Blog

Logon Duration – What Can Be Learned from 2 Million Logons?

Introduction

How fast is the user logon process in business IT environments? This question is of interest for IT pros because an answer to it can provide a benchmark for user experience.

This article presents statistics from a global dataset of logon duration metrics gathered by the ControlUp Insights platform. ControlUp Insights is a cloud-based IT analytics platform built for IT admins. Alongside traditional IT monitoring and reporting, ControlUp Insights offers access to global performance data for comparison and benchmarking. This article provides the IT community with a first peek into those global statistics.

Logon Duration – What do you mean by that?

Logon duration is a period of time that begins when the user is successfully authenticated. Typically, this happens whenever a user inputs a username and password which are accepted as correct by a domain-joined Windows system.

For a “desktop” scenario, the logon process is considered complete when the Windows Explorer shell is ready. From the user’s perspective, this is the moment when the Start button of the Windows taskbar becomes responsive and opens the Start Menu when clicked.

When the default Windows shell is not used, an alternative application is configured to be launched by the application delivery mechanism in place. For this “published app” scenario, we’ll consider the logon process complete when the published application is started.

To explain this choice of boundaries briefly, let’s just say that the idea was to represent nothing but the machine-caused delay, until the desktop is ready or until the published app takes over. Any further delay experienced from that moment can be attributed to resource bottlenecks associated with applications, and not caused by the logon process itself.

For more details regarding the ways ControlUp measures login duration, see here: http://www.controlup.com/introducing-controlup-4-1/

Why should IT care about logon duration?

Logon duration is an important user experience metric for any end-user computing environment, for several reasons.

First, logon performance may be seen as related to employee productivity. The faster a computing environment is delivered to the end user, the greater the user’s ability to react to work challenges, whatever they may be. The speed of logon upon an employee’s arrival at the office determines how quickly they are able to start generating value for the organization. For logons that happen in the middle of a workday, logon duration may influence the speed with which an employee can switch from one task to another. If logon is slow enough, users may even forget why they logged on in the first place!

Second, Windows logon is a well-structured process which can be analyzed to uncover delays and produce actionable insights. The IT department can (and should) use those insights to improve the quality of service delivered by the organization’s computing environment. While application performance may vary greatly and can be very tricky to measure, the logon process involves a small number of relatively well-known phases.

This also enables IT professionals to easily compare logon performance between different systems, both within the organization and outside it. It’s about knowing how the average login duration in your organization ranks against global stats, while also being able to detect and monitor irregular login performance.

Here’s some more on important metrics for VDI: http://www.controlup.com/vdi-end-user-experience-metrics/

Cut to the chase. So who wins?

In a sample of over 1.8 million logon events from 16 different organizations, the average logon duration was found to be 29.7 seconds (with a standard deviation of 24.4 seconds).  The median was 27.5 seconds, dividing between the faster and slower half, 8 organizations in each.

Looking at the organizational averages might give us a better idea about the results you might find in a given network. For example, in the organization with the slowest average login in the sample users waited for 67.8 seconds, while the champion network had an average login duration of only 9.6 seconds!

Like with everything in IT systems performance, your mileage may vary. So what else is new?

Without going into complicated stats, here’s one source for difference between fast and slow. It just so happens that half of the systems in the sample use predominantly published applications with an average login duration of 18.5 seconds, while the other half utilizes published desktops which become responsive after an average of 40.9 seconds! Overall, 7 out of 8 published app shops ranked in the faster half, with 7 out of 8 full desktop networks in the slower half of the sample.

Logon Speed

When it comes to user experience, consistency is an additional factor to consider. Performance of systems in which logon duration exhibits a lower standard deviation can be seen as more predictable. In this sample, the lowest standard deviation for an organization was 3.5 seconds, and the highest standard deviation was 69.5. Again, 7 out of 8 networks who used published apps as their primary delivery method ended up in the top half of the list when ranked by the standard deviation.

So on average, published app environments tend to have logins that are faster than systems that deliver full desktops. Published apps also show more consistent login duration. This insight is no big news to anyone involved in end-user computing, but it’s still nice to see the numbers.

How fast are my logins? I need to know!!!

Logon duration is a metric that ControlUp displays for every monitored user session, along with a breakdown of the logon process into phases. So for a quick answer, go ControlUp – you should be able to set it all up within an average of 5 slow logins 🙂

The PowerShell freaks among you can help themselves to some resources for investigating logon duration in depth: http://www.controlup.com/script-library/Tags/logon/

Is there more? 

Always. We have some more interesting insights regarding the duration of the individual phases of the logon process (profile load time, group policy, shell load, and other components). We built ControlUp Insights for the purpose of delivering globally acquired knowledge to IT pros, and we intend to deliver, so stay tuned!

Be Sociable, Share!

Comments (4)

  • Nick Rintalan | February 23, 2016 at 2:33 pm Reply

    Hi Eugene – this is great stuff and I was hoping someone would expose this data soon. I get asked this question all the time and I always respond with something like this – “if you’re under 30 seconds, you’re doing well. If you’re over 30 seconds, something is probably not quite right.” But now it’s great to have some data actually back my statement up – but is it true this data is only coming from 16 different companies right now? The # of logon events at 1.8M is impressive, but I think we need data from a lot more companies or orgs. Because what I’ve found in my 12 years of Consulting at Citrix is that the average login time is more like 45-60 seconds, at least in the ENT space where I spend 90% of my time. And I’ve seen everything from 6 seconds to 5 minutes. If everyone was getting 27-29 seconds, I wouldn’t be getting as many phone calls. 😉

    Anyway, let me know if the # of orgs is truly 16 and how big these orgs are if you can. This is great stuff though – congrats to the ControlUp team for including this telemetry in the product and exposing the data to the community.

    -Nick

    Nick Rintalan, Lead Architect, CCS

    • Eugene Kalayev | February 23, 2016 at 4:08 pm

      Hey Nick!
      Thanks a lot for the warm words. Really glad you find the stats interesting.

      As for your questions:
      Yes, this sample included 16 organizations, with sizes ranging from a few hundreds to a few thousands CCU.
      As for the average – I wouldn’t like to state that everyone was getting 27-29 seconds, only that it was the average between organizations. I didn’t dive into more details here for brevity’s sake, but in fact most logons in the sample were faster than that.

      As soon as I find a couple of hours to write it up, I’m going to publish a second round of stats, this time for 100+ organizations. Following your comment, I think this time I’m going to include some more detailed stats on the distribution of logons.

      Thanks again, and stay tuned!

  • Al Boyce | April 5, 2016 at 4:01 pm Reply

    Thanks for the information Eugene!
    There’s some other things I’m trying to figure out in our environment. Sometimes our average logon duration goes up from say 16 seconds to 21 seconds on a particular day and, when I look at sessions, I see a couple hundred people with durations of more than 90 seconds. That skews the average considerably and nearly all of their lag is in the “Logon Duration – Other” category.

    It would be nice if there was an easy way to determine whether these outliers come from a particular part of the network, or are hitting a couple “laggy” Citrix servers. Any ideas on where we could parse that out?

    Thanks again!
    Al

    • Eugene Kalayev | April 6, 2016 at 3:06 pm

      Hey Al,
      As always, great to hear from you!
      First of all, the situation you’re describing is very similar to what we see in most networks – some logons are abnormally long, and more often than not we find that those conditions result from detectable misconfigurations that can be fixed. I will elaborate on that in the second part of this article, which is in the works right now.
      High variance in the “Logon Duration – Other” category is another typical finding. I suggest you look at this post by my colleague Niron Koren for suggestions on dissecting the logon process phases. There’s a script-based action for ControlUp that can help you find the culprit/s.
      Regarding the way to find out systematic patterns in your slow logons, it’s something that offer in ControlUp Insights. There you can easily sort your historical sessions by logon duration and observe the common properties (like the same server or client IP range). For on-premises customers such as yourself, we are working on a great offline Insights platform, to become available later this year. I bet you’ll be among the first ones to hear about the beta version once it becomes available 🙂
      For now, as a workaround, I would suggest creating an export rule that will export the Sessions view to the disk, pick some sample files for times of interest and then sort the files by logon duration. Simple trends like specific laggy servers or remote subnets should be easy to spot. To identify trends in multiple files, perhaps we can hack a PowerShell script to identify common factors from sessions with slow logons. Let us know if you would like to look into this option.

      Best,
      Eugene

Leave a comment

Your email address will not be published.

Do you have more
 questions?
Want a real person to walk you through a live demo?
We’re happy to help out.
Feb 5 – Citrix and ControlUp webinar introducing ControlUp 4.1 and user experience management. Sign up now!
START YOUR TRIAL
Get Your Download Link
Gain access to ControlUp from your PC. Register and get a link to start your Free Trial.