Practical Limit to Number of Concurrent Editors in Umbraco 9 Back Office?
I've seen this question asked for older versions of Umbraco with very hand wavy answers. I'm looking for something a little more concrete.
Say I have an AWS EC2 instance that is pretty beefy (64GB RAM, 16 cores, very fast SSD, SQL Server, etc.). Something like an M4.4xlarge (https://aws.amazon.com/ec2/instance-types/).
Say I have hundreds or thousands of editors in the back office at the same time. Not editing the same content, just editing at the same time.
How many editors before this instance starts to slow down (e.g., publish operations take longer than 5 seconds)? For this example, let's say each editor publishes every 30 seconds.
Bonus points if you're actually aware of a site where such a scenario was supported (perhaps on Umbraco 8 since Umbraco 9 is still pretty new), or if you've built some sort of tool to test this scenario.
I imagine Lucene would be the biggest bottleneck, but I think that can be limited (e.g., configured to not index on every publish). Maybe the database queries would be the second biggest bottleneck. Perhaps the cache construction might be the third biggest bottleneck.
To get the very obvious answers out of the way, I would not find these answers particularly useful:
Ask HQ Reach out to Umbraco HQ, they might have some info (we'll also try that).
Dozens of Users Seemed to hand a few dozen users fine (I'm sure that's true, I'm looking for a reasonable upper limit).
Old Versions I tried Umbraco 6 and it handled 3,000 at once (bit dated to be of use now).
Another Forum Post This other forum post vaguely hints at some sort of a limit (I'm looking for concrete answers, and I've done some searching in the forums already).
2K Limit There used to be a bug that prevented more than 2,000 users from being created at once (I'm aware of that and assume it's no longer an issue, and the question is more about users who are all logged in and working at once).
.Net 5 Speed Umbraco 9 is much faster now that it's on .Net 5 (goes without saying, but I'm looking for a more practical answer).
The boring answer, which you probably already guessed:
It depends on what you're doing, your mileage may vary.
I know it's not in your expected answers and you won't like it very much. :)
The truth is: we haven't tried, we expect it to be loads better than v8, but it's extremely hard to create real-life tests. So, as before, we're going to rely on reports of specific areas of Umbraco limiting performance, please do let us know and we can have a look.
And yes, at some point we'd love to support load balancing the backoffice, currently we don't have the infrastructure for it and we have no timeline for when we can get that done.
At Moriyama, we are currently looking into the performance of Umbraco v9 compared to v8. We are also comparing v9 on Windows vs Linux. At the moment we have focused on the front end of the site but we will also be doing some testing of the backoffice. Your scenarios would make a great test case. I'll see what we can do to get near to your test criteria and let you know the results.
Hi Paul. Did you happen to have a sense of when you might have some performance test results? Still very interested to see them if you end up running them!
Practical Limit to Number of Concurrent Editors in Umbraco 9 Back Office?
I've seen this question asked for older versions of Umbraco with very hand wavy answers. I'm looking for something a little more concrete.
Say I have an AWS EC2 instance that is pretty beefy (64GB RAM, 16 cores, very fast SSD, SQL Server, etc.). Something like an M4.4xlarge (https://aws.amazon.com/ec2/instance-types/).
Say I have hundreds or thousands of editors in the back office at the same time. Not editing the same content, just editing at the same time.
How many editors before this instance starts to slow down (e.g., publish operations take longer than 5 seconds)? For this example, let's say each editor publishes every 30 seconds.
Bonus points if you're actually aware of a site where such a scenario was supported (perhaps on Umbraco 8 since Umbraco 9 is still pretty new), or if you've built some sort of tool to test this scenario.
I imagine Lucene would be the biggest bottleneck, but I think that can be limited (e.g., configured to not index on every publish). Maybe the database queries would be the second biggest bottleneck. Perhaps the cache construction might be the third biggest bottleneck.
To get the very obvious answers out of the way, I would not find these answers particularly useful:
The boring answer, which you probably already guessed:
It depends on what you're doing, your mileage may vary.
I know it's not in your expected answers and you won't like it very much. :)
The truth is: we haven't tried, we expect it to be loads better than v8, but it's extremely hard to create real-life tests. So, as before, we're going to rely on reports of specific areas of Umbraco limiting performance, please do let us know and we can have a look.
And yes, at some point we'd love to support load balancing the backoffice, currently we don't have the infrastructure for it and we have no timeline for when we can get that done.
At Moriyama, we are currently looking into the performance of Umbraco v9 compared to v8. We are also comparing v9 on Windows vs Linux. At the moment we have focused on the front end of the site but we will also be doing some testing of the backoffice. Your scenarios would make a great test case. I'll see what we can do to get near to your test criteria and let you know the results.
Thanks, Paul! If you do happen to add that test case, I would be VERY interested in the results!
Hi Paul. Did you happen to have a sense of when you might have some performance test results? Still very interested to see them if you end up running them!
is working on a reply...