Subject: re : real time var
tanya ,
thank you for the information . i agree that we need to talk about more what
and how global valuation can facilitate a more competitive var engine .
winston and nilay are going to give my group a presentation regarding the
current var system next tuesday . i am sure the presentation will help my team
better understand the requirements of var engine on global valuation . please
join us if you have the time . i would also appreciate your insight on this
matter .
thanks
zhiyong
tanya tamarchenko
01 / 08 / 2001 11 : 09 am
to : zhiyong wei / hou / ect @ ect , nilay basu / hou / ect @ ect , wenyao jia / hou / ect @ ect
cc : vince j kaminski / hou / ect @ ect
subject : re : real time var
ziyong ,
we met with nilay and winston last week regarding real time var calculation
possibility .
winston has an overview of var system which consists of :
- main curves simulation ;
- curve server ;
- book server ;
- id server
- clients
as a first step i want to see where the time is spent when var runs , which
percentages of time are spent by each part .
nilay is going to get this information for a few portfolios ( agg - ect , agg - gas
and ng - price - prc ) .
preliminary information : currently var run takes about 1 hour , half of this
time taken by " book server "
( we have about 4500 lower level portfolios in the portfolio hierarchy , about
5500 portfolios all together ) ,
most of the rest is taken by " clients " , " main curves simulation " does not
take much time .
i am looking also at using alternative methods of faster var calculation , but
having so many portfolios in the hierarchy will slow down
even analytical var .
we also have to think more about what " real time " calculation means and what
it should produce .
tanya