[Beowulf] propagation velocity in cables

Vincent Diepeveen diep at xs4all.nl
Thu Sep 6 16:43:10 PDT 2012


Interesting is that some year and some ago, some tradeworx guys  
downplayed the importance
of speed in chats to me when i said speed is everything.

Seems they got convinced.

In a congressional hearing where also tradeworx was present,
a professor there proposed a penalty in time after a trade, to avoid  
trading in the same instrument (say a future Facebook or something)
too much, as several traders he checked out were trading 200 times a  
second in the same instrument.

In that congressional hearing the tradeworx person also opposed any  
form of artificial delay in trading.
The argument was that it would be bad for 'liquidity'.

We now see this is because of own interest.

Congress should do new hearings simply on this, instead of ask those  
who make big cash.

It's like asking Coca Cola whether allowing cola competition is a  
good idea.

I bet nothing happened sincethen in terms of rules.

You make 0% chance as a trader at home against the high frequency  
guys, and within high frequency land, you need to pay real big
cash to have the fastest links and fastest trading software.

Fair competition is simply not possible as bigger cash pays faster  
hardware and connections,
and as markets are so volatile thanks to politicians, basically the  
financial world is
doing a latency/hardware race that's pretty comparable with NSA.

Though it'll pay the salary of many HPC guys on this list,  
objectively this is not a good thing.

Yet world wide rules would be needed, as otherwise the traders will  
just hop to the next nation with their technology.

LSE now has an internal processing speed of trades of under 111  
microseconds roughly (just over 100 microseconds).

A link that's milliseconds faster than most competitors allows doing  
many trades in an instrument prior to majority knowing
what is happening.

In itself this is a form of knowing things in advance.

Not regulating this would be a major failure of EU + USA + other  
nations.

One needs slot times removing 'first comes first wins' and directly  
also avoid traders to trade X times in a row within too short of
a time trading in the same instrument. All this in order to reduce a  
big hardware race.

The hardware race should be in HPC and for the military, not in  
financial sector, just because they have the cash to do it...
In the end the clients of course pay all this.



On Sep 6, 2012, at 5:21 PM, Prentice Bisbal wrote:

> On 09/05/2012 07:43 PM, Vincent Diepeveen wrote:
>> The interesting question is that big satellite disk on the roof as  
>> well.
>> Microwave is simply civil allowed frequency i was told. Around
>> 2800Mhz or so?
>> How fast is such communication?
>>
>> Using a powerful laser would it somehow be possible to rival speeds
>> of these cables?
>
> The current issue of Wired magazine has a feature article on this  
> topic.
> If you look at the infogram on in the article, it compares different
> networks used for trading between NYC and Chicago. Right now microwave
> networks are the fastest. It's not a perfect comparison, because the
> microwaves travel a shorter distance than the wired routes, but
> microwaves still seem to be faster.
>
> It's interesting that the one microwave network, McKay Brothers,  
> set up
> their antennas in a "near perfect circle"
>
>
> http://www.wired.com/business/2012/08/ff_wallstreet_trading/all/
>
> Prentice
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin  
> Computing
> To change your subscription (digest mode or unsubscribe) visit  
> http://www.beowulf.org/mailman/listinfo/beowulf




More information about the Beowulf mailing list