Hi Jan

   Thanks for your detailed explaination. Based on what you said, arangodb 
(mmap engine) is similar with mongodb (mmapv1). 

   Well I can wait for 3.2 anyway.

Regards

gelin yan

在 2017年6月12日星期一 UTC+8下午3:28:15,Jan写道:
>
> Hi,
>
> when index size gets bigger than the available RAM, then it depends on the 
> configuration of your operating system what will happen exactly:
> - when there is swap space configured and swap space is available, the OS 
> will assign disk-based memory for the index data, which will slow down 
> index operations from in-memory speed to on-disk speed
> - if there is no swap space available, then either operations in ArangoDB 
> will fail with an out-of-memory error, or worse, the operating system may 
> trigger its OOM killer to terminate the process with the highest RAM usage, 
> which will likely be arangod.
>
> Note that all of the above applies for ArangoDB 3.1 and earlier. From 
> ArangoDB version 3.2 on, we do offer an alternative storage engine based on 
> RocksDB. This storage engine is designed to handle data bigger than the 
> available RAM. With the RocksDB-based storage engine, index data that is 
> not in RAM will be loaded from disk when it is not in the in-memory hot 
> set. So operations that need to load data from disk will be slower than 
> operations that can access data already in RAM. The engine will 
> automatically keep the hot set in memory, so if you're only accessing a hot 
> set of data that fits into the available RAM, it should be fine. If you 
> will be accessing your entire data set and it is much bigger than RAM, then 
> all operations will slow down. But there is no way to avoid this, even with 
> other databases, which will behave similarly.
> ArangoDB 3.2 is currently in the end of its alpha stage, and will very 
> soon go into beta stage. I recommend to give it a test drive and try the 
> new storage engine that is shipped with it.
>
> Best regards
> Jan
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"ArangoDB" 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.

Reply via email to