Performance benchmark of Azure Arc-enabled SQL Managed Instance

Microsoft Azure SQL Managed Instance, a member of the Azure SQL family, is an intelligent and scalable cloud database service that combines the broadest Microsoft SQL Server engine compatibility with all the benefits of a fully managed and evergreen platform-as-a-service (PaaS) solution. With Azure Arc, Microsoft brings the Azure innovation and cloud benefits of Azure data services to any environment, including on-premises, edge, and multicloud environments, through Azure Arc–enabled SQL Managed Instance.

SQL Server has been the proven leader and the fastest database for online transactional processing (OLTP) and data warehouse (DW) workloads on Windows and Linux for years. Together with our hardware partners, SQL Server has continued to innovate with high-performing, enterprise-ready solutions that deliver unparalleled price performance.

With the general availability of Azure Arc-enabled SQL Managed Instance Business Critical, we now see enterprise customers and partners modernizing their SQL Server database engine workloads from hardware based and virtualized environments to the evergreen version of Azure SQL MI running on Kubernetes clusters in their own infrastructure. With a near 100% compatibility with the latest SQL Server database engine, customers are able to lift and shift their workloads to a cloud-native Kubernetes platform with minimal application and database changes while maintaining data sovereignty. Azure Arc-enabled SQL Managed Instance also includes built-in management capabilities that drastically reduce management overhead, simplifies deployment of a business critical high-availability configuration with contained availability groups, and has built-in intelligent query optimization features that improve the performance of existing workloads with minimal implementation effort to adopt.

Mission critical workloads require high performance to maximize throughput and minimize response times. Therefore, Dell Technologies and Microsoft partnered to test the performance and scale of Azure Arc-enabled SQL Managed Instance running with Dell PowerStore, Dell PowerEdge and Intel 3^rd generation Xeon Scalable processors.

Performance test of Azure Arc-enabled SQL MI

In the first test we wanted to compare how Azure Arc-enabled SQL MI General Purpose running on an upstream version of Kubernetes (1.23.5) on Red Hat Enterprise Linux 8.5 performs relative to SQL Server 2019 running on Windows virtual machine on VMware ESXi. We used the MSOLTPE¹ workload and the Microsoft BenchCraft transaction driver to compare the performance between these two different environments. The results of this test showed nearly identical performance between the two.

In the second test we compared SQL Server 2019 performance against Azure Arc-enabled SQL MI Business Critical which leverages contained availability groups for database level high-availability and disaster recovery, and the same performance and scalability features we have in SQL Server Enterprise edition. The environment (hardware, storage, OS/Kubernetes version) were the same that we used in the first test. Based on our test results, Azure Arc-enabled SQL Managed Instance Business Critical runs 15% faster compared to SQL Server 2019

Scalability test

We also wanted to test the scalability of Azure Arc-enabled SQL MI by running multiple managed instances. We identified two scenarios:

1. Running an increasing workload across twelve Azure Arc-enabled SQL MIs to demonstrate an increasing application load (scale-up user workload)
2. Running a fixed workload and ramping up the number of running Azure Arc-enabled SQL MI instances from 1 to 12, and then back down to 1. This is a way to demonstrate what happens when additional workloads or applications are added or when running a larger distributed or sharded data application in a scale-out workload style.

In the scale-up user workload test we used the HammerDB TPROC-C workload on 12 Azure Arc-enabled SQL MI GP instances simultaneously. The workload ramped up the number of users over time, increasing the workload in a controlled fashion to demonstrate business growth. All the 12 SQL Managed Instances in the environment proved to be behaving in a balanced way, consuming roughly the same amount of compute resources of the Kubernetes platform.

In the scale-out scenario, a fixed workload was deployed using the same HammerDB TPROC-C test to start or spin-up the same user workload with 15 users every 10 minutes on a new Azure Arc-enabled SQL MI GP instance, up to 12 instances. Each workload executed for 150 minutes and completed or “spun-down”. In the Grafana based monitoring dashboard, which is automatically deployed as part of the Azure Arc data controller, we were easily able to see the scalability of the environment. Each added instance consumed roughly the same amount of compute resources, and started and stopped in a non-disruptive fashion.

These tests demonstrate that Arc-enabled SQL MI, combined with Dell PowerStore and PowerEdge, deliver Azure innovation and cloud benefits with the performance you need for your mission critical workloads.

Read the full white paper here. 

¹ The MSOLTPE workload is derived from the TPC-E Benchmark and is not comparable to published
TPC-E results, as this implementation does not comply with all requirements of the TPC-E Benchmark.

--

Marko Hotti

Sr. Technical Product Manager

Jamie Reding

Sr. Program Manager