OrderBook


#1

Hi All

Is it possible to get a little more information on the orderbook example please guys? I am having some trouble getting it to work, and sense it is the Linux side (well documented Windows prisoner !) Just a little more help getting it to work would be appreciated.

The issues in a nutshell;
(1) running 12.04 Ubuntu and SciDB 14.3 working fine. Ran the lines; “apt-get install scidb-14.3-dev scidb-14.3-dev-tools scidb-14.3-libboost1.54-dev scidb-14.3-libmpich2-dev scidb-14.3-libboost1.54-all-dev” (as root) as the example suggests - github.com/Paradigm4/orderbook-example
(2) The problem appears with the next commands; “cd orderbook” - it does not recognise the folder - Am I making a basic Linux error here please ?
(3) All the subsequent commands also do not run
(4) Is it possible to define the headers of the csv file please ? The 11 fields
(5) Does SciDB’ers have any experience with FIX message files en.wikipedia.org/wiki/Financial_ … n_eXchange, the placing of them into SciDB database - using some basic R script to translate the files into a format similar to the ARCA example and hope to put them into SciDB
(6) The file example shows 200m rows for the day I think - I have 400m, and have 500 trading days - and the same again for a different exchange - 6 fields per line - Will SciDB be suitable ? Hope so.

Thanks in advance guys for any comments / assistance

G


#2

Hi, sorry about that. I added some more specific details to the readme file at:

github.com/Paradigm4/orderbook-example

that may help. You have to clone or otherwise download the repository from Github.

hundreds of millions of rows should be no problem. We shoot for about 1 million non-empty cells per chunk in the array on average, so try to pick chunk sizes accordingly.

Try to get that going and run through the basic ARCA example, and then we can help adapt it to your other data, OK?


#3

Many thanks Byran - let me take a look

So 500 Arrays (one per date,exchange pair) and each array with 400m ‘events’ (a line in the ARCA file) should be fine? Brilliant !

Noting, no need to aggregate across exchanges in this task

Regards


#4

Yes, that is OK.

Or make exchange a dimension and consolidate into a 3d array–might be easier in the long run to keep track of.