a large scale systems problem

Nonius · Post by **Nonius** » Fri Jun 06, 2008 10:12 pm

Ok, so I'm "tasked" with the, er, task of trying to spec out a global system to bring into one portal a bunch of different stuff related to hedge fund credit exposures and hedge fund investments. there's a bunch of different systems computing and storing differnt types of information and that information is stored all over the map and in some cases not even at the group level but in wholly (and in once case not wholly but JV) owned subsidiaries. but there are similarities of info that I want. so, there's a measure of worst case Gap risk on structured products that reference hedge fund investments. that measure is almost identical to credit exposures that we measure when trding with HF counteparties. but, the investment gap risk exposures are sitting at point A whilst the credit exposures are sittiting at point B. it doesn't stop there. I want to view exposures, ISDAs, collateral, mtms, covenants in CPPIs and other products, margin call disputes, gap risk, etc etc etc by pointing and clicking with a mouse in one simple portal, which will be my command center. Now, I don't even want to go into the details because they are way too messy.

what I need is a good reference book on how to even start, because it's actually scaring me to even think about the task. I don't want to tech spec things myself...I've dweebs who can do that. I need a good reference (book/weblinks/anything) on how to model requirements on a large scale business perspective.

Maggette · Post by **Maggette** » Fri Jun 06, 2008 10:59 pm

Since you are an expert in the field I guess you know what numbers you want, where these numbers are and what you want to do with them, when you have them, right?

So we are talking about how to model software requirements?

I am definetly no expert in this field, but I heard from two close friends who are in the "IBM buisiness process modelling JEE " (so really large buisinees scale) that the books by Karl E. Wiegers are good...maybe "Software Requirement Patterns " is for you..

Sorry if I got your question wrong and wasted your time

ZeroBeta · Post by **ZeroBeta** » Sat Jun 07, 2008 6:07 am

Is this event based information? i.e. Location_1 computes some new numbers that needs to be pulled in to Location_2 and Location_3 and so forth?

This sounds to me as an EAI (Enterprise Application Integration) problem, and the requirements could be modelled as such, and the books to search for would be about modelling requirements for large EAI solutions. I couldnt find one on requirements, but on the subject itself, this looks promising.

FYI there are three basic ways to do it:

i) 1980's: Location_1 dumps stuff into an ftp location, where applications Location_2 and Location_3 pick it up from when they need it.

ii) 1990's: Location_1 dumps stuff into a database, where Location_2 and 3 either read into it, or get it replicated locally and then read it. Or L_1 calls objects at L_2, L_3 over Corba/DCOM

iii) 2000's : Location_1 uses some messaging infrastructure to publish update messages, and Location_2, 3 subscribe to these messages. Or L1 publishes a webservice which L2 and L3 call into.

Nonius · Post by **Nonius** » Sat Jun 07, 2008 7:45 am

maggette, you hit it on the nail.

ZeroBEta, yes, it's an Enterprise Application Integration problem now that I think about it. I'd almost say it should be called an Enterprise Data Integration problem (if such a phrase exists).

On the "messaging infrastructure", that's more technical specificaiton isn't it?

at the end of the day, I mostly want see and aggregate (that might mean adding or maybe slightly more complicated operations) data.

one problem I'd like to solve is: if XYZ fund blows up, then I can point and click and see our credit and investment exposures in a flash.

thanks, I'll check your link.

maybe this thread should be in books.

Maggette · Post by **Maggette** » Sat Jun 07, 2008 9:14 am

Sorry, forgot to post the link

this is the book my friends were talking about..

cheers

edit: especially the "crash course" in chapter one might be interesting

Cheng · Post by **Cheng** » Mon Jun 09, 2008 10:06 am

We had similar issues here (read: integrating 75 different systems that did not know each other and were located all over the place). Not a book, but after evaluating tons of software we ended up with SAS to put things together. I know their HP is quite informative, you might want to have a look.

HTH.

weismat2 · Post by **weismat2** » Tue Jun 17, 2008 5:59 pm

For me this sounds like a data warehouse.

The main issue is to get, normalize and clean the data.

There are some special tools available market under ETL which help you to build the upload part and there are also special report tools out there. (SQL Server comes with a free tool, but I am not 100% sure about the quality).

Unfortunately I have not build any data warehouse for 5 years, but I think this is where you should start looking at.

SAS is one vendor in this area.

tristanreid · Post by **tristanreid** » Tue Jun 17, 2008 6:25 pm

I think you need:

1) 1 guy with some technical experience but mostly business-side guy who understands your needs. This guy doesn't need to spend 100% of his time on the project, and it could just be you, the function of this position is to push things forward and answer questions about priority of tasks. The person in this role has to be willing to learn tech-speak, which is why you want to get someone else to do it. A junior guy hoping to prove his worth would be the most valuable, the description of this role is 'Business Systems Analyst' or something like that.

2) 1 guy who is a techie, who understands how to run a website and can write scripts to download stuff from other places. Depending on your commitment to this project, you could also divide this task into: 1 database guy, 1 web guy, 1 developer.

3) 1 or 2 intern-level people who are hoping to learn teh biz but don't have teh mad skilz. Their job is to do lots of manual data entry. I think most tech-people would skip this step, but I think you'll get where you want to be a lot faster if you take this approach.

Once you've assembled the team, (1) goes around looking at websites and other data sources and starts putting together lists of everything you want to see, which data sources cost money and how much, etc. He regularly passes URLs to (2) to see how hard it would be to automatically grab data on a regular basis, and when it's not feasible (or during interim while (2) is working, (3) goes out and manually copies vast numbers of articles onto portal, via a simple text html template that (2) has provided. The career path for (3) is to become better at distilling and categorizing information by regularly asking questions or figuring it out. You ask (2) to make your web portal highly searchable and browseable, and make sure there's a way to add keywords for categorizing.

That's probably how I'd approach it, if I understand what you're trying to do.

-t.

AVt · Post by **AVt** » Tue Jun 17, 2008 8:53 pm

almost agreed, however a junior as Business Analyst will loose against the techies if not strongly supported it comes to decisions ...

rowdyroddypiper · Post by **rowdyroddypiper** » Tue Jun 17, 2008 11:00 pm

Nonius it may be complete overkill but to address your real end user concerns (i.e. disaster strikes, point and click to your exposures) you might be able to accomplish this using some form of OLAP to sit on the database. Mind you it's been like 3 years since I've tried to do anything with this, but Cognos was pretty big in the space and good when I did this last (now owned by IBM). They may also be able to help you design your database to optomize the reporting if you don't have internal people that are good in that front. Now I have no idea what kind of resources or what the scope of the project is and it's possible that some simple reporting would be sufficient.