UK Ruby on Rails, Exalead, AWS, Consultancy - InfiniDB, Infobright and MonetDB - Day 2: Infobright Comments

Comment on 'InfiniDB, Infobright and MonetDB - Day 2: Infobright' by erin

2010-07-06T20:33:03Z

Infobright does indeed start to shine above 100M rows & esp in the billions. At this number MySQL doesn't handle deletes, updates very well either. AND an ALTER in MySQL can take days.

I found that a highly optimized schema (all ints & decimals no chars/varchars) and queries that are optimized for Infobright meant that queries that took days took seconds & minutes. AND the time a query took to return didn't vary much whether there were 1 billion or 10 billion rows.

In a col based DB it is possible to have queries take hours & days if you run a SELECT * , rather than selecting a few of the columns. Also the number of items returned will affect the time a query takes to return. If I'm returning 6 rows it will be faster than the query returning 55K. The less optimized queries also tend to take a few seconds more each billion record added.

Also a big PLUS for Infobright is the compression. Because our schema is all ints & decimals the compression is pretty amazing. (last I checked it was 6gb Infobright = 414gb Mysql.) Our table is fairly wide for MySQL standards but ok in Infobright. A friend of mine has found that while NULLs can be problematic in MySQL they're great for Infobright....

My only complaint about ICE (Infobright's open source product) is that loading data BLOCKS SELECTS! They've fixed this in the enterprise version but not ICE. I'm a bit unhappy about that.

Comment on 'InfiniDB, Infobright and MonetDB - Day 2: Infobright' by Ace

2010-03-01T22:58:00Z

Nice writeup, but to make a case for or against infobright, you need to load up the DB upto atleast 1Billion rows. That is a real test.

No, I an no way related to the company that produced Infobright.

Comment on 'InfiniDB, Infobright and MonetDB - Day 2: Infobright' by Chris Anderton

2009-11-25T15:51:32Z

@igor: I removed the names of my columns to avoid revealing any secrets (not that I think there were any in there! I'll update them so they're obvious which columns are which as I just called everything X

@susan: Thanks for reading - at the moment in this four table example the volumes are 4.5m, 300k, 10m and 2m rows. The 10m rows are a subset of a larger 110m row table so it may be worth me sucking in all 110m rows to see how it changes things?

Comment on 'InfiniDB, Infobright and MonetDB - Day 2: Infobright' by Susan Davis

2009-11-25T14:56:28Z

Hi Chris,

Thanks for doing these tests - it is good info for Infobright as well as the rest of your readers.

Can you tell me what volume of data you have? While we certainly expect to see great query performance versus an OLTP type database like InnoDB, users will find that as the database gets bigger, the speed of Infobright versus MySQL will become even more pronounced.

Regards

Susan Davis, Infobright

Comment on 'InfiniDB, Infobright and MonetDB - Day 2: Infobright' by Igor

2009-11-25T14:36:23Z

WHERE t1.x = 125 AND t1.x BETWEEN "2009-01-01" AND "2009-10-10" ORDER BY t1.x DESC, t1.x DESC

What do you mean by that - how can t1.x be 125 and BETWEEN "2009-01-01" AND "2009-10-10" at the same time?