I have a site that I recently moved to Azure Websites with a SQL Azure server (first time I've used Azure)for hosting and since moving it there I'm seeing some serious performance issues. The site was hosted on a dev server that we have with Rackspace and performance was just fine there. But on Azure load times seem to have doubled to tripled on some pages. The pages have that are loading slow have a lot of relational data that I have set up using relations.
I've tried upgrading both the server and the database and neither seems to have any effect on the load times. The server stats also don't seem to be spiked out. RAM for website stays around 32% and spikes once to 42% while the largest processor spike is 2%. The SQL Azure server spikes at about 22% DTUs. I also have the website set be always on to keep it warm.
I'm at a loss here as to what would be causing such a drastic change in performance. I feel like I've heard nothing but good from the Umbraco community about Azure so I'm starting to get pretty concerned over going this route if it turns out that they simply don't have the compute power to handle the site.
Any suggestions that anyone could give to speed this up some is greatly appreciated and I'm happy to supply any additional information.
We opened a Microsoft call and they suggested that the website/umbraco application is throwing millions of exceptions (that don't appear in the logs) which is causing the garbage collector to get backed up.
That may be true but the server stats remain under 50% even as CPU usage spikes.
I'm this close to moving the website to an on premise server but I would like to avoid that as I have 14 domains that would need to be reconfigured.
I wish I could tell you that I fixed this, but unfortunately I didn't. We enabled caching on all the pages that were running slow, which turns out to be nearly the entire site. If you are lucky enough to be able to get a page loaded from cache then the site feels pretty snappy, otherwise it's pretty painful.
This has been pretty frustrating for me, I really like this site, but I absolutely hate the performance.
If you find out anything I would love to hear about it.
I know it's been a while since this thread was updated, but I've had performance problems on an Azure hosted Umbraco site lately and I wanted to share the outcome.
For us it was slightly different to what you're describing in that the delays happened when the Lucene indexes were updated. It would cause the site to actually restart each time content was published, resulting in big delays. The reason for this is that Azure web roles use a shared disk for the webroot of the site so that each instance shares the same file system. This doesn't play well with Lucene indexes.
Our fix so far has been to update the storage location for the indexes (as per the suggestions on this fix: http://issues.umbraco.org/issue/U4-5993). So far this seems to have done the trick.
The reason I bring this up is I'm wondering if your pages are built to use a lot of index queries? If they are, perhaps the issue is the slow shared-disk that Azure is using for your web instances? Storing the index in temp storage might help with that?
Site slow on Azure Websites
I have a site that I recently moved to Azure Websites with a SQL Azure server (first time I've used Azure)for hosting and since moving it there I'm seeing some serious performance issues. The site was hosted on a dev server that we have with Rackspace and performance was just fine there. But on Azure load times seem to have doubled to tripled on some pages. The pages have that are loading slow have a lot of relational data that I have set up using relations.
For example of load times:
http://tipawards.azurewebsites.net/the-heroes/ - 5384ms on Azure
http://tipheroes.openfieldcreative.com/the-heroes/ - 2548ms on Rackspace
and even worse:
http://tipawards.azurewebsites.net/resources/ - 13872ms on Azure
http://tipheroes.openfieldcreative.com/resources/ - 3193ms on Rackspace
I've tried upgrading both the server and the database and neither seems to have any effect on the load times. The server stats also don't seem to be spiked out. RAM for website stays around 32% and spikes once to 42% while the largest processor spike is 2%. The SQL Azure server spikes at about 22% DTUs. I also have the website set be always on to keep it warm.
I'm at a loss here as to what would be causing such a drastic change in performance. I feel like I've heard nothing but good from the Umbraco community about Azure so I'm starting to get pretty concerned over going this route if it turns out that they simply don't have the compute power to handle the site.
Any suggestions that anyone could give to speed this up some is greatly appreciated and I'm happy to supply any additional information.
Thanks,
Owen
Did you ever find a solution Owen?
We are having an almost identical issue.
We opened a Microsoft call and they suggested that the website/umbraco application is throwing millions of exceptions (that don't appear in the logs) which is causing the garbage collector to get backed up.
That may be true but the server stats remain under 50% even as CPU usage spikes.
I'm this close to moving the website to an on premise server but I would like to avoid that as I have 14 domains that would need to be reconfigured.
I wish I could tell you that I fixed this, but unfortunately I didn't. We enabled caching on all the pages that were running slow, which turns out to be nearly the entire site. If you are lucky enough to be able to get a page loaded from cache then the site feels pretty snappy, otherwise it's pretty painful.
This has been pretty frustrating for me, I really like this site, but I absolutely hate the performance.
If you find out anything I would love to hear about it.
Hi there,
I know it's been a while since this thread was updated, but I've had performance problems on an Azure hosted Umbraco site lately and I wanted to share the outcome.
For us it was slightly different to what you're describing in that the delays happened when the Lucene indexes were updated. It would cause the site to actually restart each time content was published, resulting in big delays. The reason for this is that Azure web roles use a shared disk for the webroot of the site so that each instance shares the same file system. This doesn't play well with Lucene indexes.
Our fix so far has been to update the storage location for the indexes (as per the suggestions on this fix: http://issues.umbraco.org/issue/U4-5993). So far this seems to have done the trick.
The reason I bring this up is I'm wondering if your pages are built to use a lot of index queries? If they are, perhaps the issue is the slow shared-disk that Azure is using for your web instances? Storing the index in temp storage might help with that?
Cheers, Mark
is working on a reply...