Broadly speaking, the diary seems to be saying that Goldman might have been front running. That they developed a computer program that gained access to information on trades before the trades were finalized and then used that information to place their own trades.
2.
bago
THEY FUCKING CONTROLLED THE TRANSACTION_COMMIT CODES?
With unencrypted transaction control identities in the packets and hardware control, that means you can delay the commit phase of a transaction.
This means if someone asks to do a trade, you can halt the trade, do your own trade, and then have the results of that trade change the price of the original trade, and possibly eve cause the original transaction to timeout, rolling back the entire trade!
It’s grand theft exchange!
The only way you could pull this off without the possibility going to jail for a very long time is if you altered the timestamps on the transmissions. The listed superused ability to read process memory can give you that, depending on if you have write permissions. Of course you own the boxes, you can give the other processes lower cpu priority to ensure that you will have time to make the writes you want to cover your tracks.
Of course if the transaction information is unencrypted and you own the boxes, you can do any number of timing tricks beyond the one I listed above.
At any rate, anyone who is packet inspecting transaction data is a fucking felon. It’s a classic man in the middle attack. The idiots who didn’t encrypt are failing, but the evils exploiting them are violating the entire principle of exchange, and need to be put in a federal pound me in the ass prison.
3.
HeartlandLiberal
In a nutshell, Goldman Sachs has been shown by recent in depth article by Matt Tabbi at Rolling Stone to have been manipulating the roller coaster of bubbles in the American Economy since the Great Depression:
Now it is becoming clear that Goldman Sachs for the past decade has been running sophisticated computer programs doing quick response and rapid programming to achieve programmed trading that has been peeling of zillions in profits. But what this also represents is blatant daily manipulation of the markets for these profits.
A Russian programmer figured out how to do this. So then and only when their long secret manipulation, which has been covertly allowed and kept under the rug from public view is being exposed, are the powers that be scrambling not to expose Goldman Sachs for the evil they have been doing, but to sweep the whole sordid mess under the rug before we, the plebs and investors who are being ripped off by massive market manipulations, figure out what is going on and rise up and put an end to these practices.
It really is just one more case of exposing the dark, criminal under side of what has evolved in American corporate and Wall Street culture. And the sooner they can convince you not to look at that man behind the curtain who is screwing you blind, the better for them to retain their power.
4.
DecidedFenceSitter
*IF* I’m reading correctly, here’s my understanding.
In information security there are three key factors Confidentiality, Integrity, and Availability. Confidentiality – Only the intended senders and receivers can read the message. Integrity – The message remains unchanged. Availability – The system is available when needed.
The key here is confidentiality, the key line from the article is from the introduction, “is technology to grab off FIX PROTOCOL, OCX, or SWIFT messages that precede every transaction_commit at the Exchanges.” Basically, it appears that thanks to no encryption, “Encryption is optional or nonexistent or easily enough made transparent. NYSE has any and all access codes involved.” And we’re discussing split second transactions, which means if you can read the mail, you know what your competition is doing.
What the STORM patent does (if I’m reading it correctly) is allow for the real-time reading of the packets passing through the servers to route them. This is required to get the achievable results of the FIX Protocol, OCX, SWIFT messages.
Key take away for the non-financial, no security folks: “Access to FIX, OCX or SWIFT messages prior to transaction_commit at the Exchanges would give a player an advantage parallel to seeing an opponent’s cards in a game of poker.”
5.
DecidedFenceSitter
My comment is in moderation, but unsurprisingly enough, we’ve got a few folks who get computer security on an online blog. :)
6.
HeartlandLiberal
To expand on my comment above, this from another diary at DailyKos on this issue:
The abbreviated version of the story, in a few sentences: A senior technology strategist and Vice President at Goldman-Sachs, Sergey Aleynikov, copied much of his firm’s “secret sauce”–an extensive set of proprietary, automated stock trading software code and algorithms–all related to “program trading.” Upon finding out about this, senior officials at Goldman-Sachs informed the FBI of all of this and had Aleynikov arrested at Newark Airport on July 3rd.
The code, as it has been noted by many, including Goldman-Sachs, allows the firm to execute securities/commodities transactions in microseconds, thus providing their company with an extreme edge over their competitors. The tacit fact is, with proper monitoring of market trades, in general and as facilitated by Goldman’s own practices, it’s entirely conceivable–albeit significantly questionable from a legal standpoint–that the firm would be enabled to “frontrun” its competition at quite a grand scale, too, since it could see trades occurring in real-time, and then execute its own trades automatically at lightning speed, before the previously-observed trades of others were even concluded.
Meta P.S. Blockquoting is still broken in this interface. I could not enclose the above two paragraphs in one blockquote tag set. And I did NOT bold the first paragraph. Whatever CSS code that is parsing the HTML is doing that all on its lonesome own, not me.
7.
bago
@HeartlandLiberal: In layman’s terms, it’s as if they look through all of the letters asking for trades on the exchange because they own the mailroom, write new letters making those trades, send them first, and then continue to send the original letter to the trading floor.
8.
Lesley
@bago
Wow, then it would be even worse than plain vanilla front running. I didn’t understand the parts about controlling the transaction_commit codes. So they might have been getting information on the trades prior to execution and then delaying execution of the trades until their own trades were placed.
9.
bago
@Lesley: It could be. You can screw with a lot if the assertions in the Kos article are true. Depending on the environment and the encryption, you can do a lot, and the notion of superuser accounts running production code should scare the bejeezus out of you.
10.
Lesley
OK, correcting myself. Usually front running is using data from your client’s trades, so you do control when the trades execute. In this case, it sounds like they could have been using data from and delaying everyone’s trades, not just their own clients.
11.
bago
@Lesley: To be more accurate, they were using data from any trades that passed through a system they owned. Slowing down competitor trades just makes that process more effective, and may have been possible. Again not enough details to be sure, but based on what is known, you can do message inspection into competitors trade commands in real time, possibly fast enough to beat exchange transaction times, thus you can get the good deal that some other trader saw as fast or faster than your competitor, essentially stealing the competitors trading knowledge.
Not just using data from your clients, but from anyone else on the system you were running.
If the article is true.
12.
Lesley
@bago
Sounds to me like they’ve tacitly admitted the program has the capability to do more or less what was described. They admit it could be used to manipulate the market, so it must have at least some of that functionality. I guess it’s down to whether we trust them not to have abused that functionality.
And it does scare the bejezus out of me.
13.
Karmakin
Here’s what I got out of it. If true, GS is observing the trade stream in real time, pulling out the basic info from it. If they see a large buy order, they jump ahead, buy some of that stock then resell it at a profit to the person who put in the buy order. If they see a large sell order, they then short the stock.
Basically what they’re doing is they’re skimming money off the top.
That said, I’m a person who thinks that the public market system is a fraud as a whole and actually fights against what it is intended to do (provide capital). So I’m probably too cynical on this.
14.
PeakVT
Sounds plausible. I think the biggest problem would be making the timestamps look realistic. If GS is frequently making a trade X milliseconds before another larger trade, that pattern would show.
Shorter Kossack Diarist: Goldman Sachs appears to be doing a high tech version of front running the entire market.
16.
bago
@PeakVT: Hence my comment about superusers and timestamps. Again, depends on lots of precise technical information such as the operating system, the permissions that each process ran in, what was shim’d out to who, logfiles, disk permissions, etc.
Without access to the machine configurations, you can’t tell if it is possible. Without access to the logs and binary you can’t know if it was actually done.
However the author seems to speak in a manner that indicates that this is a possibility, and should be investigated.
If it is true, then this means that Goldman appropriated itself as a semi-governmental entity and has been taxing everyone. Even though half of the treasury department is ex-Goldman, to borrow a phrase, “Don’t steal, the government hates competition”.
17.
Walker
If this is true, it is grounds for a class action lawsuit from every single person who has bought a stock in the last decade.
18.
inthewoods
I think it’s a lot of smoke and no real fire. There’s no proof, broadly, that anything has happened. That doesn’t mean it hasn’t, but the current evidence is weak. You’ll need stronger stuff (that may well exist) to take down GS on something like this.
The Kos article does come across as a bit 9/11 tinfoil hat imho – but hey, that’s just me.
19.
Redhand
It’s fodder for a massive insider trading investigation and criminal prosecutions. But, is this yet another blatant securities law violation they’re going to get away with, . . . because they are GS?
20.
Walker
I think it’s a lot of smoke and no real fire. There’s no proof, broadly, that anything has happened.
Reading the article more thoroughly, and the comments, I am starting to believe that this dairy is a bit premature. It needs more proof.
We have known for years that GS has been investing in high throughput data dissemination software. It is possible for them to get a leg-up (e.g. get the information first) and do all of this simply because their software was better. If they actually had a software advantage, it would be stupid to risk it on something as damaging as reading before the transaction commit.
21.
Napoleon
If they actually had a software advantage, it would be stupid to risk it on something as damaging as reading before the transaction commit.
Lets say they are reading it after transaction commit, would everybody else in the world be the the same position to take advantage of that fact?
22.
Walker
Lets say they are reading it after transaction commit, would everybody else in the world be the the same position to take advantage of that fact?
If they had the software to read it fast enough, yes. But it is unclear that anyone else was spending the money on this that Goldman Sachs was.
Goldman Sachs hires A LOT of high performance computing, distributed systems, and data management PhDs (several of the students in my group have been offered jobs). They are the Google of the financial world. All their software is developed in house. Any attempts to sell them third party software (e.g. Stonebraker’s Streambase) have failed because they believe their proprietary stuff is better.
Any start-up can hire a quant, but you need a critical mass to build a competitive software financial infrastructure.
23.
jwb
One thing I’m confused about is the patent. What nutcase patents something illegal? So I have to believe that at least what’s described in the patent is perfectly legal.
24.
Walker
One thing I’m confused about is the patent. What nutcase patents something illegal? So I have to believe that at least what’s described in the patent is perfectly legal.
The comments on the Kos piece say that the patent is for routing in voice networks. The company IDT is just a telco; STORM was not designed for financial trading. Only its implementation in a financial system would be illegal. The patent was developed at IDT, not Goldman Sachs.
This is a highly suggestive coincidence, but right now there is nothing more than that. There may be fire here, but we need more evidence.
25.
Bill Arnold
Basically, seems to be accusing GS of engaging in very high speed automated insider trading, or worse, with no actual evidence. Just (1) it’s possible (2) Aleynikov has some skillz that are hypothetically applicable (3) GS made a pile of money. (There may be (4) etc in there as well.)
(I am not familiar with this domain, but … the volumes are extremely high and milliseconds matter (the early bird gets the worm), so there is a tradeoff between slowdowns due to encryption/decryption/signature verification, and speed gained by trusting the intermediaries.)
The diarist is alleging that Goldman hacked the NYSE. If true, it’s the biggest securities crime ever.
27.
A Cat
I always wondered what GS needed a sub 15ms response time and this explains it.
Assuming the network traffic was over a stream based protocol, not packet based, a sub 15ms response time would allow plenty of time to grab those messages and decode them and get a response out all before the final message saying the trade had been executed hit any one elses servers.
Having read the full article at KoS, I have to say some people at the NYSE and their members need to be fired. Colocating your servers in a NOC where your members will be on the same damn HUB, as apposed to a switch, as you is asking for trouble. Infact, it makes the whole thing seem improbably now. Every single NYSE member would have to be employing inexperienced Idiots in their MIS/IT departments if they allowed their servers to communicate with the NYSE servers where a simple packet sniffer could grab and parse all their data.
The first assumption is the more likely and more probable, they were parsing the messages that announce an interest or the pretrade of a stock and reacting quicker then the NYSE servers or even the originators programs.
The fact its just now coming to light in 2009 shows how slow the media is at picking this up or how little they investment firms think this stuff through. When I had first heard about program trading in 2000 this was the first thing that popped into my mind.
Consider Bernie Madoff. Sterling reputation. Well connected, former chairman of NASDAQ.
Consider all of the data he provided over the years to his clients regarding the transactions he conducted on their behalf.
Consider how over the years, several people approached oversight authorities regarding their suspicions about Madoff’s trading practices.
Consider how these suspicions never went anywhere until Madoff himself confessed he had been running a $50 billion dollar scam.
Now think bigger.
Could Madoff’s been the biggest scam out there? Perhaps. But perhaps not. The scale of the current economic crisis is enourmous, almost incomprehensible. We will never know unless someone looks into it and puts the pieces of the puzzle together correctly.
29.
RareSanity
In theory, having software that can connect to a common point faster than other pieces of software is not illegal or even unethical.
However, if I infer what I believe diary writer is asserting, it is not just that GS was able to connect to the common point quicker than everyone else.
WARNING: EBAY ANALOGY TO FOLLOW…
Imagine that the “trade” is a ebay auction that is ending. There exists “snipping” software that will query what the highest current bid is, set a bid slightly higher, and then submit that bid at the last possible millisecond to insure they are not outbid. This is fair and ethical because at the close of the auction, they are in fact the highest bidder.
However, what if two people are using the software, and they are set to trigger their bids at the same time? Even though impossible for them both to access the common point at the same exact time, let’s say hypothetically they do. The winner of the auction would simply be the one that had the higher increment over the current bid than the other. Or, if their bids were equal, the one whose packets arrived nanoseconds before the other. Or, in Goldman’s parlance, using an innovative, proprietary application that is more affective than any other at beating everyone to the punch.
In fact (if the accusations are true) what Goldman is doing, because of access it has that is not available to others, is first finding out what the other party is sending in as their final bid, making sure theirs is a penny (or some slight value) higher, then allowing both of the bids to arrive milliseconds before the auction ends. All of this is happening without the other party knowing it is even possible, let alone actually happening.
Goldman wins…every time!
That, my friends, is cheating. They have access to real-time information that no one else can access. They have used this extraordinarily important, (and exclusive) information to develop a software package that can take advantage of said information to gain an unfair advantage in the marketplace, which is supposed to be “open”. Call it insider trading, call it an antitrust violation, call it flat out stealing. Whatever you want to call it, make sure you call it illegal.
Of course this is extremely over simplified, but you get the point.
30.
Ken
inthewoods @ 18 wrote: There’s no proof, broadly, that anything has happened.
No, but (if the technology is as described) GS is in somewhat the same position as a person found with burglar’s tools in their possession.
31.
TimO
It’s spelled Godamn Sacs! Get it right!
32.
Cranky
Woah! Slow down people. A lot of you are reading a diary at Dkos and assuming that all of the assertions in it are facts that were uncovered by the FBI and reported by the media.
Nothing could be further from the truth!
The diary is full of crazy speculation that borders on conspiracy theory. STORM has nothing to do with the GS trading systems, it’s a very basic tool for routing telephone calls through an old school telco switch (think Nortel DMS or Lucent 5ESS). The diarist uses the fuzzy language of the patent app to make some bold leaps of logic about what something like this might be used for if put to nefarious purposes, but he does so without any understanding of what STORM is even capable of doing. Classic Tin Foil hat stuff.
I know STORM, I worked on STORM (it was never even fully built because the switches it was going to route are no longer even in use). There is no way that anything in STORM could be used to sniff packets on a financial trading system. None. Nada. Zip.
Any conclusions that are drawn using that faulty reading of the patent app are pure bunk.
Control of to network and server access codes is unclear. A number of shops run their network and job control operations with UNIX shell scripts. Bloomberg and Merrill are typical. Problem with shell scripts is that they get run with superuser credentials — soon as you have 1000 scripts, everybody gets access to the passwords.
It doesn’t have to be the “root” user. Any networking user with access to TCP can grab values from memory.
All of the assertions here hold the author up as a complete buffoon.
34.
Cranky
Yeah, I noticed that too and was going to pint that out in the Dkos diary but I didn’t want to create a pissing match about who has most Unix knowledge. Shell scripts do not run with super user privilege unless the setuid bit is set on the file and some systems wont even let you do that with a shell script.
And seeing what’s in the TCP stack wouldn’t do you much good unless you were on a network segment that was shared by other entities. Apparently the diarist has never heard of a network bridge and private links. I cant even sniff other computers on my home network thanks to my $20 bridging hub. I’m going to go out on a limb and say that the NYSE probably has even more sophisticated networking gear than I do.
35.
southpaw
I’m not as skeptical of Goldman’s scofflaw ways as some here, but there’s an alternate theory of why GS would need the fastest throughput. Rather than front running trades that had already been ordered, they could be trying to get in ahead of the predicted program trading of other firms.
An example, say Slow Bank has an observable program trading pattern that includes selling 20% of its position in any security that experiences a price decline of greater than 4% during the trading day. And say that it takes Slow Bank five seconds (5s) from the time a security crosses that -4% threshold to process the data and execute its sell order. Now, say for the sake of argument that, Fast Bank can (a) detect that trading pattern and (b) figure out with some measure of confidence which securities are part of Slow Bank’s portfolio. If that’s the case, and Fast Bank can also execute a short trade for one of those securities less than 5s after it learns of a -4% move in the price, then Fast Bank can reliably profit off Slow Bank’s trading program. Further, I’m inclined to think that would all be perfectly legal.
Comments are closed.
Share this ArticleLike this article? Email it to a friend!
Lesley
Broadly speaking, the diary seems to be saying that Goldman might have been front running. That they developed a computer program that gained access to information on trades before the trades were finalized and then used that information to place their own trades.
bago
THEY FUCKING CONTROLLED THE TRANSACTION_COMMIT CODES?
With unencrypted transaction control identities in the packets and hardware control, that means you can delay the commit phase of a transaction.
This means if someone asks to do a trade, you can halt the trade, do your own trade, and then have the results of that trade change the price of the original trade, and possibly eve cause the original transaction to timeout, rolling back the entire trade!
It’s grand theft exchange!
The only way you could pull this off without the possibility going to jail for a very long time is if you altered the timestamps on the transmissions. The listed superused ability to read process memory can give you that, depending on if you have write permissions. Of course you own the boxes, you can give the other processes lower cpu priority to ensure that you will have time to make the writes you want to cover your tracks.
Of course if the transaction information is unencrypted and you own the boxes, you can do any number of timing tricks beyond the one I listed above.
At any rate, anyone who is packet inspecting transaction data is a fucking felon. It’s a classic man in the middle attack. The idiots who didn’t encrypt are failing, but the evils exploiting them are violating the entire principle of exchange, and need to be put in a federal pound me in the ass prison.
HeartlandLiberal
In a nutshell, Goldman Sachs has been shown by recent in depth article by Matt Tabbi at Rolling Stone to have been manipulating the roller coaster of bubbles in the American Economy since the Great Depression:
The Great American Bubble Machine
Now it is becoming clear that Goldman Sachs for the past decade has been running sophisticated computer programs doing quick response and rapid programming to achieve programmed trading that has been peeling of zillions in profits. But what this also represents is blatant daily manipulation of the markets for these profits.
A Russian programmer figured out how to do this. So then and only when their long secret manipulation, which has been covertly allowed and kept under the rug from public view is being exposed, are the powers that be scrambling not to expose Goldman Sachs for the evil they have been doing, but to sweep the whole sordid mess under the rug before we, the plebs and investors who are being ripped off by massive market manipulations, figure out what is going on and rise up and put an end to these practices.
It really is just one more case of exposing the dark, criminal under side of what has evolved in American corporate and Wall Street culture. And the sooner they can convince you not to look at that man behind the curtain who is screwing you blind, the better for them to retain their power.
DecidedFenceSitter
*IF* I’m reading correctly, here’s my understanding.
In information security there are three key factors Confidentiality, Integrity, and Availability. Confidentiality – Only the intended senders and receivers can read the message. Integrity – The message remains unchanged. Availability – The system is available when needed.
The key here is confidentiality, the key line from the article is from the introduction, “is technology to grab off FIX PROTOCOL, OCX, or SWIFT messages that precede every transaction_commit at the Exchanges.” Basically, it appears that thanks to no encryption, “Encryption is optional or nonexistent or easily enough made transparent. NYSE has any and all access codes involved.” And we’re discussing split second transactions, which means if you can read the mail, you know what your competition is doing.
What the STORM patent does (if I’m reading it correctly) is allow for the real-time reading of the packets passing through the servers to route them. This is required to get the achievable results of the FIX Protocol, OCX, SWIFT messages.
Key take away for the non-financial, no security folks: “Access to FIX, OCX or SWIFT messages prior to transaction_commit at the Exchanges would give a player an advantage parallel to seeing an opponent’s cards in a game of poker.”
DecidedFenceSitter
My comment is in moderation, but unsurprisingly enough, we’ve got a few folks who get computer security on an online blog. :)
HeartlandLiberal
To expand on my comment above, this from another diary at DailyKos on this issue:
The Outrage Against Goldman Sachs Builds
Meta P.S. Blockquoting is still broken in this interface. I could not enclose the above two paragraphs in one blockquote tag set. And I did NOT bold the first paragraph. Whatever CSS code that is parsing the HTML is doing that all on its lonesome own, not me.
bago
@HeartlandLiberal: In layman’s terms, it’s as if they look through all of the letters asking for trades on the exchange because they own the mailroom, write new letters making those trades, send them first, and then continue to send the original letter to the trading floor.
Lesley
@bago
Wow, then it would be even worse than plain vanilla front running. I didn’t understand the parts about controlling the transaction_commit codes. So they might have been getting information on the trades prior to execution and then delaying execution of the trades until their own trades were placed.
bago
@Lesley: It could be. You can screw with a lot if the assertions in the Kos article are true. Depending on the environment and the encryption, you can do a lot, and the notion of superuser accounts running production code should scare the bejeezus out of you.
Lesley
OK, correcting myself. Usually front running is using data from your client’s trades, so you do control when the trades execute. In this case, it sounds like they could have been using data from and delaying everyone’s trades, not just their own clients.
bago
@Lesley: To be more accurate, they were using data from any trades that passed through a system they owned. Slowing down competitor trades just makes that process more effective, and may have been possible. Again not enough details to be sure, but based on what is known, you can do message inspection into competitors trade commands in real time, possibly fast enough to beat exchange transaction times, thus you can get the good deal that some other trader saw as fast or faster than your competitor, essentially stealing the competitors trading knowledge.
Not just using data from your clients, but from anyone else on the system you were running.
If the article is true.
Lesley
@bago
Sounds to me like they’ve tacitly admitted the program has the capability to do more or less what was described. They admit it could be used to manipulate the market, so it must have at least some of that functionality. I guess it’s down to whether we trust them not to have abused that functionality.
And it does scare the bejezus out of me.
Karmakin
Here’s what I got out of it. If true, GS is observing the trade stream in real time, pulling out the basic info from it. If they see a large buy order, they jump ahead, buy some of that stock then resell it at a profit to the person who put in the buy order. If they see a large sell order, they then short the stock.
Basically what they’re doing is they’re skimming money off the top.
That said, I’m a person who thinks that the public market system is a fraud as a whole and actually fights against what it is intended to do (provide capital). So I’m probably too cynical on this.
PeakVT
Sounds plausible. I think the biggest problem would be making the timestamps look realistic. If GS is frequently making a trade X milliseconds before another larger trade, that pattern would show.
The Grand Panjandrum
Shorter Kossack Diarist: Goldman Sachs appears to be doing a high tech version of front running the entire market.
bago
@PeakVT: Hence my comment about superusers and timestamps. Again, depends on lots of precise technical information such as the operating system, the permissions that each process ran in, what was shim’d out to who, logfiles, disk permissions, etc.
Without access to the machine configurations, you can’t tell if it is possible. Without access to the logs and binary you can’t know if it was actually done.
However the author seems to speak in a manner that indicates that this is a possibility, and should be investigated.
If it is true, then this means that Goldman appropriated itself as a semi-governmental entity and has been taxing everyone. Even though half of the treasury department is ex-Goldman, to borrow a phrase, “Don’t steal, the government hates competition”.
Walker
If this is true, it is grounds for a class action lawsuit from every single person who has bought a stock in the last decade.
inthewoods
I think it’s a lot of smoke and no real fire. There’s no proof, broadly, that anything has happened. That doesn’t mean it hasn’t, but the current evidence is weak. You’ll need stronger stuff (that may well exist) to take down GS on something like this.
The Kos article does come across as a bit 9/11 tinfoil hat imho – but hey, that’s just me.
Redhand
It’s fodder for a massive insider trading investigation and criminal prosecutions. But, is this yet another blatant securities law violation they’re going to get away with, . . . because they are GS?
Walker
Reading the article more thoroughly, and the comments, I am starting to believe that this dairy is a bit premature. It needs more proof.
We have known for years that GS has been investing in high throughput data dissemination software. It is possible for them to get a leg-up (e.g. get the information first) and do all of this simply because their software was better. If they actually had a software advantage, it would be stupid to risk it on something as damaging as reading before the transaction commit.
Napoleon
Lets say they are reading it after transaction commit, would everybody else in the world be the the same position to take advantage of that fact?
Walker
If they had the software to read it fast enough, yes. But it is unclear that anyone else was spending the money on this that Goldman Sachs was.
Goldman Sachs hires A LOT of high performance computing, distributed systems, and data management PhDs (several of the students in my group have been offered jobs). They are the Google of the financial world. All their software is developed in house. Any attempts to sell them third party software (e.g. Stonebraker’s Streambase) have failed because they believe their proprietary stuff is better.
Any start-up can hire a quant, but you need a critical mass to build a competitive software financial infrastructure.
jwb
One thing I’m confused about is the patent. What nutcase patents something illegal? So I have to believe that at least what’s described in the patent is perfectly legal.
Walker
The comments on the Kos piece say that the patent is for routing in voice networks. The company IDT is just a telco; STORM was not designed for financial trading. Only its implementation in a financial system would be illegal. The patent was developed at IDT, not Goldman Sachs.
This is a highly suggestive coincidence, but right now there is nothing more than that. There may be fire here, but we need more evidence.
Bill Arnold
Basically, seems to be accusing GS of engaging in very high speed automated insider trading, or worse, with no actual evidence. Just (1) it’s possible (2) Aleynikov has some skillz that are hypothetically applicable (3) GS made a pile of money. (There may be (4) etc in there as well.)
(I am not familiar with this domain, but … the volumes are extremely high and milliseconds matter (the early bird gets the worm), so there is a tradeoff between slowdowns due to encryption/decryption/signature verification, and speed gained by trusting the intermediaries.)
The Raven
The diarist is alleging that Goldman hacked the NYSE. If true, it’s the biggest securities crime ever.
A Cat
I always wondered what GS needed a sub 15ms response time and this explains it.
Assuming the network traffic was over a stream based protocol, not packet based, a sub 15ms response time would allow plenty of time to grab those messages and decode them and get a response out all before the final message saying the trade had been executed hit any one elses servers.
Having read the full article at KoS, I have to say some people at the NYSE and their members need to be fired. Colocating your servers in a NOC where your members will be on the same damn HUB, as apposed to a switch, as you is asking for trouble. Infact, it makes the whole thing seem improbably now. Every single NYSE member would have to be employing inexperienced Idiots in their MIS/IT departments if they allowed their servers to communicate with the NYSE servers where a simple packet sniffer could grab and parse all their data.
The first assumption is the more likely and more probable, they were parsing the messages that announce an interest or the pretrade of a stock and reacting quicker then the NYSE servers or even the originators programs.
The fact its just now coming to light in 2009 shows how slow the media is at picking this up or how little they investment firms think this stuff through. When I had first heard about program trading in 2000 this was the first thing that popped into my mind.
Rumours of Chaos
Consider Bernie Madoff. Sterling reputation. Well connected, former chairman of NASDAQ.
Consider all of the data he provided over the years to his clients regarding the transactions he conducted on their behalf.
Consider how over the years, several people approached oversight authorities regarding their suspicions about Madoff’s trading practices.
Consider how these suspicions never went anywhere until Madoff himself confessed he had been running a $50 billion dollar scam.
Now think bigger.
Could Madoff’s been the biggest scam out there? Perhaps. But perhaps not. The scale of the current economic crisis is enourmous, almost incomprehensible. We will never know unless someone looks into it and puts the pieces of the puzzle together correctly.
RareSanity
In theory, having software that can connect to a common point faster than other pieces of software is not illegal or even unethical.
However, if I infer what I believe diary writer is asserting, it is not just that GS was able to connect to the common point quicker than everyone else.
WARNING: EBAY ANALOGY TO FOLLOW…
Imagine that the “trade” is a ebay auction that is ending. There exists “snipping” software that will query what the highest current bid is, set a bid slightly higher, and then submit that bid at the last possible millisecond to insure they are not outbid. This is fair and ethical because at the close of the auction, they are in fact the highest bidder.
However, what if two people are using the software, and they are set to trigger their bids at the same time? Even though impossible for them both to access the common point at the same exact time, let’s say hypothetically they do. The winner of the auction would simply be the one that had the higher increment over the current bid than the other. Or, if their bids were equal, the one whose packets arrived nanoseconds before the other. Or, in Goldman’s parlance, using an innovative, proprietary application that is more affective than any other at beating everyone to the punch.
In fact (if the accusations are true) what Goldman is doing, because of access it has that is not available to others, is first finding out what the other party is sending in as their final bid, making sure theirs is a penny (or some slight value) higher, then allowing both of the bids to arrive milliseconds before the auction ends. All of this is happening without the other party knowing it is even possible, let alone actually happening.
Goldman wins…every time!
That, my friends, is cheating. They have access to real-time information that no one else can access. They have used this extraordinarily important, (and exclusive) information to develop a software package that can take advantage of said information to gain an unfair advantage in the marketplace, which is supposed to be “open”. Call it insider trading, call it an antitrust violation, call it flat out stealing. Whatever you want to call it, make sure you call it illegal.
Of course this is extremely over simplified, but you get the point.
Ken
inthewoods @ 18 wrote: There’s no proof, broadly, that anything has happened.
No, but (if the technology is as described) GS is in somewhat the same position as a person found with burglar’s tools in their possession.
TimO
It’s spelled Godamn Sacs! Get it right!
Cranky
Woah! Slow down people. A lot of you are reading a diary at Dkos and assuming that all of the assertions in it are facts that were uncovered by the FBI and reported by the media.
Nothing could be further from the truth!
The diary is full of crazy speculation that borders on conspiracy theory. STORM has nothing to do with the GS trading systems, it’s a very basic tool for routing telephone calls through an old school telco switch (think Nortel DMS or Lucent 5ESS). The diarist uses the fuzzy language of the patent app to make some bold leaps of logic about what something like this might be used for if put to nefarious purposes, but he does so without any understanding of what STORM is even capable of doing. Classic Tin Foil hat stuff.
I know STORM, I worked on STORM (it was never even fully built because the switches it was going to route are no longer even in use). There is no way that anything in STORM could be used to sniff packets on a financial trading system. None. Nada. Zip.
Any conclusions that are drawn using that faulty reading of the patent app are pure bunk.
dbt
All of the assertions here hold the author up as a complete buffoon.
Cranky
Yeah, I noticed that too and was going to pint that out in the Dkos diary but I didn’t want to create a pissing match about who has most Unix knowledge. Shell scripts do not run with super user privilege unless the setuid bit is set on the file and some systems wont even let you do that with a shell script.
And seeing what’s in the TCP stack wouldn’t do you much good unless you were on a network segment that was shared by other entities. Apparently the diarist has never heard of a network bridge and private links. I cant even sniff other computers on my home network thanks to my $20 bridging hub. I’m going to go out on a limb and say that the NYSE probably has even more sophisticated networking gear than I do.
southpaw
I’m not as skeptical of Goldman’s scofflaw ways as some here, but there’s an alternate theory of why GS would need the fastest throughput. Rather than front running trades that had already been ordered, they could be trying to get in ahead of the predicted program trading of other firms.
An example, say Slow Bank has an observable program trading pattern that includes selling 20% of its position in any security that experiences a price decline of greater than 4% during the trading day. And say that it takes Slow Bank five seconds (5s) from the time a security crosses that -4% threshold to process the data and execute its sell order. Now, say for the sake of argument that, Fast Bank can (a) detect that trading pattern and (b) figure out with some measure of confidence which securities are part of Slow Bank’s portfolio. If that’s the case, and Fast Bank can also execute a short trade for one of those securities less than 5s after it learns of a -4% move in the price, then Fast Bank can reliably profit off Slow Bank’s trading program. Further, I’m inclined to think that would all be perfectly legal.