Quantcast
Channel: Category Name
Viewing all articles
Browse latest Browse all 10804

Running Critical Application Workloads on Microsoft Azure D-Series Virtual Machines

$
0
0

On the Azure Customer Advisory Team (AzureCAT), we’ve been testing the performance of one of the latest generations of hardware components now being introduced into our public cloud called the D-Series. And what’s especially cool is how they can help boost performance significantly for critical workload applications compared to the earlier VM series. This is extremely important for solutions based on Microsoft SQL Server as we described in a previous white paper. Our new findings extend those tests.

Customers told us that they wanted a straightforward way to transition their applications from traditional data centers to Microsoft Azure virtual machines (VMs), but performance is key with their critical workloads, and they weren’t always getting it.

The D-Series offers two key features related to performance, neither of which require you to make any particular application changes:

  • Local storage (temporary) based on solid-state drives (SSDs)
  • Higher number of attached data disks (up to 32 for D14 VMs)

In our performance tests, we used these new features to tune applications and saw gains in performance. For example:

  • Placing TempDB file on local SSD storage on a D13 VMgave approximately 4.5 times the throughput of an A7 VM with attached data disks, at a fraction of previous latency for the same SQL Server-generated IO patterns.
  • D14 VMs with 32 attached disks can provide up to 85 percent more write IOPS and bandwidth compared to an A7 VM with 16 attached disks.

We documented four scenarios in which the D-Series made a significant difference for our customers in the white paper, Running Critical Application Workloads on Microsoft Azure Virtual Machine.

It describes:

  • How persistent disk latency can directly impact application response times.
  • How limited throughput from persistent disks can impact application performance when SQL Server tempdb use is significant.
  • How to use SSD-based fast storage in the application tier to speed temporary file processing.
  • How to reduce compile and startup time for a large ASP.NET web application by moving the %temp% folder on a temporary drive in a D-Series VM.

In essence, new D-Series VMs in Azure can help run performance-critical workloads on both the data tier and application tier, offering better performance overall for CPU, storage, and networking, with a price performance ratio that can be favorably compared to other VM series.

Certain application scenarios such as OLTP database servers benefit mainly from local SSD-based temporary storage for extending buffer pools and hosting temporary operations. Application servers benefit from faster and low latency local storage and also from the increased CPU performance provided by this new generation of VMs.

Download the white paper for details and suggestions for improving application performance. And check out our performance expectations article for the D-Series.


Viewing all articles
Browse latest Browse all 10804

Trending Articles