Re: Data Model Suggestion

2015-06-23 Thread Vladimir Rodionov
ly aggregates for 24 hours. This will > give > > a better performance boost. > > > > > > > > From: Pariksheet Barapatre [mailto:pbarapa...@gmail.com] > > Sent: Tuesday, June 23, 2015 8:07 PM > > To: user@phoenix.apache.org > > Subject: Re: Data Model Su

Re: Data Model Suggestion

2015-06-23 Thread James Taylor
nce will be better with 24 columns. Better add > one more column(25th) to keep daily aggregates for 24 hours. This will give > a better performance boost. > > > > From: Pariksheet Barapatre [mailto:pbarapa...@gmail.com] > Sent: Tuesday, June 23, 2015 8:07 PM > To: user@pho

RE: Data Model Suggestion

2015-06-23 Thread Puneet Kumar Ojha
@phoenix.apache.org Subject: Re: Data Model Suggestion Hi Puneet, Upsert could be easy - e.g. If we want to load data for 2015062412 --- ADID | COUNTRY | HOUR ID | CF.IMP

Re: Data Model Suggestion

2015-06-23 Thread Pariksheet Barapatre
nefits on use cases. Like breaking into column families , how > updates behave for a particular rowkey if you go with 24 columns. > > > > > > *From:* Pariksheet Barapatre [mailto:pbarapa...@gmail.com] > *Sent:* Tuesday, June 23, 2015 7:07 PM > *To:* user@phoenix.apache.org >

RE: Data Model Suggestion

2015-06-23 Thread Puneet Kumar Ojha
, 2015 7:07 PM To: user@phoenix.apache.org Subject: Re: Data Model Suggestion Thanks Skanda for your suggestion. If we go with 24 columns , won't it reduce number of rows over large data set and will improve scan performance. If I want to get daily report, I just have to scan only one row.

Re: Data Model Suggestion

2015-06-23 Thread Pariksheet Barapatre
Thanks Skanda for your suggestion. If we go with 24 columns , won't it reduce number of rows over large data set and will improve scan performance. If I want to get daily report, I just have to scan only one row. Your thoughts? Thanks Pari On 23 June 2015 at 18:57, Skanda wrote: > Hi Pari > >

Re: Data Model Suggestion

2015-06-23 Thread Skanda
Hi Pari For your use-case, having it as part of the rowkey should be a better design than creating so many columns. Regards Skanda On Tue, Jun 23, 2015 at 6:49 PM, Pariksheet Barapatre wrote: > Hello All, > > This is more like a HBase question but as I am planning to use Phoenix as > a access

Data Model Suggestion

2015-06-23 Thread Pariksheet Barapatre
Hello All, This is more like a HBase question but as I am planning to use Phoenix as a access layer, I hope phoenix user will help me. I would like to create time series data to get on-the-fly analytics. This use case is for adTech. Report - what is houly,daily,weekly impression counts at count