluster for the small SKUs is limited, it's preferable to use larger VMs that have greater RAM. 某些对 RAM 资源需求更高的查询类型(例如使用 joins 的查询)需要更多的 RAM。More RAM is needed for some query types that put more demand on the RAM resource, such as queries that use joins. 因此,在考虑缩放选项时,建议纵向扩展到更大的 SKU,而不要通过添加更多的实例进行横向扩展。That's why, when you're considering scaling options, we recommend that you scale-up to a larger SKU rather than scale-out by adding more instances.

计算 SKU 选项Compute SKU options

下表描述了 Azure 数据资源管理器群集 VM 的技术规格:The technical specifications for the Azure Data Explorer cluster VMs are described in the following table:

名称Name 类别Category SSD 大小SSD size 核心数Cores RAMRAM 高级存储磁盘 (1 TB)Premium storage disks (1 TB) 每个群集的最小实例计数Minimum instance count per cluster 每个群集的最大实例计数Maximum instance count per cluster
Dev(No SLA) Standard_D11_v2Dev(No SLA) Standard_D11_v2 计算优化compute-optimized 75 GB75 GB 11 14 GB14 GB 00 11 11
Standard_D11_v2Standard_D11_v2 计算优化compute-optimized 75 GB75 GB 22 14 GB14 GB 00 22 88
Standard_D12_v2Standard_D12_v2 计算优化compute-optimized 150 GB150 GB 44 28 GB28 GB 00 22 1616
Standard_D13_v2Standard_D13_v2 计算优化compute-optimized 307 GB307 GB 88 56 GB56 GB 00 22 1,0001,000
Standard_D14_v2Standard_D14_v2 计算优化compute-optimized 614 GB614 GB 1616 112 GB112 GB 00 22 1,0001,000
Standard_DS13_v2 + 1 TB PSStandard_DS13_v2 + 1 TB PS 存储优化storage-optimized 1 TB1 TB 88 56 GB56 GB 11 22 1,0001,000
Standard_DS13_v2 + 2 TB PSStandard_DS13_v2 + 2 TB PS 存储优化storage-optimized 2 TB2 TB 88 56 GB56 GB 22 22 1,0001,000
Standard_DS14_v2 + 3 TB PSStandard_DS14_v2 + 3 TB PS 存储优化storage-optimized 3 TB3 TB 1616 112 GB112 GB 22 22 1,0001,000
Standard_DS14_v2 + 4 TB PSStandard_DS14_v2 + 4 TB PS 存储优化storage-optimized 4 TB4 TB 1616 112 GB112 GB 44 22 1,0001,000
  • 可使用 Azure 数据资源管理器 ListSkus API 查看各区域已更新的计算 SKU 的列表。You can view the updated compute SKU list per region by using the Azure Data Explorer ListSkus API.
  • 详细了解各种 SKULearn more about the various SKUs.

后续步骤Next steps