Measuring performance impact of BizTalk Feature Pack 1 analytics

Posted by Marius Storsten

Find me on:

Microsoft just recently launced their Feature Pack 1 for BizTalk 2016. They claim that enabling the analytics part (view operational data in Power BI) does not hurt or impact performance in any way. So I created a very simple BizTalk test to check this out.....

First, a bit about the test-setup (very simple as mentioned):

  1. BizTalk 2016 with Feature Pack 1 installed
  2. Single server setup with SQL server on the same box
  3. Single processing host (of course I should have had a dedicated tracking host, but I skipped this to compare directly on the host)
  4. Passthrough messaging pattern with a single receiveport and a sendport subscribing to the receiveport
  5. Analytics enabled under Group settings
  6. Analytics enabled for receiveport and sendport

For testing purposes, a simple script generated 2 messages per second to the receiveport. The message size was 1 KB.

To measure if there was any differences in CPU during the test, I simply used Perfmon to chart the BizTalk service.

Test 1 - 2 messages per second without analytics enabled

without-featurepack1-analytics-1.jpg

 

Test 2 - 2 messages per second with analytics enabled

with-featurepack1-analytics-1.jpg

 

Summary

In short, I was quite amazed by the numbers in this very simple test. This test is even using passthrough and just two ports. The performance impact is huge! 

As mentioned I should have used a dedicated tracking host for this. However, the total impact on the CPU would be the same.

Not sure how much MS is tracking by enabling analytics but the numbers should give you a good indication. On the bright sight you can access a lot of data in Power BI, but based on these numbers I would not recommend to have this as an "always-on" feature but more as a "on-demand" feature when you need additional insight.

 

Tags: TECH

Subscribe by Email

Most Popular Posts