Update on RavenDB at NBC Universal
I just got an email from NBC Universal about the RavenDB usage. We’ve previously talked with them about their RavenDB usage, and I was very pleased to hear that they have completed their upgrade to RavenDB 2.x.
I was even happier to hear the results of the upgrade:
- Index times decreased (75% improvement)
- Replication + Indexing times decreased (40% improvement)
- Development server raven restore times decreased (75% improvement)
- Memory leak/crash frequency decreased (1) (60% improvement – stronger devices are experiencing no crashes)
- Support from Hibernating Rhinos will now improve as they are more capable of supporting the new version
I like getting such news from customers.
- The scenario they are talking about is overloading the system. It exhibits itself as a high memory utilization. On small devices, that can lead to crashes due to out of memory errors.
Comments
Memory leak/crash frequency decreased (60% improvement – stronger devices are experiencing no crashes)
I don't understand this. So RavenDB regularly crashes and has memory leaks!?
Jiggaboo, The scenario they are talking about is overloading the system. It exhibits itself as a high memory utilization. On small devices, that can lead to crashes due to out of memory errors.
Ayende, I had the same reaction Jiggaboo did with regards to RavenDb crashing. It may be a good idea to clarify that in the article.
Good point, done.
Still not sure I understand. I don't see why increasing load (even overloading) can lead to memory leaks and this being a normal thing.
Jiggaboo, Overload on the system can be expressed as using too much memory. This is what they called memory leak, even though we are actually using that memory.
Comment preview