http://ayende.comAyende @ RahienCopyright (C) Ayende Rahien 2004 - 2021 (c) 202660"for me relational = joins because that is the way you model your data" That is not how the creator of the relational model defines it and VoltDB themselves are pretty explicit in stating their goals where a relational db to complete with NoSQL alternatives. From their FAQ :"VoltDB is a SQL RDBMS for database applications that have extreme transaction processing (XTP) workloads" And "VoltDB is a relational store that uses SQL for its query language and stored procedures for its transactional unit of work. " So what are they lying? Again every relational implementation has limitations based on it's architecture. Relations are Tables, SQL is relational algebra, even without joins. Your just trying to move the goal posts. http://ayende.com/4552/nosql-and-data-warehousing#comment11http://ayende.com/4552/nosql-and-data-warehousing#comment11Sat, 03 Jul 2010 21:36:27 GMTNoSQL is a very wrong term,not relational is way better. i can give you a SQL interface over whatever you want without joins! and if you would read a bit more about VoltDB you will see that the joins are very limited! f you think about it is a in-memory row store with a ability to do lookups to other read-only tables! you can`t create a fully normalized schema with VoltDB you will be doing too many joins. for me relational = joins because that is the way you model your data,if you don`t have joins(or have a limited form of joins) you start modeling your data in a very different way. i really recommend you to install VoltDB(very easy install) and try the auction db,then go to the adhoc query tool(need python for that) and try to do some joins,see what you get :) by the way VoltDB is more comparable to TimesTen or if we don`t look at their fancy SQL interface to redis(which i think is way better than VoltDB) http://ayende.com/4552/nosql-and-data-warehousing#comment10http://ayende.com/4552/nosql-and-data-warehousing#comment10Sat, 03 Jul 2010 16:06:54 GMTVoltDB is not a RDBMS,not in a traditional way,first you can do unlimited JOINs only to replicated tables(those table aren`t partitioned and should be mostly read-only and small if you want to scale) or to 1 non-replicated table. also your queries need to be single-sited(their term for hitting only one partition of the table those only hitting one server) in order to get a fast query. this is how i see VoltDB: a auto shrading layer on top of a in-memory relational-like storage without even auto balancing or adding nodes online,in my opinion using a sharded Tokyo Cabient or some other fast KV store you will get better performance. in short:VoltDB is a NoSQL(in sense it is not really relational) with a SQL interface :) http://ayende.com/4552/nosql-and-data-warehousing#comment9http://ayende.com/4552/nosql-and-data-warehousing#comment9Fri, 02 Jul 2010 09:33:38 GMTThanks for answering my question. I was thinking in the same line as you. Since analysis and reporting of data is essential to DW/BI, current NoSQL solutions are not up to par with this yet. Given time, more tool sets will be available to accommodate for this need in the NoSQL arena. http://ayende.com/4552/nosql-and-data-warehousing#comment8http://ayende.com/4552/nosql-and-data-warehousing#comment8Thu, 01 Jul 2010 15:55:58 GMT"one of the things that tend to lead people to the NoSQL land is the desire to scale in some manner (more data, more users, higher concurrency, cheaper TCO) than is possible using a SQL solution." Go tell that to the guys at Vertica, Greenplum, or VoltDB. SQL seems to scale just fine for them. SQL nor the relational model has much if anything to do with scalability, its more about which parts of ACID you turn off. http://ayende.com/4552/nosql-and-data-warehousing#comment7http://ayende.com/4552/nosql-and-data-warehousing#comment7Thu, 01 Jul 2010 15:39:31 GMTWhat kind of reporting tools are out there that work well with a NoSQL solution? Thanks, http://ayende.com/4552/nosql-and-data-warehousing#comment6http://ayende.com/4552/nosql-and-data-warehousing#comment6Thu, 01 Jul 2010 14:44:05 GMTSteve, Take a look at my NoSQL category, I have a lot of info about that http://ayende.com/4552/nosql-and-data-warehousing#comment5http://ayende.com/4552/nosql-and-data-warehousing#comment5Thu, 01 Jul 2010 13:38:23 GMTAlex, Sure, which is why I said that the NoSQL solution has less flexibility than the alternative. Map/reduce solutions are good for data crunching to _generate_ the report, as part of the solution (preparing ahead of time). http://ayende.com/4552/nosql-and-data-warehousing#comment4http://ayende.com/4552/nosql-and-data-warehousing#comment4Thu, 01 Jul 2010 13:37:22 GMT@Steve, check out the articles in the NoSQL category here on Ayende's blog: [http://ayende.com/Blog/category/565.aspx](http://ayende.com/Blog/category/565.aspx). In particular, this artical is really good at showing where NoSQL can scale in situations where a relational model cannot: [ayende.com/.../that-no-sql-thing.aspx](http://ayende.com/Blog/archive/2010/03/27/that-no-sql-thing.aspx)http://ayende.com/4552/nosql-and-data-warehousing#comment3http://ayende.com/4552/nosql-and-data-warehousing#comment3Thu, 01 Jul 2010 13:35:04 GMTAyende, could you provide or point me to an explanation of why this is true: "one of the things that tend to lead people to the NoSQL land is the desire to scale in some manner (more data, more users, higher concurrency, cheaper TCO) than is possible using a SQL solution." I gather that will be the first question I have to answer to the DBA :) http://ayende.com/4552/nosql-and-data-warehousing#comment2http://ayende.com/4552/nosql-and-data-warehousing#comment2Thu, 01 Jul 2010 12:49:27 GMT> I would probably choose a Column Family database for such a task, since my primary concern is probably being able to handle large amount of data. Leaving aside scaling, if I look at the current wide-column family of solutions (Cassandra, HBase, Hypertable) in terms of queries all I get is primary key indexes and range scans. I'd say that this is quite restrictive in terms of BI. You'll probably want to plug in something like Hadoop, Hive or Pig to really get closer to the features needed for a data warehouse. http://ayende.com/4552/nosql-and-data-warehousing#comment1http://ayende.com/4552/nosql-and-data-warehousing#comment1Thu, 01 Jul 2010 12:40:01 GMT