SmartQuant Discussion

We've been pretty quiet for a couple of months. Actually we've been working on the CLUB Server for VisualQuant. CLUB is CLoUd of Blocks. With the CLUB Server you can run a diagram locally in VisualQuant and send any block to a remote server with one mouse click.

What can we do with this?

- co-locate your strategy or application
- distribute your strategy or application globally to achieve the lowest latency. Fo example you can trade a cross exchange arbitrage strategy so that your startegy decision logic is running on a server in Moscow, one leg of your execution is co-located in Frankfurt and another leg is in Chicago. Why Moscow? Bacause it's in the middle on the map and you have about the same latency for both legs. This way you can be the first to build arbitrage spread and execute your orders. And after all you can control all this from any place on the globe.
- backtest your startegy on a powerful remote server
- distribute your strategy logic among several servers to improve backtesting speed
- optimize your strategy on a cluster of servers
- etc. etc.

Here is a simple demo video

http://www.smartquant.com/visualquant/v ... Server.wmv

and about the same but with two servers

http://www.smartquant.com/visualquant/v ... erver2.wmv

Cheers,
Anton

Nice Idea.

Interesting. Do you plan on offering some sort of colo-service along with this, or is that something every user will have to do by himself?

I think this is a question for QuantHouse

what protocol do you guys use for inter server communication?

Currently it's .NET remoting but it can be any transport. For example QuantHouse has a superfast event bus with their FeedOS...

Dr. Anton,

Do you have any documentation for CLUB Server other than the two videos available? I downloaded the beta and entered in a server, wasn't sure what the format for the connection string (i tried it by IP) and as soon as I clicked ok after adding a server the application immedietely crashed. Excited about this product, it is designed to handle farm level tasks like redundancy, optimization in addition to distributed management of code correct?

Hi,

we will add some docs next week. Yes, I also see that the software carshes if you enter wrong connectiong string. This is not very nice indeed. We'll fix this.

Meanwhile I think VQ should work with the CLUB out of the box if you install both on the same computer. No need to change connecttion string in this case.

PS. Currently there is one limitation. Blocks with property inputs can not be sent to the CLUB. This will be improved.

We are also working on a programming model where you can use blocks as normal classes and yet be able to build a globally distributed application by simply telling a block that it should be run on a server. Something like


I think this should be really cool

Regards,
Anton

Hi Anton!
Great features you are working on, really great job.
Does your team plan to create a server application for CLUB technology. What I meant by server application? Just imagine me having my personal servers, lets imagine I have 4 of them, connected to the local network. I want to run CLUB technology locally, without any information sent to the Internet servers, or more complicated I want more distributed app I would like to have my servers and Internet servers work at the same time solving the same problem engaged in backtesting/optimizing my solution. So have you already included these feature, if not are you planning to make this one possible? thx.

Hi,

I am not sure I understand your question correctly, but CLUB server is a server application. You can deploy several CLUB servers locally, on your LAN or internet and submit blocks from different VisualQuant applications to be executed on these servers.

Though perhaps you are asking something else?

Cheers,
Anton

Anton, yes you have shed light on these for me (Actually it wasnt the answer I wanted to hear, it made me feel puzzled and I rewathched the video again and I found the answer by myself, so basically speaking it have helped me) however I have another questions
Imagine yourself having three different computers, connected into one network - they are having different computation resources. Your aim is to run strategy optimization/backtesting using all your computation resources of these three computers with a maximum outcome. Is it possible for you technology? Do you have a special computation resources manager who will sending tasks to three computers with maximum computation efficiency?

If yes to make matters worse all your computers is connected not via local network, but via Internet, so all computers are distributed throughout the net. Will the delay between packets sending/receiving influence overall computation capacity? Do you have special manager for solving computation/execution problems that will choose among computation resources which combination of resources will be the most efficient for the given optimization/backtesting task?