Azure E series (E20S_V3 and E20DS_V4)

Question

We are already on azure IaaS model since last 3 years and currently planning to spin up a VM in production subscription which will host SQL server 2016 enterprise edition and we are comparing between these two series E20ds_v4 and E20S_V3 with in E series.The only difference I see is the temp storage & it’s throughput and a $79 price difference/mo and rest all specs are same.Can some please share your thoughts on what is the major difference between "E20ds_v4" and "E20S_V3" VM's ? what does "ds" stands for ? for a production scale oltp what would be a better choice ?

Answer 1

Es, Eas, Ds and Das Series offers the optimum memory to vCPU ratio required for OLTP workload performance as you can read here. DSv_3 and Es_v3-series are hosted on general purpose hardware with Intel Haswell or Broadwell processors.

Use VM sizes with 4 or more vCPU like E4S_v3 or higher, or DS12_v2 or higher as a best practice for SQL Server VMs on Azure.

M Series offers the highest memory to vCPU ratio required for mission critical performance and is ideal for data warehouse workloads. M-series offers the highest vCPU count and memory for the largest SQL Server workloads and hosted on memory optimized hardware with Skylake processor family.

Use HammerDB to measure performance and scalability of each SQL VM option.

Use premium SSDs for the best price/performance advantages. Configure ReadOnly cache for data files and no cache for the log file. Use Ultra Disks if less than 1 ms storage latencies are required by the workload. Meanwhile, premium file shares are recommended as shared storage for a SQL Server failover cluster instance. Premium file shares do not support caching, and offer limited performance compared to premium SSD disks. Standard storage is only recommended for development and test purposes or for backup files and should not be used for production workloads. Use a minimum of 2 premium SSD disks (1 for log file and 1 for data files).Enable read only caching on the disk(s) hosting the data files. Stripe multiple Azure data disks to get increased storage throughput. Place TempDB on the local SSD D:\ drive for mission critical SQL Server workloads