Merge pull request #105728 from dimitri-furman/dimitri-furman

Added hardware gen clarification for DTU

Author: PRMerger7

articles/sql-database/sql-database-purchase-models.md

@@ -137,6 +137,20 @@ The input values for this formula can be obtained from [sys.dm_db_resource_stats
 
 Pools are well-suited for databases with a low resource-utilization average and relatively infrequent utilization spikes. For more information, see [When should you consider a SQL Database elastic pool?](sql-database-elastic-pool.md).
 
+### Hardware generations in the DTU-based purchasing model
+
+In the DTU-based purchasing model, customers cannot choose the hardware generation used for their databases. While a given database usually stays on a specific hardware generation for a long time (commonly for multiple months), there are certain events that can cause a database to be moved to another hardware generation.
+
+For example, a database can be moved to a different hardware generation if it is scaled up or down to a different service objective, or if the current infrastructure in a data center is approaching its capacity limits, or if the currently used hardware is being decommissioned due to its end of life.
+
+If a database is moved to different hardware, workload performance can change. The DTU model guarantees that the throughput and response time of the [DTU benchmark](https://docs.microsoft.com/azure/sql-database/sql-database-service-tiers-dtu#dtu-benchmark) workload will remain substantially identical as the database moves to a different hardware generation, as long as its service objective (the number of DTUs) stays the same. 
+
+However, across the wide spectrum of customer workloads running in Azure SQL Database, the impact of using different hardware for the same service objective can be more pronounced. Different workloads will benefit from different hardware configuration and features. Therefore, for workloads other than the DTU benchmark, it is possible to see performance differences if the database moves from one hardware generation to another.
+
+For example, an application that is sensitive to network latency can see better performance on Gen5 hardware vs. Gen4 due to the use of Accelerated Networking in Gen5, but an application using intensive read IO can see better performance on Gen4 hardware vs. Gen5 due to higher memory per core ratio on Gen4.
+
+Customers with workloads that are sensitive to hardware changes, or customers who wish to control the choice of hardware generation for their database can use the [vCore](https://docs.microsoft.com/azure/sql-database/sql-database-service-tiers-vcore) model to choose their preferred hardware generation during database creation and scaling. In the vCore model, resource limits of each service objective on each hardware generation are documented, for both [single databases](https://docs.microsoft.com/azure/sql-database/sql-database-vcore-resource-limits-single-databases) and [elastic pools](https://docs.microsoft.com/azure/sql-database/sql-database-vcore-resource-limits-elastic-pools). For more information about hardware generations in the vCore model, see [Hardware Generations](https://docs.microsoft.com/azure/sql-database/sql-database-service-tiers-vcore#hardware-generations).
+
 ## Frequently asked questions (FAQs)
 
 ### Do I need to take my application offline to convert from a DTU-based service tier to a vCore-based service tier?

articles/sql-database/sql-database-service-tiers-vcore.md

@@ -92,10 +92,11 @@ To enable M-series hardware for a subscription and region, a support request mus
 |Hardware generation  |Compute  |Memory  |
 |:---------|:---------|:---------|
 |Gen4     |- Intel E5-2673 v3 (Haswell) 2.4 GHz processors<br>- Provision up to 24 vCores (1 vCore = 1 physical core)  |- 7 GB per vCore<br>- Provision up to 168 GB|
-|Gen5     |**Provisioned compute**<br>- Intel E5-2673 v4 (Broadwell) 2.3-GHz and Intel SP-8160 (Skylake) processors<br>- Provision up to 80 vCores (1 vCore = 1 hyper-thread)<br><br>**Serverless compute**<br>- Intel E5-2673 v4 (Broadwell) 2.3-GHz and Intel SP-8160 (Skylake) processors<br>- Auto-scale up to 16 vCores (1 vCore = 1 hyper-thread)|**Provisioned compute**<br>- 5.1 GB per vCore<br>- Provision up to 408 GB<br><br>**Serverless compute**<br>- Auto-scale up to 24 GB per vCore<br>- Auto-scale up to 48 GB max|
+|Gen5     |**Provisioned compute**<br>- Intel E5-2673 v4 (Broadwell) 2.3-GHz and Intel SP-8160 (Skylake)* processors<br>- Provision up to 80 vCores (1 vCore = 1 hyper-thread)<br><br>**Serverless compute**<br>- Intel E5-2673 v4 (Broadwell) 2.3-GHz and Intel SP-8160 (Skylake)* processors<br>- Auto-scale up to 16 vCores (1 vCore = 1 hyper-thread)|**Provisioned compute**<br>- 5.1 GB per vCore<br>- Provision up to 408 GB<br><br>**Serverless compute**<br>- Auto-scale up to 24 GB per vCore<br>- Auto-scale up to 48 GB max|
 |Fsv2-series     |- Intel Xeon Platinum 8168 (SkyLake) processors<br>- Featuring a sustained all core turbo clock speed of 3.4 GHz and a maximum single core turbo clock speed of 3.7 GHz.<br>- Provision 72 vCores (1 vCore = 1 hyper-thread)|- 1.9 GB per vCore<br>- Provision 136 GB|
 |M-series     |- Intel Xeon E7-8890 v3 2.5 GHz processors<br>- Provision 128 vCores (1 vCore = 1 hyper-thread)|- 29 GB per vCore<br>- Provision 3.7 TB|
 
+\* In the [sys.dm_user_db_resource_governance](https://docs.microsoft.com/sql/relational-databases/system-dynamic-management-views/sys-dm-user-db-resource-governor-azure-sql-database) dynamic management view, hardware generation for Gen5 databases using Intel SP-8160 (Skylake) processors appears as Gen6. Resource limits for all Gen5 databases are the same regardless of processor type (Broadwell or Skylake).
 
 For more information on resource limits, see [Resource limits for single databases (vCore)](sql-database-vcore-resource-limits-single-databases.md), or [Resource limits for elastic pools (vCore)](sql-database-vcore-resource-limits-elastic-pools.md).