Adding a data store
Locked
 
|
Greetings All,
My name is Keith Wood and I currently work as the business development manager for a little known consulting company based in Frankfurt, Germany. As part of some business development work that I am currently doing, I find myself wanting to suggest that someone add the required functionality such as would allow users of QuantLib to Fetch/Process/Store data in a relational database. I envisage this as being the next step up from QuantLibAddIn where rather than providing the hooks to a spreadsheet, you get hooked into a data schema. I was thinking something along the lines of a generic open-source model that allowed you to "install" a bunch of tables and "stuff" on your favorite flavor of relational data provider. I find myself in the somewhat unique position of being able to call upon the engineering group of Sybase as resources to implement this. Is there any interest from your side in talking about this further. If so, who should I be talking to. Regards Keith Wood ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
Hi Keith, On Fri, 2007-03-09 at 22:22 +0100, Keith Wood wrote: > I find myself wanting to suggest that someone add the required > functionality such as would allow users of QuantLib to > Fetch/Process/Store data in a relational database. > I find myself in the somewhat unique position of being able to call upon > the engineering group of Sybase as resources to implement this. It seems like an offer one can't refuse. > Is there any interest from your side in talking about this further. > If so, who should I be talking to. Posting to this mailing list is OK---in fact, it's the best place to discuss the issue. As for me, I'm mostly concerned with the core library, so I don't have the clearest view of the QuantLibAddin architecture. But I guess someone with more knowledge in that particular field (Eric?) will reply shortly. Later, Luigi ---------------------------------------- Discontent is the first necessity of progress. -- Thomas A. Edison ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
In reply to this post by Keith Wood
Hi Keith,
Many thanks for getting in touch. Some progress has already been made on a design for extending QuantLibAddin to support serialization, and my first thought is whether a common framework could serve as a basis both for serialization (say for distributed computing) and an RDB store. Here are the links to the previous discussion on serialization: http://sourceforge.net/mailarchive/forum.php?thread_id=6507779&forum_id=4300 http://sourceforge.net/mailarchive/forum.php?thread_id=6539684&forum_id=4300 http://sourceforge.net/mailarchive/forum.php?thread_id=6540997&forum_id=4300 http://sourceforge.net/mailarchive/forum.php?thread_id=6543552&forum_id=4300 http://sourceforge.net/mailarchive/forum.php?thread_id=6700844&forum_id=4300 Please also take a look at the Value Objects feature, contributed to QuantLibAddin by Plamen Neykov. Each time a QuantLib object is constructed in ObjectHandler, a Value Object is associated with it, the VO comprises a snapshot of the inputs to the constructor of the QL object, and the idea is that VOs would be used as the basis for, say, reconstituting the same object on another machine. VOs can be interrogated with functions ohPropertyNames() and ohPropertyValues(). I'd be interested to hear your reaction to these initial ideas. Kind Regards, Eric On 3/9/07, Keith Wood <[hidden email]> wrote: > Greetings All, > > My name is Keith Wood and I currently work as the business development > manager for a little known consulting company based in Frankfurt, > Germany. As part of some business development work that I am currently > doing, I find myself wanting to suggest that someone add the required > functionality such as would allow users of QuantLib to > Fetch/Process/Store data in a relational database. > I envisage this as being the next step up from QuantLibAddIn where > rather than providing the hooks to a spreadsheet, you get hooked into a > data schema. I was thinking something along the lines of a generic > open-source model that allowed you to "install" a bunch of tables and > "stuff" on your favorite flavor of relational data provider. > I find myself in the somewhat unique position of being able to call upon > the engineering group of Sybase as resources to implement this. > > Is there any interest from your side in talking about this further. > If so, who should I be talking to. > > Regards > Keith Wood > > ------------------------------------------------------------------------- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to share your > opinions on IT & business topics through brief surveys-and earn cash > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > _______________________________________________ > QuantLib-dev mailing list > [hidden email] > https://lists.sourceforge.net/lists/listinfo/quantlib-dev > ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
Hmm, serialisation / data store. I think we might be talking
"differently" here. Let me deal with the FPML bit first. This might be a little un-politically correct, but here goes anyway. I know a number of people who hold senior roles on the FPML committee for an assortment of different threads. It is a committee based thing that takes a little time to review stuff, agree stuff, document stuff and then publish stuff. Evidence of this is seen in the manner in which the SwapsWire people have built their own SWML and extended-FPML that deals with things that are not yet in the published FPML. There is a similar type of thing being done by the nice people at DTCC to encapsulate their electronic confirmation matching service called DerivServ. The nice people at MarkIT are also working with an extended variant of the FPML stuff. At some point this will all converge, but don't hold your breath as by the time the FPML people catch up, the others will have moved on. My suggestion had to do with the design and implementation of a data schema that would enable people who use QuantLib to store stuff in tables in a relational database. All of the holiday stuff belongs in a table. All of the day count convention could usefully be stored in a collection of tables. Having gone to the trouble of getting some market rates from somewhere, using them to calculate the zero's, strips, smiles, complicated volatility surfaces and so on for "today's" pricing, it might be nice to save "stuff" in a relational table (or several). FPML, SWML and all the other "flavours" that have to do with XML stuff, lend themselves to being mapped to a collection of relational tables in a sensible manner. There are already collections of tools that do most of that for you. There seems to be a need to apply some assistance to the choice of which-XML, and this is loosely connected to a similar decision about which-SQL. I will now take the liberty of talking to the nice people at Sybase to see just how generous they feel about professional assistance in this little venture. (No Promises, no commitments, intended as a feedback thing to test the waters) I will get back to you shortly. Regards Keith p.s. please feel free in the meantime to bombard me with questions and or flame me for not being politically correct. eric ehlers wrote: > Hi Keith, > > Many thanks for getting in touch. > > Some progress has already been made on a design for extending > QuantLibAddin to support serialization, and my first thought is > whether a common framework could serve as a basis both for > serialization (say for distributed computing) and an RDB store. > > Here are the links to the previous discussion on serialization: > > http://sourceforge.net/mailarchive/forum.php?thread_id=6507779&forum_id=4300 > > http://sourceforge.net/mailarchive/forum.php?thread_id=6539684&forum_id=4300 > > http://sourceforge.net/mailarchive/forum.php?thread_id=6540997&forum_id=4300 > > http://sourceforge.net/mailarchive/forum.php?thread_id=6543552&forum_id=4300 > > http://sourceforge.net/mailarchive/forum.php?thread_id=6700844&forum_id=4300 > > > Please also take a look at the Value Objects feature, contributed to > QuantLibAddin by Plamen Neykov. Each time a QuantLib object is > constructed in ObjectHandler, a Value Object is associated with it, > the VO comprises a snapshot of the inputs to the constructor of the QL > object, and the idea is that VOs would be used as the basis for, say, > reconstituting the same object on another machine. VOs can be > interrogated with functions ohPropertyNames() and ohPropertyValues(). > > I'd be interested to hear your reaction to these initial ideas. > > Kind Regards, > Eric > > On 3/9/07, Keith Wood <[hidden email]> wrote: >> Greetings All, >> >> My name is Keith Wood and I currently work as the business development >> manager for a little known consulting company based in Frankfurt, >> Germany. As part of some business development work that I am currently >> doing, I find myself wanting to suggest that someone add the required >> functionality such as would allow users of QuantLib to >> Fetch/Process/Store data in a relational database. >> I envisage this as being the next step up from QuantLibAddIn where >> rather than providing the hooks to a spreadsheet, you get hooked into a >> data schema. I was thinking something along the lines of a generic >> open-source model that allowed you to "install" a bunch of tables and >> "stuff" on your favorite flavor of relational data provider. >> I find myself in the somewhat unique position of being able to call upon >> the engineering group of Sybase as resources to implement this. >> >> Is there any interest from your side in talking about this further. >> If so, who should I be talking to. >> >> Regards >> Keith Wood >> >> ------------------------------------------------------------------------- >> >> Take Surveys. Earn Cash. Influence the Future of IT >> Join SourceForge.net's Techsay panel and you'll get the chance to >> share your >> opinions on IT & business topics through brief surveys-and earn cash >> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV >> >> _______________________________________________ >> QuantLib-dev mailing list >> [hidden email] >> https://lists.sourceforge.net/lists/listinfo/quantlib-dev >> > ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
On Mar 13, 2007, at 7:58 PM, Keith Wood wrote: > My suggestion had to do with the design and implementation of a data > schema that would enable people who use QuantLib to store stuff in > tables in a relational database. All of the holiday stuff belongs in a > table. All of the day count convention could usefully be stored in a > collection of tables. I'm not sure that I follow here. I can see that the definition of, say, an equity option (exercise date, strike...) can be stored in a table. A calendar I might see---even though our calendars are not a collection of holidays, but rather a collection of rules such as "the third Monday in September" or "May 1st unless it's a Sunday, in which case the following Monday is a holiday instead". But a day-count convention is basically an algorithm. I might just have misinterpreted you, but how are you going to store them? Later, Luigi ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
So a simple vanilla swap is where I pay fixed and receive floating.
I can save most of that, applications such as Summit, Murex, Calypso and so on do this all the time. If I initiate this swap today, then for a 3M Euro-Euribor swap, in about 88 days time I will stop guessing what the floating rate is for the 2nd payment and FIX it with the prevailing market rate. This might change the NPV of the swap. I could create a portfolio of derivative instruments and then look at the effects of changes in market rates on the valuation of those instruments over time. I can accept that holidays are based on a rule. Calendar-Week-1 is the Week where the first Thursday is a January. I am not sure how you would do the Easter thing, but I accept that there is a rule for this somewhere. Whilst it is nice to know when the Eurex people stay at home, there are also religious holidays for the individual states in Germany. I accept that you can do rules for all of these, and that so far nobody was interested in when Christi did the himmelfart thing from a Bavarian viewpoint. The daycount stuff is rather trivial for a quant-type (no offence) but there are a bunch of people who still ask what the convention is for MM, Bond, Repo and Swaps in EUR. A trivial collection of tables but it saves me from having to remember, when I can go and do a select. (See it is even trivial for a non-quant-type) I am looking at a reversed view of this. I already have a large data warehouse that contains all of the trades and positions, now I want to apply the QuantLib functions on those instruments. I can fetch stuff into the arrays and structures and then do the kinky valuation thing. I might want to do this on a regular basis and then also store the result set back in the warehouse. I have a very specific interest. I am working on a data warehouse that allows me to store the ticks and quotes from more than one exchange based platform. If I have all, and I mean all of the ticks and quotes from the various different exchanges where they trade Daimler-Chrysler I can do some interesting things when it comes to calculated volatility. Currently, quite a lot of the tick-by-tick stuff is hidden away in what people call "Dark Liquidity". MiFID will change that here in Europe. suddenly the Tick & Quote information from OTC and internal trades will be published. A lot of people who currently calculate risk on the basis of High/Low/Open/Close will have access to much more granular information. I have a standard SQL based means to collect and store that information in and efficient manner that will also allow you to read from the tick-pool in near real-time (<10ms). Having this in a big bucket is a nice idea, it would be nice if there was an "open" collection of financial analysis tools that allowed people to access and analyse this information. I have a data schema that allows me to collect lots of interesting information about business entities. I have a data schema that allows me to collect lots of interesting information about financial instruments. The business entity stuff is a collection of Citibank, BarCap, RBOS, HBOS data that has been scrubbed and reconciled. The instrument stuff is based on the internal instrument definitions used by Eurex, Xetra, Xontra and Clearstream. I have a feed handler and subscription that ensures that this is updated. I am working on a Tick&Quote data model that allows me to collect data from most of the recognised electronic platforms in the world. Rather than trying to read 400000 ticks for the Bund-Future from a file, I plan to be able to get this from a table. I currently have some test data for 1-day that has 10,823,981 ticks in it. (it was a quite day) I also have all of Eurex for 30-days. Whilst I accept that there are not that many ContinuousFloatingLookbackOptions in there, doing Monte Carlo and Stochastic stuff to this might be interesting for some people. The nice people at Sybase like that I have done all of this using their product line. It is loosely based on their RAP product line where they have Sybase-IQ as the VLDB time series database and an in-memory version of the regular Sybase-ASE data server for the relational stuff. They are sufficiently impressed and have asked me what else they could do to assist in further enhancing this basic platform. I figured that putting some element of persistence around QuantLib would be nice, as this would add an open source collection of accepted and approved financial math to the picture. I have not yet mentioned QuantLib to them yet, as I wanted to get some idea for what the QuantLib people thought about this. Sybase is a big commercial data vendor that has a previous history of being nice to the open world. Hopefully I did not get too arrogant in all of the above, or make any assumptions that I should not have done. equally as hopefully, you can let me know what you think about what might at first seem like a wacky idea. Regards Keith Luigi Ballabio wrote: > > On Mar 13, 2007, at 7:58 PM, Keith Wood wrote: >> My suggestion had to do with the design and implementation of a data >> schema that would enable people who use QuantLib to store stuff in >> tables in a relational database. All of the holiday stuff belongs in a >> table. All of the day count convention could usefully be stored in a >> collection of tables. > > I'm not sure that I follow here. I can see that the definition of, > say, an equity option (exercise date, strike...) can be stored in a > table. A calendar I might see---even though our calendars are not a > collection of holidays, but rather a collection of rules such as "the > third Monday in September" or "May 1st unless it's a Sunday, in which > case the following Monday is a holiday instead". But a day-count > convention is basically an algorithm. I might just have misinterpreted > you, but how are you going to store them? > > Later, > Luigi > ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
In reply to this post by Keith Wood
Hi Keith,
That's an impressive summary of the resources available to you for your project, both in terms of the data and the potential support from Sybase. > Hmm, serialisation / data store. I think we might be talking > "differently" here. I accept that serialisation and data store are separate undertakings, I'm just thinking that as we tackle data store we should keep in mind the design for serialisation, in case the two projects have some common building blocks. If we get into the details and find that the two projects have zero in common, then so be it. I suspect we'll find there's some overlap. > Let me deal with the FPML bit first. This might be a little > un-politically correct, but here goes anyway. > I know a number of people who hold senior roles on the FPML committee > for an assortment of different threads. > It is a committee based thing that takes a little time to review stuff, > agree stuff, document stuff and then publish stuff. > Evidence of this is seen in the manner in which the SwapsWire people > have built their own SWML and extended-FPML that deals with things that > are not yet in the published FPML. > There is a similar type of thing being done by the nice people at DTCC > to encapsulate their electronic confirmation matching service called > DerivServ. > The nice people at MarkIT are also working with an extended variant of > the FPML stuff. > At some point this will all converge, but don't hold your breath as by > the time the FPML people catch up, the others will have moved on. Interesting story, I had no idea the whole process was so fraught. When it comes time to move into FpML, I wouldn't lose sleep over the implementation details, I'd just do whatever seems most sensible at the time, with the option to adjust later as things evolve. > I will now take the liberty of talking to the nice people at > Sybase to see just how generous they feel about professional assistance > in this little venture. (No Promises, no commitments, intended as a > feedback thing to test the waters) Sounds good, please keep me posted and let me know if there's anything I can do to help. Regards, Eric ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
«
Return to quantlib-dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |