[Beowulf] propagation velocity in cables

Prentice Bisbal prentice.bisbal at rutgers.edu
Mon Sep 10 08:21:17 PDT 2012

On 09/10/2012 10:29 AM, Andrew Piskorski wrote:
> On Fri, Sep 07, 2012 at 01:43:10AM +0200, Vincent Diepeveen wrote:
>> In that congressional hearing the tradeworx person also opposed any
>> form of artificial delay in trading.
> The guys at the (probably small) data and research shop Nanex point
> out that High Frequency TRADING isn't a problem.  The problem is high
> frequency QUOTING, aka quote spam.  Not working with HFT data myself,
> I don't know if they're correct, but their arguments sound very
> reasonable:
>    http://www.nanex.net/aqck2/3532.html
>    http://www.nanex.net/Research/Proposal/NanexProposal.html
>    http://www.nanex.net/aqck/2817.That
> (The connection of any of this stuff to COTS HPC is somewhat tenuous
> though, I'd say.)

It sounds to me like they're talking about the practice of "quote 
stuffing", which is essentially a denial-of-service attack on the 
trading system. If you do that on the internet in general, it's a major 
crime. Do that on a trading system, and it's just part of your trading 
strategy. Over the weekend, I finally read the "Raging Bulls" article in 
Wired Magazine. It touches on all the topics we've brought up in this 
broad thread.

The discussion of HFT to COTS HPC is somewhat tennuous, but I do think 
it's relevant. I work/live in the NYC area, and I've been contacted by 
numerous recruiters for HFTs. They're going after HPC guys like you 
wouldn't believe, since we have the closest skill set to what they need 
as admins. I also participate in an HPC+GPU meetup group in NYC. A large 
majority of the attendees are in the Financial industry, and I'm sure 
many are in HFT shops. So there is relevance, but maybe it's the other 
way around - HPC is relevant to HFT, but HFT is not relevant to HPC.

Someone earlier made a comment about the secrecy surround HFT, and that 
published numbers on system purchases are irrelevant. I can confirm the 
secrecy. I've seen registration forms for HFT conferences in NYC where 
you can check a box so that your employer is not listed on your ID 
badge, and you can even eliminate your last name from your ID badge so 
other attendees at the conference will only know you buy first name, and 
not know who you work for.


More information about the Beowulf mailing list