Hi,

We are still evaluating Databases. MongoDB was looking good, and now we 
like what OrientDB has to offer, since we will also be needing a graph 
database and the ability to use a flexible schema database with it is like 
a dream come true. However... 

I noticed in some (a bit older) benchmarks, that ODB is on the slower end 
of the performance scale compared to other document and graph data stores. 
Is this due to the split it tries to make between the two types? Is there 
promise that it will ever get as quick as the others? Or is the lack of 
performance a compromise needed for having such a very flexible and yet 
ACID compliant datastore?

Do people using ODB still need some sort of caching, which is normal 
practice with RDBMS, to take up slack in performance? 

Do people using ODB, still need an indexing system, like Elasticsearch, to 
take up slack in performance (when needing lots of indexing)?

We are looking at a multi-tenant system and have decided, due to the 
expressiveness of the system we are building, it would be much easier (and 
safer) to just give each customer their own database. Some questions on 
that.

1. Is the disk footprint of each ODB database relatively small? Does it do 
anything like the disk pre-allocation stuff that MMap in MongoDB does?
2. Would a good number of databases mean the performance of an ODB instance 
would be degraded at all? Let's assume the hardware and I/O can handle the 
number of databases. Purely talking about ODB, would it start to falter in 
anyway, because multiple databases are being used?
3. Is it a multi-threaded system?

On other topics:
1. Does ODB have an issue with document size growing above allocation size 
(if it has it) and any necessary movement of the data on disk like Mongo 
has (with MMap)?
2. The bug tracking of ODM seems, well, like it isn't being done well. This 
might be only my first impression. Can anyone expand on their experience?
3. I tried to find documentation on it, but how does ODB handle network 
partitioning? MongoDB can cause a client to read an inconsistent state and 
even incorrect data (in a very tight time window), when there is a network 
partition between replication nodes and during the continued spit. Does ODM 
have any such similar issues?

Thanks for any help on any questions in advance.

Scott

-- 

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