4

i have recently discovered MonetDB and i am evaluating it for an internal project, so probably my questions are from a really newbie point of view. Maybe someone could point me to a site and/or document where i could find more info (i haven't found too much googling)

regarding scalability, correct me please if i am wrong, but what i understand is that if i need to scale, i would launch more server instances and discover them from the control node, is it right? is there any limit on the number of servers?

the other point is about storage, is it possible to use amazon S3 to back MonetDB readonly instances?

update we would need to store a massive amount of Call Detail Records from different sources, on a read-only basis. We would aggregate/reduce that data for the day-to-day operation, accessing the bigger tables only when the full detail is required. We would store the historical data as well to perform longer-term analysis. My concern is mostly about memory, disk storage wouldn't be the issue i think; if the hot dataset involved in a report/analysis eats up the whole memory space (fast response times needed, not sure about how memory swapping would impact), i would like to know if i can scale somehow instead of reingeneering the report/analysis process (maybe i am biased by the horizontal scaling thing :-) )

thanks!

Iosu S.
  • 154
  • 11
  • 1
    Could you quickly explain what/why you want to scale? What resource do you think will be exhausted in a single machine setup (Disk, RAM, CPU)? Generally, scaling out is not one of the target problems in the MonetDB development. We generally try to get the maximum out of a single machine and only scale out when absolutely necessary. – Holger May 08 '13 at 09:30
  • thx Holger, i have updated the question with the info – Iosu S. May 09 '13 at 11:36

1 Answers1

3

You will find advantages of monetdb easily on net so let me highlight some disadvantages
1. In monetdb deleting rows does not free up the space
Solution: copy data in other table,drop existing table, and rename the other table
2. Joins are little slower
3. We can can not give table name as dynamic variable
Eg: if you have table name stored in one main table then you can't make a query like "for each (select tablename from mytable) select data from tablename)" the sql You can't make functions with tablename as variable argument.

But it is still damn fast and can store large amount of data.

Gaurav
  • 2,003
  • 1
  • 25
  • 50