[R] R as chart engine in web-service
jholtman at gmail.com
Mon Jun 16 18:43:10 CEST 2008
I think that you have to define some type of criteria on what you
expect the performance to be. There is a cost/benefit tradeoff in
having multiple servers with a load balancer and this all depends on
how much you want to spend on a solution. You might want to prototype
it in R to at least get an understanding of the data transformations
that are required and then after you have some performance data and a
target for usage, you can then make a better decision.
On Mon, Jun 16, 2008 at 8:26 AM, Daniel Cegielka <domibre at gmail.com> wrote:
>> Can you put some criteria on 'better'/'faster'. What throughput are
>> you expecting? How many requests per second will you have to handle?
> It's very difficult to tell how many request per second I will have to
> handle. If you have normal site it will be max. 2-5 tps (on chart website)
> but if you have real time quotes it will be about 300 tps. In RT quotes I
> intend use java applet to make RT charts with independent connection.
> I thing about service like BigChart.com - put some symbol (GOOG, DJIA etc).
> and the chart like:
> In this kind of web services its important how faster the chart engine is in
> general. If there will be to many request per second I can add more machines
> and use load balancer (haproxy or pound).
>> What size/type machine will you be running on?
> Probably 1-2 Quad Core Intel Xeon 2x6MB Cache, 2.0GHz per machine with 8GB
> (or 16GB) RAM. Zope 3 as web server on Gentoo Linux (x86). Memory will be
> mounted as a local disk (tmpfs) and python and R binaries will by copy into
> RAM. Python and R are compiled with ICC compiler.
>> Is data being accessed from a database which
>> may influence the timing?
> R will connect with Tree Data Server (database):
> Data can be keep in memcached: http://www.danga.com/memcached/
> So I think there is very small influence on the timing.
> And I think about quantmod as a R chart package:
> The main problem with R is that, R uses x11 server to render images and this
> way is very slow and I thing that faster would be the Cairo (or GDD) R
> package but I haven't tested this yet.
> R engine is very universal and this is the strong of R choice in this kind
> of web service... then begin write new (faster?) c/c++ plugin. What you
> daniel cegielka
> -----Original Message-----
> From: jim holtman [mailto:jholtman at gmail.com]
> Sent: Monday, June 16, 2008 1:55 AM
> To: Daniel Cegielka
> Cc: r-help at r-project.org
> Subject: Re: [R] R as chart engine in web-service
> Can you put some criteria on 'better'/'faster'. What throughput are
> you expecting? How many requests per second will you have to handle?
> What is the timing of your current method? What size/type machine
> will you be running on? Is data being accessed from a database which
> may influence the timing? So you need to give some specificity of
> what your requirements are.
> You can probably get faster (but maybe not better) with a plug-in, but
> I would assume that it would also be more effort.
> On Sun, Jun 15, 2008 at 6:39 PM, Daniel Cegielka <domibre at gmail.com> wrote:
>> Hi R-users
>> I think about some web service with stock charts and I plan use python as
>> web framework with R as mathematical and chart engine. I use rpy to
>> R with python. It works good.
>> Is it good idea to use R as chart engine? Maybe it's better (faster) to
>> c/c++ plugin?
>> R-help at r-project.org mailing list
>> PLEASE do read the posting guide
>> and provide commented, minimal, self-contained, reproducible code.
> Jim Holtman
> Cincinnati, OH
> +1 513 646 9390
> What is the problem you are trying to solve?
+1 513 646 9390
What is the problem you are trying to solve?
More information about the R-help