Right, if you weigh that in for latencies you're not going to get <1ms anywhere. That's true, some brokers have excellent connectivity to the exchanges because they're on private circuits close by or actually colo at the exchanges themselves (Equinix in Chicago or IXEurope in Frankfurt for example). If they also have an effective and well written risk engine that minimizes risk processing times and their order engine doesn't run at cap you're most likely well off. However, to find the best solution for a trader it's important to know about the products traded, avg. orders/sec, cancels, etc.
Hmm ... ok. I basically want to trade everything in the stocks/futures universe, but I'm a spreader / relative value trader. So execution times are huge for me. For example (although maybe a bad one), let's say I want to arb SPY on the ECNs vs the ES on CME. I could either get data and direct access to all of them separately, write gateway software, etc. Or I can pay commish to a brokerage that provides all that. If I were to go with the latter, who rules the roost? A trade like this would probably be something like 8-10 orders/sec., I'm guessing.
I guess I will stick my head in. First of all, to get a very low latency, your broker must have direct point to point connections to all of your market data and routing destinations, and no, something like a Radianz won't do, as the data are redistributed from a Radianz PoP. So by definition, you are looking at a broker that runs its own ticket plant. Two, for order routing, you need a borker that let your system to route directly to their respective exchange gateways, while abstracting the exchange protocols from your system (I am assuming that your system does not have hassle of writing all the different exchange protocols, right?) You also mentioned at least stock and futures, for the heck of it, let's throw options (vanilla equity and index) in the mix. Then there are very few firms that would fit the bill. GS/SLK/FOC (first options) would easily fit the bill, naturally. And I know that the GS prime brokerage service runs their own direct feed ticker plant (separate from SLK/Redi and the GS's own internal ticket plant to feed internal desks, etc). Well, your guess is as good as mine what it would take to become a GS PB client. The only other player in the product coverage space might be Merrill, but I seriously doubt how much they have their technology act together. And also to keep in mind the geographical differences, RedSky is in Chicago, so their routing to NY destinations would suffer the distance latency, LimeBrokerage is in NY, so anything they route to Chicago, well, you get the picture. There are very good reasons why most chicago firms route their equity trades to ARCA. Most of the fastest systems I know are strictly proprietary systems, and these firms either have no clients, or these systems are not available for clients (proprietary trading only).
Good reply. Basically, what you said is why I'm asking the question. I can put together some pieces, but not the others.
I think for GS BP you're looking at 10mil min. I heard good things about their technology side but never experienced it first hand though. I've had lots of problems with ML in the past so that wouldn't be my top choice. For futures, ABN is pretty good but I have no idea how good they are on the equity side. London
I think a weakness is assuming the big banks have the best technology. It is a time of smaller and more nimble places are outdoing the monsters. If you look at the Nasdaq list of liquidity providers, you'll see a lot of places that represent who is getting orders - because at its' simplest, you have to figure the providers are more in need of speed - to the market quickly. It really winds up being a volume game, and the volume is flocking to the providers who can HANDLE the increased flow, and quote traffic, going on right now. So, you can have a PB, or you can use an IB into a PB, and there are a slew of brokers providing the technology and execution services out there for a group that does fly a bit under the radar. And, to do so, with low leverage, that exists out there currently. YOu do need to make sure that point to point is direct, yes. You also have to ask how it is managed, what their volume is, and how not on;y latencies - any one can say "wow, we're fast" but how do they perform with their volume at peak times - open, close, Fed, numbers. It shouldn't take that long to look around and figure out who the monsters are in this space. If they specialize in something, be it US equities, or futures, it's a matter of figuring gout if they cal also handle the "other" side of the trade. If they can get say CME data/trades through the same API, your automated strategies should be fine. Any broker is going to be happy to talk to you if you're doing volume, and if you needed another market, it's just another variable cost in the equation. If they wanted the GUm Stock Market (fictional) added...if there was volume, and the technology could be added, it would be added on. The biggest challenge is to figure out who really specializes in what you need the most. Any place that just mentions they "Connect everywhere" isn't really stating they have the best hook up to say LSE. They're just saying, wow, we have a shared line or a service provider helping us along the way. You may need, if I'm right here, someone that has direct point to point connections and MANAGE those connections, serve the boxes, and keep up with the changes in quote traffic. Some people here may be able to point out what you need. I'm just saying you've got to take a lottttt of factors into account to maximize your technology and minimize your headaches.
Enterprise data feeds have latency < 1ms But not cheap... probably into 5 figures/month for a full setup... A different universe from low-end feeds. Here is an example: http://about.reuters.com/productinfo/informationmanagement/material/Reuters_DataFeedDirect.pdf If someone knows more about implementing this... please advise.
Somebody made an interesting point that banks do not have the best technology and the smaller firms actually have a competitive advantage. I somewhat agree, to an extent, only if small firms invest heavily in cutting edge technology, high caliber people can the small firms compete effectively. And spending 5-10-15M in technology and some ppl is not considered a huge amount for a lot of the leading smaller firms. Large banks are ladden with bureaucratic layers that makes decisions by committee, and that's an inherently inefficient mode of operation. But on the other hand, cutting edge technology tend to be expensive, experimental and require special knowledge just to work with. It is not unusual in highly technology dependent smaller firms for technology to account for 25-35% of overall budget (including compensation). I had a RMDS Direct (what they called it back in late '04) setup when I was running e-trading for a certain firm, the feed, as you guess it, requires full direct point to point connection to all raw data sources, including OPRA (which is > a T3 nowadays, of course x2 for redundancy). The setup was costing about $135k a month for feed handlers and managed service alone, then figure about 80k / month more for the feed P2P lines. Not to mention about 8 edge feed handler servers and 2 full racks of data center space, and it still took up 3 FTE just to manage and run the setup And it was considered even by my Reuters rep that I did it on the "extremely cheap", since I did not have full redundancy for all the edge handlers, and 3 FTE can not provide 24x7 coverage (1 handler specialist, 2 RMDS managers / operators). A certain market making firm that I know well spent about 8-10M for a fully built-out Direct service, including 3M just to expand the data center space. But nowadays the Reuters Direct is not considered state of the art anymore, since the Reuters Bridge seems to be a bottleneck in data aggregation and transformation (less than optimal latency). A more popular setup among investment banks and large trading firms have been Wombat, which is a raw, unfiltered setup, a typical Wombat setup for US destinations alone is usually in the 15-20M range, or so I have heard from ppl who have done it. http://www.wombatfs.com/index.htm