Optiq + Phoenix

[Tom Theng]

Hi,

I was wondering if Optiq works with Phoenix? Any know gotchas or shortcoming ?

Thx!

[Ted Dunning]

Optiq certainly *could* work with phoenix.  I am pretty sure that it doesn't now.

Bolting Optiq onto something is a very significant job if only because you have to think a lot about what you exactly want it to do for you.  But it should definitely be possible.

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

[Julian Hyde]

Ted is correct that it’s non-trivial to build an adapter to a new data source. I wouldn’t say “very significant” — you can get quite a useful adapter by just pushing down table scans, and doing all other relational operations (filter, aggregate, join etc.) in the client. If you want further optimizations, you need to push down more operations, and that takes thought.

But I believe that Phoenix does work reasonably well. After all, it presents a JDBC interface, and Optiq’s JDBC adapter is fairly mature.

Gabriel, Bruno, Steven or Karel: Can you give more details about your integration between Optiq and Phoenix?

James Taylor and I have discussed how Optiq could be used to power Phoenix (thus we’d be talking about an HBase adapter) but nothing has come from those discussions at this point.

Julian

PS Since Optiq is now an Apache incubator project, let’s start using the Apache mailing list, http://mail-archives.apache.org/mod_mbox/incubator-optiq-dev/. I encourage you all to subscribe ASAP.

[Ted Dunning]

On Mon, Jun 9, 2014 at 4:11 PM, Julian Hyde <julia...@gmail.com> wrote:

I encourage you all to subscribe ASAP.

Ooops.

Good reminder.  I was already subscribed and forgot to push the discussion there.

(and me a mentor and all ... disgraceful)

[Tom Theng]

Great thx everybody .. this is exactly what I'm looking entities in phoenix/hbase and aggregates for slicing and dicing in optiq ... 

FWIW I can report back with a quick write up on experiment when i'm done ...

On Tue, Jun 10, 2014 at 11:05 AM, Steven Noels <ste...@apache.org> wrote:

(moving)

On Mon, Jun 9, 2014, at 04:11 PM, Julian Hyde wrote:



Ted is correct that it’s non-trivial to build an adapter to a new data
source. I wouldn’t say “very significant” — you can get quite a useful
adapter by just pushing down table scans, and doing all other
relational operations (filter, aggregate, join etc.) in the client. If
you want further optimizations, you need to push down more operations,
and that takes thought.



But I believe that Phoenix does work reasonably well. After all, it
presents a JDBC interface, and Optiq’s JDBC adapter is fairly mature.



Gabriel, Bruno, Steven or Karel: Can you give more details about your
integration between Optiq and Phoenix?

Hey everybody,



we can definitely confirm that this is possible. On the entity-side, we
have Phoenix, on the customer-level aggregates side, there's Optiq -
both accessing data which ultimately resides in HBase. And you can
access the Phoenix data using the Optiq JDBC driver transparently.



Steven.