Tessell for SQL Server Benchmark

At Tessell, we are committed to delivering the highest-performing database services in the market. To achieve this goal, we explored Azure Managed Instance for SQL Server to assess its throughput. The most accurate way to gauge database performance is by conducting performance benchmarks for databases operating in a cloud environment with an industry benchmark tool.

Azure Managed Instance for SQL Server is a scalable cloud database service that is always running on the latest SQL Server Patch and fully managed by Microsoft.

Tessell provides a fully Managed DBaaS service @ your terms, allowing you complete control over patch set and versioning.

In this article, we conduct performance benchmarks to assess, analyze, and compare the performance of SQL Server database engines running on Azure - Tessell and Azure Managed Instance for SQL Server. We utilized the HammerDB benchmarking tool for the benchmarking process to make a repeatable, verifiable comparison.

Tessell for SQL Server is offered in the following variations:

High-Performance Compute (HPC): Run your databases on high-performance cloud infrastructure with locally attached NVMe storage arranged in RAID 0 fashion to achieve millions of IOPS. Take cloud-native snapshots the same way you do for standard cloud storage. Stop/start the service at any time without any loss of data. Mirrored disks, also known as disk mirroring or RAID 1, are a storage configuration in which two identical disk drives are used in parallel to store the same data. This setup provides data redundancy and fault tolerance. When one disk fails, the other continues to operate seamlessly, ensuring data availability. In the context of SQL Server in an HPC environment, this redundancy is crucial to maintain continuous data access and minimize downtime.

Standard Compute: Standard Azure VM shapes for your standard workloads. Enjoy fully managed PaaS service @ your terms for versioning, etc. Stop/start the service on demand for utility consumption.

For this benchmarking exercise, we will use the Tessell HPC shapes.

Set up Azure Managed Instance

We are creating a SQL Managed Instance with the latest 5th Gen ( 8 vCore ) CPU. We will run a heavy load on the system and load it with ~120-130 GB of data.  

The instance takes a whopping ~6 hours to get created.

We have also created the jump server in the same network to launch the Hammer DB test.

Once provisioned, Hammer DB was configured with the below parameters. The same Jump Box will also be utilized to test the SQL Server on Azure Tessell.

Following is the configuration used for the Benchmarking Test

{

    "connection": {
        "mssqls_server": "mssqlmi001.2f347019b10a.database.windows.net",
        "mssqls_linux_server": "localhost",
        "mssqls_tcp": "true",
        "mssqls_port": "1433",
        "mssqls_azure": "true",
        "mssqls_authentication": "sql",
        "mssqls_linux_authent": "sql",
        "mssqls_odbc_driver": "ODBC Driver 18 for SQL Server",
        "mssqls_linux_odbc": "ODBC Driver 18 for SQL Server",
        "mssqls_uid": "hpctest",
        "mssqls_pass": "Tessell@0123456789",
        "mssqls_encrypt_connection": "true",
        "mssqls_trust_server_cert": "true"
    },

    "tpcc": {

        "mssqls_count_ware": "1000",
        "mssqls_num_vu": "16",
        "mssqls_dbase": "tpcc",
        "mssqls_imdb": "false",
        "mssqls_bucket": "1",
        "mssqls_durability": "SCHEMA_AND_DATA",
        "mssqls_total_iterations": "10000000",
        "mssqls_raiseerror": "false",
        "mssqls_keyandthink": "false",
        "mssqls_checkpoint": "false",
        "mssqls_driver": "timed",
        "mssqls_rampup": "2",
        "mssqls_duration": "5",
        "mssqls_allwarehouse": "false",
        "mssqls_timeprofile": "false",
        "mssqls_async_scale": "false",
        "mssqls_async_client": "10",
        "mssqls_async_verbose": "false",
        "mssqls_async_delay": "1000",
        "mssqls_connect_pool": "false"
    }
}

Now run the Test for calculating the throughput for Azure MI using Hammer DB

The total NOPM  (new orders per minute) was achieved in 2183 and 5185 TPM.

Now, let's Provision an HPC shape on Tessell with an 8 vCPU machine (NVMe Storage ) arranged in the RAID 0 fashion.

As mentioned above, a similar Test was conducted, and the following output was observed.

‍

Conclusion:

As you can see from the results, Tessell HPC SQL Server gave 20X higher IOPS as compared to Azure Managed Instance  for the identical Hammer DB workload

As is generally the case, you can benefit significantly in license costs if your application is limited by IOPS.

With Tessell HPC, there is no IOPS metering/throttling, so you always get price predictability alongside massive performance uplift.

Conclusion: Your demanding workloads can significantly benefit from Tessell HPC database service at reduced TCO

To get started, you can sign up for a free trial here.