Why is it that ALL brokers do not understand true latency?

"1. short without checking inventory (they are market makers).

2. unlimited buying power so no checking for buying power cuz they can.

Latency, well they eliminated that latency and have only.

1. quotes- depending?? uncontrollable* but just say 40 milscecs
2. executions: ecn's, market makers-(most likely using fiber) , and specialist) about 0-40 milsecs ave probably lower on most ecn's, nothing if market maker themselves. Listed is higher if you use true acks.
3. back up databases if that-depending? controllable*

In all, they can come in around 0-80 milsecs. Depending on the quotes. Not bad.

"

If I came in at 80 milliseconds, i'd be out of business................
 
doubt it, exchanges are in coming around 100.



Quote from sigsegvboogman:

"1. short without checking inventory (they are market makers).

2. unlimited buying power so no checking for buying power cuz they can.

Latency, well they eliminated that latency and have only.

1. quotes- depending?? uncontrollable* but just say 40 milscecs
2. executions: ecn's, market makers-(most likely using fiber) , and specialist) about 0-40 milsecs ave probably lower on most ecn's, nothing if market maker themselves. Listed is higher if you use true acks.
3. back up databases if that-depending? controllable*

In all, they can come in around 0-80 milsecs. Depending on the quotes. Not bad.

"

If I came in at 80 milliseconds, i'd be out of business................
More...
 
Quote from sigsegvboogman:

Nitro:

"I never suggested you are full of shit - only that you are worrying about things that you could not possibly take advantage of, and that the people that can you aren't helping them because they know what the real issues are.

nitro"

I am the people who can take advantage of the low latencies. I have for years and years. The reason I have survived is because I have eat, breathed and slept these issues. The end result has been survival and profitability. I also know the end result of loosing speed for all of my algos: It equals less to money and failed trading because I am late. Whether it be marketmaking, directional decision making, everytime I take step back in latency by an amount smaller than most people see, I see the end results. I am at the point where I know roughly how much more moeny I make with every certain "increment" of latency reduced. That knowledge has been pretty solid for almost 6 years now.

With regards to IB:

Nice guys, but they have always been too slow for my needs. Mike Domka has always been a champ and a very nice person to work with exploring if we could have worked together. So far they have been too slow. Now, that doesn't mean they are slow for other high speed guys, just for me based on their stats and my needs.

Wow, just got back. Thank god I got out of New York early to drive home with all of that mess. I have alot to read and catch up on here.
More...

Sig:

Since you've done a lot of work on latency, can you share some of your knowledge with the guy who may have a store-bought computer box, and a dsl line in his house? What is the best way to shave milliseconds, is it in hardware or software? My current ping time is 40ms. Trades are manually entered (by exchange rule), but other traders try for the same opportunities, so speed is still important.
 
Well, I think we are comparing apples and oranges here. I think the most latency in your execution path will be your finger clicking on the mouse.

You more than likely are using Windows for an operating system, so there is an inherent latency there no matter what. You can do things in Windows to speed that up (like disabling the paging file or setting to "best performance"). Just make sure you have 1 gig of RAM or more to do so. At that point, try to get as less hops as you can on your line: However that may be attained.

It's not really technical, but from your perspective, those are some ways to do so.

Like I said, what I do and what you do are like comparing apples to oranges. Best of luck and Happy Holidays.
 
Quote from JMartinez:
2. unlimited buying power so no checking for buying power cuz they can.

A little off topic, but...
I think this is a myth. It's unreasonable to think that any firm that wants to stay in business for more than a minute does not have risk management systems that will prevent it from exceeding its desired exposure for a given asset class, instrument, trader, or aggregate firm position.

That's not to say that there aren't exceptions (like Mizuho :) ).

Even still, I know that the amount of time spent checking my own risk management params in my system is, without question, completely insignificant.
 
No myth. Blow ups are myths too :) :)




Quote from alanm:

Quote from JMartinez:
2. unlimited buying power so no checking for buying power cuz they can.

A little off topic, but...
I think this is a myth. It's unreasonable to think that any firm that wants to stay in business for more than a minute does not have risk management systems that will prevent it from exceeding its desired exposure for a given asset class, instrument, trader, or aggregate firm position.

That's not to say that there aren't exceptions (like Mizuho :) ).

Even still, I know that the amount of time spent checking my own risk management params in my system is, without question, completely insignificant.
More...
 
Quote from sigsegvboogman:

You more than likely are using Windows for an operating system, so there is an inherent latency there no matter what. You can do things in Windows to speed that up (like disabling the paging file or setting to "best performance"). Just make sure you have 1 gig of RAM or more to do so. At that point, try to get as less hops as you can on your line: However that may be attained.
More...

And yet I rarely see any trading system using a true real-time operating system like embeded Linux or QNX or the like. Even under the regular Linux the clock resolution (aka the semaphore minimum time) is still measurable, it is better than the "default" windows setup (which has atrocious clock resolution, crap like 10 to 25 ms). Don't get me wrong, I am an old UnixFart (since '82).

The interesting thing about tuning Windows for better clock res is that a lot of game developers does it very well (see Doom, Quake, etc). So one can (and I have) taken cues from custom video and DirectX apps (!) to get windows for much finer resolution. The amazing thing is that the driver refresh rates (roughly corresponds to minimum clock res) under Windows can be better than a "stock" Linux setup, and is about the same (since we then run into hardware) as a custom tuned Linux kernel. Task management in Kernel releases prior to 2.4.x was sub-optimal anyways, before Cox rewrote all of it (I think it was included in 2.4.x kernel release, I don't run a stock kernel, so kernel versions I have lost track).
 
You must log in or register to reply here.
Back
Top