Tag Archives: ultra Low Latency

FPGA options for ultra low latency HFT with stunningly convincing Youtube video with Algo Logic with some technical papers

FPGA options for ultra low latency HFT with stunningly convincing Youtube video with Algo Logic with  some technical papers

These highly useful tips came from a very forthcoming Youtube subscriber so many thanks to them for presenting this. The mentioned YOutube video below is one of the best reasons you should go with FPGA if you are very serious about ultra low latency in your HFT solution.

Join my FREE newsletter if we will ever go FPGA 

Hang on, I got the paper:

Click to access Low_Latency_Library_for_HFT-Algo-Logic-4831a009.pdf

This is some interesting information.
The guy in the youtube video is also the author of this paper. He runs a company called “Algo-Logic”, which is similar to Cheetah Solutions. They provide an end-to-end FPGA solution with an included proprietary development platform in addition to the hardware and support.

He also has some technical papers on the subject:

Click to access asic94.pdf

Description of a high-speed parsing engine:

http://www.arl.wustl.edu/projects/fpx/references/Content_Routing-HotI-2006.pdf.

Hardware clustering algorithms:

Click to access FPL2006-Clustering_in_FPGAs.pdf

Semantic data:

Click to access HardwareSemantic-Aero_2006-10_0802.pdf

Sounds more interesting and cutting edge than developing a fix connector with no performance guarantees for a buggy tradelink, if you ask me.

Also, I found out that TT is not the only company that offers integrated market data over fix+execution to retail traders. CTS T4 also has a fix api with both market data and execution.

 

 

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

More useful FPGA options for ultra low latency HFT with stunningly convincing Youtube video with Cheetah

 

More useful FPGA options for ultra low latency HFT with stunningly convincing Youtube video with Cheetah

These highly useful tips came from a very forthcoming Youtube subscriber so many thanks to them for presenting this. The mentioned YOutube video below is one of the best reasons you should go with FPGA if you are very serious about ultra low latency in your HFT solution.

Join my FREE newsletter if we will ever go FPGA 

 

 

FPGA Solutions
I found an article behind a paywall that you might find interesting:

http://www.computer.org/csdl/proceedings/hoti/2012/4831/00/4831a009-abs.html
“A Low-Latency Library in FPGA Hardware for High-Frequency Trading (HFT)”
” The application sustains 10Gb/s Ethernet line rate with a fixed end-to-end latency of 1µs — up to two orders of magnitude lower than comparable software implementations.”

Here’s a company that sells fpga solutions showing a model of their platform:

http://www.cheetah-solutions.com/pages/ovw/cf.html

There are apparently already a few companies like Cheetah Solutions and Impulse software that offer full end-to-end solutions for FPGAs. Of course, by using industry standards like matlab/simulink and going directly to the mfrs like Dini, you can cut out the middle man.

If this works out, then the latency of tradelink, or any other software platform for that matter becomes irrelevant. And FIX would already be optimized for the fpga.

This really is cutting edge stuff, most of the material on this and the companies that have sprouted up as a result have only come up in the past 1-2 years.

I found on video this subject that compares software solutions to fpga hardware based solutions:

The slides for the presentation:

http://hoti.org/hoti20/slides/Lockwood_AlgoLogic.pd

More useful commercial FPGA options for ultra low latency HFT with stunningly detailed Youtube video with Cheetah

These highly useful tips came from a very forthcoming Youtube subscriber so many thanks to them for presenting this. The mentioned YOutube video below is one of the best reasons you should go with FPGA if you are very serious about ultra low latency in your HFT solution.

 

 

 

 

 

 

 

 

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

Retail Multicharts for HFT? Other Options including FPGA for ultra low latency

Retail Multicharts for HFT? Other Options including FPGA for ultra low latency

From the same previous Youtube user:

Also, for the FPGA solution, I found this:

http://eecatalog.com/chipdesign/2012/10/20/new-fpga-based-nic-features-lowest-latency-and-lowest-price/

The Dini Group has a few boards that are apparently optimized for FIX:

http://www.dinigroup.com/new/DN_FBSP.php

Starting price is 5k.

–> Big thanks for this note which is very useful

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

Use of Matlab Simulink with FPGA hardware option for ultimate ultra low latency HFT option if you want DMA

Use of Matlab Simulink with FPGA hardware option for ultimate ultra low latency HFT option if you want DMA

FPGA Hardware

Altera

Xilinx

This could be the best option but do understand you don’t get the slickness and support of a powerful platform like Deltix. The hardware can be very expensive at $3k per board. To develop models is through Matlab Simulink which can be code generated to FPGA. This process could be exhaustive as compared to Deltix but you have ability to fully customize your own algos/models/strategies. Also, you will need to spend another $20k for ModelSimulator to validate models before they are synthesized onto the board. If you are looking for the absolute lowest latency, FPGA is the way to go. There will be lots of self learning as compared to lack of resources through a platform like Deltix. Also, full licensing costs of Matlab with Simulink and Simulink could reach $20+k. FPGA is really useful if you want to trade in a Direct Market Access arrangement with is another very expensive option but this is what the big boy operators do. We need to consult with a vendor on these options.

Read our FREE newsletter to see if we go with this route? 

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

For true ultra low latency HFT, the language wars is a case closed with CUDA and C or C++, sorry Java and Python lovers

For true ultra low latency HFT, the language wars is a case closed with CUDA and C or C++, sorry Java and Python lovers

UPDATE: This posting is now irrelevant!! CUDA does support Java, Python, and even Matlab. Can you say DOH? Don’t bother reading this…

If you are planning to do any kind of GPU, the only one that really matters is CUDA. Don’t bother with OpenGL as the control seems to be better under CUDA. The pain of OpenGL makes me cringe. Where’s the Tylenol?

Follow what I do with my free newsletter

As for the programming language wars, there is no further debate as the only two languages natively supported for CUDA is C++/C or Fortran. End of story. This could change but …

So what does that mean for all those Java, Python etc lovers, are you seriously going to add a potentially bug ridden API converter with an extra layer to add potential latency to your HFT system.  Now that would be dumb. Plain and simple, you should be out of a job if you go that route.  If you worked in my operation, I would fire you.

If you care about affordable and easy way to develop true HFT platforms, go CUDA with C++. Again, there is no further debate on it so case closed.

Also, here some more references :

For Java use http://www.jcuda.org/

ForPython and its ilk:

http://stackoverflow.com/questions/2947211/can-i-program-nvidias-cuda-using-only-python-or-do-i-have-to-learn-c

For true reaction including myself:

Quote from amazingIndustry:
please do not tell anyone you are running live trading strategies in R, Matlab or Python, people who know this specific business will laugh at you


What exactly is “this” specific business? If you are looking at stuff tick-level, sure, go with C++. If you care about dynamics of implied volatility or simple technical indicators in a non-latency dependent way, why would I want to be f*cking around with memory, pointers and virtual functions? As I said, there are multi-billion dollar funds that are running fully in interpreted languages.

http://www.elitetrader.com/vb/showthread.php?s=46c3936f02cbf4fd246ec4069306d44d&threadid=246526&perpage=6&pagenumber=6

Read this comment:

I believe that, with PyCUDA, your computational kernels will always have to be written as “CUDA C Code”. PyCUDA takes charge of a lot of otherwise-tedious book-keeping, but does not build computational CUDA kernels from Python code.

share|edit|flag answered Jun 1 ’10 at 4:42Alex Martelli
232k17359714
 
Indeed, though there are interesting projects that do; see my answer. 😉 – dwf Jun 20 ’10 at 0:54
We use PyCUDA at work and get good results with it. The advantage of writing plain CUDA C is that you can get help and snippets from other people without translating from your API specific python code. We use PyCUDA at work and get good results with it. The advantage of writing plain CUDA C is that you can get help and snippets from other people without translating from your API specific python code.

 

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

Last Week to Register: Join the Discussion and 150+ of Your Trading Colleagues Already Registered for the 2nd Annual Ultra Low Latency Evening Event

Last Week to Register: Join the Discussion and 150+ of Your Trading Colleagues Already Registered for the 2nd Annual Ultra Low Latency Evening Event – New York, May 24th, 2011

May 24, 2011 – Westin Times Square New York – 5:45 pm Registration

Exchange consolidation. Co-location issues. Rate of connectivity and the search for ever lower latency. This environment the trader finds himself in, with the need for speed driving the latest innovations, and the government engaged in push-back, is challenging and one ripe for discussion. The application of algorithms in the high frequency space has been addressed in the equity market but what’s the next area of growth? What developments are occurring in FX, Options, Futures, Commodities and the other asset classes?

Our evening panels address issues including: Winning the speed game while remaining compliant: Regulations, changes in the markets and its impact on trading in FX and other asset classes; Fragmentation, liquidity, connectivity and the markets; the impact of HFT on the markets; Build, buy and beyond: selecting and implementing the best tools to achieve a successful multi asset strategy in trading; Assessing and addressing risk and What’s the next phase and the next “hot” area?

Join us for an evening discussion, one of the series of High Frequency Trading events produced by the Mankoff Company, and learn from your colleagues in the industry. Our panels bring together trading experts, leading practitioners and innovators in the field who bring real-life experience to the talks. Learn why over 1,000 of your industry colleagues have attended The Mankoff Company events which combine applicable “top of mind” discussions with advanced networking opportunities.

The panels are immediately followed by a networking reception allowing for further engaged discussion on the subject matter.

To register and for more information on who’s speaking:http://www.themankoffcompany.com/FXNYCHFT/

 

 

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

FPGA Ultra Low Latency Risk Solution: A ULLINK Live Demonstration Breakfast Event

FPGA Ultra Low Latency Risk Solution: A ULLINK Live Demonstration Breakfast Event New York, May 25th at 8:00am

For registration and details, please e-mail stephanie.fields@ullink.com.

 

 

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

Some great ideas to achieve low latency and ultra LL for arbitrage, decrease LL on networks

Some great ideas to achieve low latency and ultra LL for arbitrage, decrease LL on networks

Here are some great tips on how to decrease low latency and ultra low latency for arbitrage. This also helps decrease low latency on networks.
If you can afford go for Cisco Nexus devices at data Centre with Blade Server from Hp.Best Hardware platform workers on 10 GB network.
You should have a look at our Storage technology cards, we increase database performance from 4-10x, networks are important but the storage is the biggest bottleneck in any networked storage computing environment. Have a look at www.fusionio.com

As a Country Manager for TNS you probably already know most of this, but for arbitrage the keys are the best available dedicated links between the exchanges and the carrier access points, and then the best routes within the carrier networks.
Fast networks, servers and infrastructure are all very important, but once in place how do you measure all of that? How do you keep both your vendors and your IT folks honest? Perhaps you look towards a hardware based solution that enables you to capture all of your packets (market data, trade excution, network, etc…) in real time, down to the nanosecond level, with 0 packet loss. In addition, what if you were given the ability to replay that information anytime you want for application testing, benchmarking, etc…

I’m using VPS for a long time, and it’s great to me. It can reduce network latency and provide me with very good stability.

But currently, I’m a little worried about the privacy problem that the VPS provider may steal our trading system.
Interesting stuff huh?? It is all about the best quality in hardware for network cars, servers, and being as close to the exchanges.

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!