Saturday, February 09, 2008

Fixnetix vs Wombat

I have heard of some companies starting to look at Fixnetix as a possible replacement and/or alternative to Wombat for low-latency market data feeds.

Has anyone done a bakeoff between the two? How about Activ?

Wombat is starting to promote its stack that include Coral8 for CEP/latency monitoring and OneTick for a tick database. Reuters has the same with Streambase and Vhayu. This means that Fixnetix and Activ will probably be looking towards that direction as well. Can an alliance with Aleri be in the cards for any of these vendors?

©2008 Marc Adler - All Rights Reserved


Anonymous said...

Hi Marc,
I believe FiXNetIX is a hosted Wombat solution. At least this was the case when I was in the UK with previous big Bank. Same thing as Thomson's hosted Wombat idea.

IMHO Activ took a major wrong turn when they decided to go down the path chosen by the ex-Primark guys (think big bloated API with group subscriptions) rather than the successful AT-Financial model (bare bones but bloody quick).

In the end whoever takes the plunge and rolls their own (probably directly embedded in their trading apps) will obviously have an advantage. Its not that hard either if you just want CTA/UTP/ARCA/INET - probably easier than trying to get a stable Wombat client going that doesn't mysteriously lose subscriptions or crash due to threading issues.



James said...

I know of one house that used to use Wombat feed handlers but flipped over to Fixnetix direct feeds instead. Other than the fact that they are very happy I cannot get any more information from them. So I guess that answers your question!!!!

Anonymous said...

I've seen reuters direct feeds out-perform the others (wombat, activ, and fixnetix) in recent bake-offs. I think wombat's gotten in a bit of trouble in the feed space with nyse acquisition.

eurovoip said...

I think you're comparing apples with peaches; ie, a real market data software house with an integrator of connectivity/hosting services similar to BT Radianz or Yipes.

A real end-to-end Ultra Low Latency Market Data solution requires, for the provider, to
1 - have its own infrastructures WITHIN the exchange to CAPTURE the data at the source,
2 - design/develop/maintain its own feedhandler technology to STANDARDIZE data directly at the source,
3 - design/implement/maintain its own fiber optic network to leverage the speed of light advantage to DISTRIBUTE datafeed to trading engines.

To be able to provide this end-to-end solution, providers need to be able to :
A - rely on strong shareholders to face the huge investment we are talking about,
B - have an experienced staff of developers working for a couple of years on feedhandler development and market data technologies,
C - control each element of this end-to-end chain with its own technology, staff and processes.

Coming back to your blog, here is my answer to James Fixnetix :

Wombat is a leading software house which delivers technology to standardize feeds. No infrastructures, fiber optic network, DMAs,.. so no end to end solution!

And what is Fixnetix ?
Point A : ? Anything solid here ?
Point B : they rely on wombat products. It's hard to believe that, just in a couple of weeks (since the acquisition of Wombat by NYSE Euronext which certainly puts at risk their partnership), they were able to develop by themselves what professional companies took 6 years+ to design and develop. Only unexperienced people could believe that.
Point C : goto point A and B

Reuters :
Point A : what about flexibility and agility
Point B : is the development still under control ? Isn't it outsourced ? Clients should check.
Point C : no end to end solution possible as collecting and distributing network have been outsourced to BT.

to whoever is reading this blog and , as a professional client, is looking for a real end-to-end ultra low latency market data solution, do not forget in your selection process QuantHouse :

and do the benchmark !

then, you'll be able to make sure that your decision is the right one to allow your firm to trade ahead

Jerry =)

Anonymous said...

Ah Jerry, methinks you should know better - where do I start with a post that says "this is apples vs. peaches - a real vegetable does the following ..." ?

Lets just go through them one by one:

1 - speed of light = ~5us/mile, time through a gigE switch ~50usecs (or thereabouts), difference in latency due to distance if you are in the NYC metro connecting to major venues is therefore pretty negligible. The big players cant skim enough off market microstructure to get value out of a few usecs and the available metrics aren't sensitive enough for customers to differentiate between products with different proximities. The primary goal of more trading systems is to avoid routing out altogether and to internalize as much matching as possible. Latency doesnt matter much here as long as you have sufficient flow.

2 - not sure why/what u mean by standardize. Data normalization is moot if you end up with different rules from different venues that depend on venue specific quote/trade conditions. Certainly Wombat do a poor job - they represent index values as prices, use double precision floats for price representations, overly complex orderbook deltas, etc. Most people buy feed handlers to avoid having to maintain the code to decode the different wire protocols but this is becoming less important as venues merge and move towards FAST FIX variants (which is also junk).

3 - eh? What about SFTI or Radianz Ultra?

A - remove the connectivity piece and its just software - all you need is a Linux box and some data.

B - it takes years to build market data components in the same way it takes years to write device drivers or learn to fly a plane (wink at marc). If you start from scratch and work your way from up through logs to square wheeled bicycles then its obviously going to be painful, but if you stand on the backs of a few giants, have the right tools and data access there is no reason why you shouldnt be able to churn out reliable feed handlers in a few weeks or less. If it took man years per handler I would have retired by now :-)

C - there you might be right, but the key is to minimize the hops and handoffs rather than the distance (within reason). Using ATM (DS3/OC3) implicity adds ~62.5usecs per hop due to the 8,000 pkts/sec timing - thats equivalent to ~13miles of distance per hop.

Wombat handlers will significantly outperform RDFD. If you have multiple consumers and have to use RDFD->RMDS there is no comparison. Having said that it takes a fair amount of pixie dust to get Wombat handlers configured optimally so it is likely that some people end up with RDFD (configured and managed by Reuters) performs better than a Wombat handler - in fact you are lucky if it even runs out the box nowadays. If you are not going to go RDFD->RMDS then you probably dont want RDFD anyway due to the inherent redistribution restrictions that Reuters places on the data.

> do not forget in your selection process QuantHouse

who? I mean seriously - is anyone actually using this? The London guys usually stick Finetix somewhere on the list (second page near the bottom after the FPGA loonies) but so far I've heard nothing about Quanthouse as a market data provider (maybe someone mentioned CEF in Chicago once?) - not even a cold call! Tervela at least gave me a pen :-)


Anonymous said...

Hi Marc, what's your opinion on fully managed hosted direct feeds? Is there a strong interest from the bulge bracket or top hedgies?

What's the prevaling thought on the costs to continue to maintain direct feeds in house versus outsourcing? Is there a good option out there?

Anyone have opinions on Spryware, InfoDyne, or Interactive Data in this space?

Anonymous said...

no need to get into multiple championing of different providers....but i will say that i've seen bakeoff stats for wombat vs. rdfd vs. fixnetix and rdfd is simply faster. it may be the case that this will only be so when in combination with rmds, but rdfd performance tops the others.

Anonymous said...

I've seen results of a bake-off performed by morgans in London with wombat, RMDS and i think a couple of others. An outsider MST ( beat them all with their Market Data Direct product.


Anonymous said...
What is the story here. MST website links to Fixnetix?