Not using memory to store strategies

if I enable storage of strategies to drive instead of mem it still loads all strategies to memory in re-tester.  So loading 35 k strategies :) takes 50 gig of ram. And even if I have more ram then 50 gig  SQx still get unresponsive.. So whats the problem here why have a setting storing strategies to drive if it do not work.
Attachments
No attachments
  • Votes 0
  • Project StrategyQuant X
  • Type Bug
  • Status Refused
  • Priority Normal

History

m
#1

mabi

24.03.2019 09:09

Task created

MF
#2

Mark Fric

25.03.2019 10:28

Status changed from New to Waiting for information

the file-based databanks don't mean that the strategies will not be stored in memory, only that they are immediately saved to disc when inserted to databank. So they won't be lost if program crashes.


Could you attach a few of your strategies so I can check what takes so much memory? 

Perhaps we can add an option to not save Monte Carlo and WF results to files to save memory when you load them next time.

m
#3

mabi

03.04.2019 16:26

Attachment EURUSD_H1_0324.7z added

Zipped a few..  
MF
#4

Mark Fric

29.05.2019 13:28

Status changed from Waiting for information to Refused

as I wrote above, strategies are always stored in the memory. So the recommendation is to not try to work with that many strategies. When you have 35k strategies, with 1000 orders each, it is 35 mil. orders. When you also have MC or WF results stored for the strategy, it can easily be 100x that.


In any case, I tried to load 20k strategies (made a copy of those you sent). It took around 11 GB RAM, and work with SQ is relatively normal, no big lagging in the UI.

We made many optimizations in the recent builds, they probably helped as well.

m
#5

mabi

29.05.2019 23:27
121 is much better then 119 was.

Votes: 0

Drop files to upload

or

choose files

Max size: 5MB

Not allowed: exe, msi, application, reg, php, js, htaccess, htpasswd, gitignore

...
Wait please