Forums  > Software  > OneTick  
     
Page 1 of 1
Display using:  

BuzzMeeks


Total Posts: 6
Joined: Jul 2009
 
Posted: 2009-07-13 14:27
Anybody have any experience with OneTick? From what I've been told it was developed by a group of Goldmans quants who set up on their own, and OneTick allegedly fills gaps left by kx, vahyu and the like....

brainyoga


Total Posts: 160
Joined: Jun 2004
 
Posted: 2009-07-21 00:19
I have not experienced onetick first hand, but I have used kdb+ for many years, and have moved around a lot amongst their clients and prospects. What gap do you think has been left by them?

size does matter

d4vid


Total Posts: 1
Joined: Jul 2009
 
Posted: 2009-07-22 17:28
I've worked with OneTick and Kx in the past. I'm unsure of the gaps you refer to, but OneTicks use of standard langauage APIs is a plus for some users. What do you need to know?

BuzzMeeks


Total Posts: 6
Joined: Jul 2009
 
Posted: 2009-07-22 18:05
Sorry brainyoga, I missed your earlier reply.

I just want to know if anybody out there who has used both can tell me pros and cons for OneTick vs kdb as I am faced with a requirement for "last tick" historical data whereas "yesterday's" data has always been sufficient in the past our strategies.

An ex colleague of mine uses OneTick and he has been evangilising about it as being the best thing since Nandos, but I would like a neutral view if possible. I am probably going to rule out Vahyu and Vertica as solutions after some initial digging around but will probably start looking seriously at kdb and onetick some time after the summer.

d4vid, thanks for your response, you have told me something I knew already about OneTick :) I agree its a big plus though as the Q language is not the most intuitive or career friendly.

I'd be more interested in peoples views on functionality, performance and architecture (kdb vs onetick).

brainyoga


Total Posts: 160
Joined: Jun 2004
 
Posted: 2009-07-31 23:26
I don't follow the standard language API argument as kdb+ has always had a c, java, c# api and they are very simple as they are just (de)serialization routines. Q is more intuitive to those who have studied functional/vector/set languages such as sql, lisp/scheme, haskell and matlab than more mainstream languages such as c, c# or java. Career friendly - my experience of learning Q was that it made me a better programmer; I've not heard of it hurting anyone's career.

I can speak about kdb+ functionality, performance and architecture.

Functionality - it is a virtual machine (byte code interpreter with super instructions), with built-in data types for vectors, dictionaries (hashtables) and tables. It has a bunch of built-in functions for some very powerful joins, and Q can sometimes look a bit like sql - e.g.
select size wavg price by sym from trade where date=2009.06.30, sym in `MSFT`CSCO, time within 10:00 11:00
There are feedhandlers out of the box but the c api is so simple it is quite straightforward to develop your own. And it hooks in to the common tools like excel and matlab.

Performance - Arthur Whitney has been optimizing the algorithms used in kdb+ for the last 30 years (Morgan Stanley and UBS had him for a few years to work on their trading systems). The executable is a couple of hundred kB, and is blazingly fast.

Architecture - It's a 64bit app and can be run in multithreaded, multiprocess and in a distributed mode on windows, solaris, linux and osx. It scales extremely well, although designs tend to revolve around concepts of in-memory database, on-disk database and gateways.

Last year there was a lot of talk about disk compression in the kdb+ community, but Kx chose to leave compression up to the file system software (e.g. ZFS gives a default compression ratio of about 3.3 for TAQ data) rather than implement a proprietary version themselves. Apparently another vendor had chosen to implement their own proprietary compression algorithm and was found to have slowly corrupted data, discovered some time after deployment.

There's a google group http://groups.google.com/group/personal-kdbplus and you can get a non-commercial version of kdb+ from kx.com. Kx has an internal group for commercial users.

size does matter

krish


Total Posts: 1
Joined: Oct 2011
 
Posted: 2011-10-11 13:11

Have been using OneTick for over two years now, and i can tell you that its quite user friendly , there are tons of event Processors which you can use to do your routine tasks, and its really really fast ( comparable to that of KDB).

However you really cant program complex programs on One Tick unlike K+ ( probably less than 5% stuff) , thats probably only one drawback, however if you want really really complex programs even K+ cant handle that . In that case you probably have to Interface with Matlab or R.

All in all its beautiful to use , its killer fast, the error reporting is cool , you hardly ever face any problems once things are up and running.

 


ESMaestro


Total Posts: 127
Joined: Jun 2009
 
Posted: 2011-10-11 22:51
Anyone have ball-park cost estimate for a single-user license (non-enterprise)?

FDAXHunter
Founding Member

Total Posts: 8224
Joined: Mar 2004
 
Posted: 2011-10-12 07:53
I'm not sure they do "single users". Reason is that an installation always encompasses several components: A database server, a CEP server, dev server(s) and clients. Even for a single-user case, it's a decent installation.

We received an indicative quote two years ago that ranged from 80,000 USD to 300,000 USD/year (depending on options suchs as number of feed handlers/number of cores/servers/etc). All of these providers (OneTick, StreamBase, Vertica, kx, Apama, etc.) tend to cluster around the starting price point of around 100,000 USD-200,000 USD, give or take 50,000 USD again depending on options). A global installation for an investment bank, however, can easily cross a few million USD.

Of course, your mileage may vary and products like this tend to be subject to very negotiable fees (and how much money you have. If you are a tier 1 investment bank, your quote will likely be a factor of 5-10 higher than that of a Namibian grain trader, at least that's how software license agreements of this sort tend to be handled in the industry. Deviations "from list price" tend to be massive). I obviously only have one quote to compare with.

I haven't looked at it since, but I wasn't very happy with it (features & functionality wise). I'm pretty bearish on all these "financial database" providers in general.

I can give you more color if you want to e-mail.

The Chaos Army seems suspiciously well-organized.

signalseeker


Total Posts: 232
Joined: Oct 2006
 
Posted: 2011-10-12 10:03
> krish: however if you want really really complex programs even K+ cant handle that . In that case you probably have to Interface with Matlab or R.

I am sorry but you have no idea what you are taking about.

The dark is light enough.

FDAXHunter
Founding Member

Total Posts: 8224
Joined: Mar 2004
 
Posted: 2011-10-12 11:48
Well, the fact that he's calling it "K+" (probably referring to "kdb+") is somewhat indicative of his familiarity with it Smiley.

The Chaos Army seems suspiciously well-organized.

BuzzMeeks


Total Posts: 6
Joined: Jul 2009
 
Posted: 2011-10-12 12:42
Exposed.

OneTick has taken off as a business since I started this post a couple of years ago - particularly in the quant hedge fund space. kx and OT are the only real options for serious players - although there have been persistent rumours for some time now that Reuters are investing heavily to try to improve (and get some ROI) from Vhayu.

We'll see....

LouisLovas


Total Posts: 1
Joined: Oct 2011
 
Posted: 2011-10-20 00:37

OneTick has a lot of advantages as a tick database, but also with an integrated CEP engine to handle real-time analysis as well. Analytical queries can span both real-time and historical for trades, quotes and order book data.

There is a technical whitepaper on OneTick available from the following website:

http://www.onetick.com/web1/omd_webinars.php#whitepapers


Scotty


Total Posts: 692
Joined: Jun 2004
 
Posted: 2013-06-10 15:08
Well, we are evaluating it at the moment. It certainly handles a lot of the 'data handling' issues with intra-day data and has some nice event handling.

We are using the python APIs to link Bloomberg data going in and also for trade execution on the way out.

So far it looks pretty good.

“Whatever you do, or dream you can, begin it. Boldness has genius and power and magic in it.”

BuzzMeeks


Total Posts: 6
Joined: Jul 2009
 
Posted: 2013-06-10 15:43
In the 4 years since I started this post things have moved on. Looser schemas, unstructured data and clustered solutions for massive parallel querying seem to be the soup du jour for the quant guys that once used OneTick. MongoDB is where its at nowadays - opensource, quick and powerful.

There is a strong trend towards opensource in FinTech now as propietary software in this space has always been overpriced to cater for firms with deep pockets. The money man has gone - market data, order routing/FIX and now data storage/analytics are all heading to the open model.

Forget OneTick and kdb.

Scotty


Total Posts: 692
Joined: Jun 2004
 
Posted: 2013-06-10 16:07
.

“Whatever you do, or dream you can, begin it. Boldness has genius and power and magic in it.”

Scotty


Total Posts: 692
Joined: Jun 2004
 
Posted: 2013-06-10 16:08
Yeah, the big data thing. I'm going to leave that to those big data people. I'm also going to leave the ultra-high frequency stuff to the hardware guys.

I think there is still mileage in identifying and exploiting relationships in time series of trade and quote intra-day data. Possibly supplemented by lower frequency indicator time series for conditioning.

I want to outsource the data handling to someone like OT or kdb and focus on the algorithm development.

No Buzz? (edit, in other words, it would be good to discuss).

“Whatever you do, or dream you can, begin it. Boldness has genius and power and magic in it.”

jslade


Total Posts: 867
Joined: Feb 2007
 
Posted: 2013-06-12 01:09
KDB is pretty damn clever; I considered paying $50k/year for it for my personal stuff. If I were a bank, I'd certainly consider it as well.
The risk for one of these fashionista big data stores is finding people to make it go, hoping they implement correctly for the thing to work right 2 years up the road, hoping the core libraries don't change underneath you at some crucial moment, hoping the maintenance/development costs aren't too horrible, and trusting it will actually do what you need.
Biggest risk for KDB is finding people to make it go. Compared to paying people to build similar stuff, it's cheap.
This won't always be true, but I bet it's true now.

"Learning, n. The kind of ignorance distinguishing the studious."

redandtheblue


Total Posts: 345
Joined: Aug 2007
 
Posted: 2013-06-14 17:49
Open source packages aren't there yet. Blaze by continuum looks awesome. I really do not see any viable open source packages yet.

brainyoga


Total Posts: 160
Joined: Jun 2004
 
Posted: 2013-07-02 18:23
Agreed. Open source has always looked attractive but never quite delivered in this space. And sure, open source moves forward, but so do commercial offerings.

size does matter

radikal


Total Posts: 112
Joined: Dec 2012
 
Posted: 2013-07-20 20:43
I've got my eye on Blaze as well; pytables/h5 is currently in use a fair amount in HFT space. It's not viable really for some use cases...but for those who typically work on a day's worth of data at a time, h5 is pretty awesome.
Previous Thread :: Next Thread 
Page 1 of 1