Re: Impact of persistent NVM Memory (pnvm) on ZODB's performance

> On Oct 5, 2017, at 11:55 PM, Yashwardhan Singh <[email protected]> wrote:
> 
> I am new to zodb and am interested in knowing if persistent non-volatile memory, will help in speeding up zodb. Is there any benchmark available for zodb for performance evaluation? With persistent nvm. the objects which needs to be persistent can be directly allocated on it and thus might help in getting rid of pickling cost. Any comments?

What an excellent question.  As was the other question about Asycn with the ZODB. 

So let me put this question in context.  First they had hard drives, and all of the software targeted hard drives.  Then along came non-volatile memory and people started running flle systems on top of non-volatile memory.   (SSD).  Cell phones for example have no hard drive, but at least iOS has a Unix file system.   Why not just do a  persistent objects on non-volatile memory?  Sounds like a great idea to me.  

As for ZODB, I recall reading on this mailing list that if you run it on an SSD, the hard drive is no longer the bottle neck, rather it is the ZEO server which becomes the bottleneck.  But maybe that was before the ZODB was 
able to write to multiple files at the same time. 

It would be great to get a version of the ZODB which writes directly to non-volatile memory, and bypasses any file system on an SSD. .  But even if we did that, where would you get an ISP or cell phone which allows you to write directly to non-volatile memory?

Warm Regards
Chris





-- 
You received this message because you are subscribed to the Google Groups "zodb" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
For more options, visit https://groups.google.com/d/optout.