Hacker News Reader

Comment by chollida1 | Hacker News Reader

> Some facilities such as the Mahwah, New Jersey, NYSE (New York Stock Exchange) data center have rolls of fiber so that every cage has exactly the same length of fiber running to the exchange cages.3

I remember the first time someone told me that. I thought they were kidding. Then I actually got to see the data center.

The exchanges, and HFT firms, take this very seriously. The speed at which they execute is just unbelievable.

The amount of technology and networking stack knowledge is just mind blowing.

I remember tuning my linux distribution's networking stack to gain an advantage only to find out that everyone had moved onto using asic's and infiniband hardware.

I continue to be amazed at the sheer talent and brain power that goes into this field:)

View on HN

Spread Networks invested an estimated $300 million in building a new fiber line from Chicago to New York in order to shave 3 off ms. However, they found out after completing the line that faster speeds were already being achieved by microwave connections built in secret. It's an interesting arms race:

http://www.cnbc.com/id/100695563

Kernel bypass is also common. Having the OS manage your networking stack adds too much latency.

That is the most common one I've heard of.

Obviously if they are building fpgas into the switch itself then well of course they'd already tried running customized ethernet driver firmware and all.

It really depends on what asset class you are trading, like the author mentioned. For equities the performance demands are out of control (< 50 micros). For FX, it's still pretty tight but much more manageable.

How do I get into this field? I'm a 20 year old making a shit ton of money by doing mobile work but I find it so boring. Nothing about it is innovative. I feel like a high level computer construction worker. It's brainless for the most part, it just takes time.

I want to get into some really cool stuff. I've been wanting to get into investment banking from a technical standpoint, HFT, algorithmic trading etc, but honestly don't know where to start.

What are the de facto books/websites/reads for this field?

The worlds of HFT and the rest of algorithmic trading are very far apart. For HFT pickup the exchange documentation for each exchange. Memorize how order types. Get a copy of RegNMS. Memorize that as well. To be effective an HFT needs to exploit the market structure. Learn how processors work. How do you avoid a pipeline stall or a cache miss? Become an expert in c++. Maybe pickup some verilog for those pesky FPGAs.

The rest of algorithmic trading offers a lot more flexibility. In either case read 'Trading and Exchanges'. It will give you a good overview of the mechanics of all different venue types. Then its thinking about how to build better algorithms then the next guy, or get better data.

I second the recommendation of 'Trading and Exchanges'. It's a great book to get insight into how the underlying exchanges work and how to exploit certain architecture to produce profitable strategies.

> How do I get into this field?

Well turn the question on its head and ask what can you bring to a hedge fund to allow it to be more successful. That's the test we use.

Hedge funds are similar to start ups in that they are small businesses where everyone needs to pull their own weight and be able to handle their own area without supervision as everyone is busy enough handling their own area of concern.

Or put another way, say the firm has 10 people and each share equally in the year end bonus pool. If the pool is $100 then each person makes $10 in bonus.

Therefor you must be able to increase the bonus pool by atleast $10, to $110 in the next year or two, in order for the firm to break even by hiring you.

So what do you bring to the table that allows the firm to increase its profits? how good are your networking, math, C or C++, optimization skills, what market structure insights do you have?

Thanks for the answer. I meant more of an absolute beginner in finance, in general, nevermind the stock market or trading or writing bots or hedge funds. Thanks for the insight.

For HFT the learning requirements are extensive.

It is /almost/ essential to have a PhD in CompSci/EE from a top school to do HFT. Alternatively one should demonstrate extensive hardware/networking and optimisation skills obtained from other low-latency industries.

All of the top work is being done on FPGAs and latency is now on the order of microseconds (probably lower).

As for lower frequency algorithmic trading, that is a game that one can play at the 'retail' level if you're willing to spend (quite a lot of) time learning.

I run a website about algo trading. If you want to get a taste for what is involved have a look at some of the articles here: http://quantstart.com/articles/#algorithmic-trading

You run quantstart? Haha awesome, thanks for commenting. I've been familiarizing myself with the content on your website. Thanks.

Feel free to send me an email (mike@quantstart.com) if you want any other help!

I will for sure!

I hope "amazed" is just poor wording.

I think I'm being trolled but I'll bite.

What's wrong with the word amazed?

If I said I worked at Google and I continue to be amazed at the tech behind how they served up ads would I still get the same remark from you?

Maybe I could be doing more with my skill? I don't know but I really love the learning curve I'm on working with cutting edge technology and pushing the performance envelope.

>What's wrong with the word amazed?

Amazement (dopamine rushes) confuses us.

Biologically, we respond to amazement as part of the learning process. But all too often, we receive these dopamine rushes simply as a consequence of the amount of resources and effort someone (or a group) puts into something.

e.g. You can go to Las Vegas, see thousands of people running around topless in gorgeous outfits and be amazed, yet what's the actual value to you?

I see this ALL too often in art. People get "amazed" by things that require a great deal of effort/resources (look how many transistors in this CPU!), but this temporary narcosis blinds us to evaluating /whether it's a good idea/. So many art projects are simply the result of an ordinary person commanding extraordinary resources from a corporate entity to construct a large number of instances of a single, mundane object or a single, large-scale version of the object. In these cases, the art is not in the sculpture, but in the conning to get the resources. I see budding artists respond to this amazement and trying to get huge funding.

Internet bloggers (Cory Doctorow comes to mind) abuse the word "amaze" all the time. I just checked boingboing.net right now and saw amazing used to describe wallpaper (no dude, this is /natural history/ wallpaper).

I realized that there's a frequent uproar against HFT, so I assume people here are pretty tired of it and are not willing to discuss it/have their minds made up about it already. That's why I won't go rhetoric.

edit: I guess, I did though.

I can't help to reply to the adwords question - although you can argue that advertising/marketing is soulless, at least it creates real world value. Someone makes a product, someone else buys it. Exchange of goods happens.

Derivatives are uncontrallable in the my humble opinion. Especially if regulated by uber greedy people. And it makes me sad sad when I see the highest skilled people go and put their talent in something that is pretty much useless, apart from making a few, selected people even richer than they are. But I guess, whatever pays the bills.

But that's not why I made my comment. I made my comment, because what I find even more sad, is people going numb to injustice and admiring this situation and the people who are making it possible.

He wasn't admiring the economic situation, he was admiring some truly amazing high-end computer networking.

If HFT produces nothing else (and I generally agree with your economic view of it), at least it'll have financed the next generation of low-latency networking gear.

So no one would buy anything if there was no advertising?

Or maybe you are worried they would buy less? In that case, would the world really be a worse place? People only buying what they need...

Car insurance is a derivative.

I think much of the effort is wasted. When you hire embedded designers to program FPGAs and ASICs just to be a few nanoseconds faster in your rent-seeking, it's a bloody shame.

Technology is making the existing system much more efficient, lowering the total amount of "rent-seeking" in the system.

If this was any other industry it would be lauded. I don't understand the moralizing about this industry while we applaud start ups for building ever more complex twitter aggregators.

Last time I checked, twitter programmers don't have a record of destroying economies nor deep capturing regulators.

What economy did low latency trading destroy? Also, is there any more deep capturing of regulators in the finance space post low latency trading than there was pre?

This is an ignorant statement. The cause of the economic crash in 2008 was leveraged bets on lots of bad debt. These bets were made by fratboys in suits, shaking hands and making phone calls and sending emails. The financial instruments they used to do this were bespoke one-off deals, not securities traded on public exchanges.

High frequency trading had nothing to do with any of this. HFT deals in fungible instruments traded in public markets.

But if you do it to mine bitcoins, you're a visionary.

Reply on news.ycombinator.com