The semantics with "clusters" in ODB

[scott molinari]

Hi,

From a horizontal scaling, distributed database standpoint, I feel the semantics used in ODB aren't quite right (although they are pretty cool nonetheless). I apologize for being pedantic, but to me, a "cluster" should be a group of nodes in the distributed data setup. As I understand computing distribution, what ODB calls "clusters" should actually be called "shards", as it is the concept of sharding that is being done with the clusters, right? 

http://www.webopedia.com/TERM/D/database_shard.html

Or am I totally out to lunch?

Scott

[Luigi Dell'Aquila]

Hi Scott,

I think you are right, and I think all the rest of OrientDB team will agree with you. The name "cluster" is completely misleading when talking about physical data portions. 

Some time ago I proposed to use the name "data file" to refer to physical files that contain data (that is what clusters are now), but "shards" makes a lot of sense as well (at a logical level).

The real problem here is that all our docs use the word "cluster" to refer to data files and a lot of users are used to that.

From my point of view, we could start to change this naming and gradually review all our docs

My 2 cents

Luigi

--

---
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 orient-databa...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[scott molinari]

Hi Luigi,

I am glad I am not out to lunch! LOL!

Because it is such a big change, I'd consider it something to tackle in 3.0 actually. I wouldn't even mind helping with the docs.

How about the code? Are there a lot of changes to be made, so the code also uses the proper semantics?

This page would make a whole lot more sense after the change. :-)

http://orientdb.com/docs/last/Distributed-Sharding.html

Sentences like this were a "WTF" moment for me initially, coming from MongoDB.

OrientDB automatically creates a new cluster per each class as soon as node joins the distributed cluster.

Much better....

OrientDB automatically creates a new shard per class as soon as a node joins the distributed cluster.

:-)

Scott

[Eric24]

+1

[Francisco Reyes]

but to me, a "cluster" should be a group of nodes in the distributed data setup

+1

If we decide on a name I could help with the documentation, or have someone in my company do it, so have we decided on a name yet?

[Edit to clarify - documentation]

[scott molinari]

For those of you interested, you can vote for this OEP (OrientDB Enhancement Proposal) for Version 3.0, which I've created for this subject.

https://github.com/orientechnologies/orientdb-labs/issues/12

If you have any other improvements to the proposal I could add, please let me know!

Scott