Q: some questions on db structure on alenka about alenka HOT 5 CLOSED

Hi Amihay
1.Inserts, updates and deletes are not implemented yet in Alenka although it is on a to-do list
2. You have guessed correctly, correlation is implemented by a rowid.

Thanks !

Anton

from alenka.

Hi Anton , thanks for the feedback, I would like to have phone call ,I was able to compile alenka on amazon vm + cuda (you can buy such vm ) , but when running i got strange errors.

If it is possible , I'll like to coordinate such call , if so , can you send your email to agonenil@gmail dot com to discuss it farther .

from alenka.

Hi !
Sorry, I was on a vacation, but I'm back now.
I'm available on Skype and I would be happy to help you with any issues.
I should note however that Alenka is not a something that you would
want to use in a production - it is just my personal reseach.

Regards,
Anton

On Sun, Jul 22, 2012 at 2:30 PM, Moti
[email protected]
wrote:

Hi Anton , thanks for the feedback, I would like to have phone call ,I was able to compile alenka on amazon vm + cuda (you can buy such vm ) , but when running i got strange errors.

If it is possible , I'll like to coordinate such call , if so , can you send your email to agonenil@gmail dot com to discuss it farther .

---

Reply to this email directly or view it on GitHub:
#1 (comment)

from alenka.

First of all - this is an awesome project. I didn't want to start a new question for a question related to the database structure - so here goes. Have you considered supporting Cassandra/HBase or some other columnar, noSQL database instead of a custom format?

Keep up the awesome work!

from alenka.

Hi Ananth !
The reason I used a custom format is that I use compression/decompression
algorithms that are not supported by HBase. The data I store are compressed
with a combination of FOR (frame of reference), FOR-DELTA and dictionary
compression. Reading uncompressed data into a GPU would have a significant
overhead because of slow speed of data transfers between main memory and a
gpu.
If there is a way to store the compressed data together with some metadata
in HBase structure I would be happy to support it.

Regards

Anton

On Wed, Sep 5, 2012 at 9:30 PM, Ananth [email protected] wrote:

First of all - this is an awesome project. I didn't want to start a new
question for a question related to the database structure - so here goes.
Have you considered supporting Cassandra/HBase or some other columnar,
noSQL database instead of a custom format?

Keep up the awesome work!

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/1#issuecomment-8308760.

from alenka.