Relating the Dynatrace Browser Agent Metrics to UEM metrics
Relating the Dynatrace Browser Agent Metrics to UEM metrics
For Eg Page Action - Loading of Page /c/xyz.. we have metrics in UEM in Page Action Pure Paths like Response Time etc. But it looks like it is total load.
How can i Get First Impression, Onload , Total Load times split in UEM , so we can create trending graphs on all these metrics
Hi,
At the moment you do not have First Impression, Onload or Total Load Time in UEM like you get it for the browser agent. The First Impression will not be available for UEM as it depends on the rendering events of the browser which can only be captured by a native hooking and not via JavaScript.
I will take this request and discuss it with PM.
Thanks Klaus
how do you folks measure "perceived render time".. wouldnt that need native hooking too?
Is there an exhaustive comparison between Browser Agent metrics and UEM metrics? Ie, what is distinct, and what is overlapping? That will help us better decide when to use what..
Check the doc - it contains a detailed explanation: Perceived Render Time
Hi Andi,
Is there an exhaustive comparison between Browser Agent metrics and UEM metrics? Ie, what is distinct, and what is overlapping? That will help us better decide when to use what..
I think that the decision should be based on the Use Case.
a) User Experience Management for ALL of your Production Users -> UEM
b) Deep Dive Diagnostics in Test and Dev or to troubleshoot a particular problem that you can verify on IE or FF -> Browser Agent
the two solutions have a very different approach to capturing data. UEM is lightweight using a JavaScript Agent Approach which is great for large scale production environments where you want to monitor millions of visits and page actions. This is great for production monitoring as well as fault domain isolation, e.g: do we have a problem in our CDN (using the 3rd party content detection feature), do we have regional problems (using the IP Lookup and mapping to GeoLocation), are there any JavaScript errors (we capture JS errors that impact the end user), are there any performance related issues on pages and how does this all impact User Experience and Conversions.
The Browser Agent is more heavy weight and requires a Browser Plugin to capture very detailed JavaScript, DOM Executions and Rendering Activities. This is great for Developers, Test Engineers, Test Automation as they can analyze down to the JS Method Level what is going on.
So - I am not sure if it makes sense to compare the individual metrics that these two approaches provide. I think you need to decide what your use case is and then decided whether you go with UEM or Browser Agent
Hope this helps
Thanks for the detailed answer.. that does help, and validates my understanding of the differences between the two.
Our use case is simple.. our performance lab wants to get some high level metrics on client-side page performance. We would also like to standardize around one set of high level metrics.. right now UEM seems preferable as it will be easy for anyone in our team to get this data.. not have the headache of configuring the browser plugin.. and it can provide a direct comparison to Production UEM data.
The basic high level metrics we are interested in are:
- server-side time
- Firstimpression (its not very clear what the boundary for this measure is.. what on the page qualifies as FirstImpression)
- Peceived Render Time (im hoping its equivalent to, or more accurate than FirstImpression).. looks like i'll have to manually chart this measure to usefully report and trend on this value for different test/build comparisons
- Document Interactive (from the w3c navigation timer).. how does this differ from the above?
- Total Page Load (for everything to get loaded, including AJAX calls).
We want to trend / compare / average / min / max the above values..
Makes perfect sense what you try to achieve. UEM and the KPIs that you list will work great.
My question now is: do you analyze the UEM data from your Production Environment? or do you use UEM also in testing and use these numbers?
If you are focusing on a testing environment you could still go with the Browser Agent and automate your test scenarios with tools such as Selenium.
But - if you are happy with the UEM data and the easy accessible way I would go with that.
To answer your remaining questions
a) FirstImpression: that is captured by our Browser Agent. We hook inot the native rendering engine of IE and FF and with that get to know when the browser draws something to the canvas. So - the First Impresesion is the first time the browser actively renders something to the screen. Which doesnt mean that the page is completely rendered - but - that the user got the first visual indication that something is happening
b) Perceived Render time: definitely more accurate than FirstImpression if you really want to know how long it takes until the user sees the page fully rendered
c) doc Interactive: this is a metric as defined by the W3C Navigation Time. The browser tells us when the document is ready to be used by the end user. More details on this metric can be found on the official W3C pages
andi
So Andi.. if I go with UEM, and have it measure:
1) server-side time
2) perceived render time
3) doc interactive : same as "dom interactive"? http://www.w3.org/TR/html5/syntax.html#the-end
I assume this is almost equivalent to OnLoad.. and longer than "perceived render time"..
4) total load time : equivalent to Page Action dashlet's "duration" column?
If the above assumptions are correct, how do you generate a performance report that is able to avg / min / max / trend the above.. for multiple iterations of the same page request..
Discovered Andi's webinar on ShowSlow.. http://www.youtube.com/watch?v=Yqj2v16Ly4M .. this is exactly what we need, ie store / average / trend / graph / report our test results.
Are there any good instructions on how ShowSlow and the Dynatrace Browser Plugin can integrate? Or is this overkill.. ie the Dynatrace Collector can do a better job than ShowSlow?
I'm hoping for a solution that will aggregate / trend / report on all of our collected results.. and then let us drill-down (by clicking a plot point on the graph) into Dynatrace for specifics into the client-side browser purepath, and server-side purepath.
The Integration with ShowSlow and AJAX Edition works well and should be well documented on the community.
However - our Browser Agent (Commercial Offering) integrates great into our Test Automation Feature where we basically do all the trending for you and also alert on any regressions.
For more information on this check out dynaTrace in Continuous Integration - The Big Picture and Automated Performance Analysis in Continuous Integration
May be i should take the onLoad handler executions of page method's Elapsed time for Onload Time and 200 web request execution total as the First Impression
