Github user nickva commented on the pull request:

    
https://github.com/apache/couchdb-couch-replicator/pull/37#issuecomment-217233800
  
    @janl @kxepal 
    
    Here is a poor man's cluster graph. From local dev cluster, single source 
to 500 targets replications. Data on ramdisk. The time window show 20-30 
seconds after the source is deleted.
    
    
![no_jitter](https://cloud.githubusercontent.com/assets/211822/15052501/0542e8b8-12cc-11e6-8aec-73e83ef8478a.png)
    
    With no jitter code, scheduler utilization shoots close 100% in shorter 
period of time, same thing with io usage. Then after all replications sleep 
they'll have another spike like that when they wake up together.
    
    
![with_jitter](https://cloud.githubusercontent.com/assets/211822/15052507/09f6e648-12cc-11e6-9227-53d5cd70e133.png)
    
    With jitter, scheduler utilization stays around 50, load is spread more 
uniformly. Same story with IO.



---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---

Reply via email to