Newbie needs to see a large project

Alan Gauld alan.gauld at btinternet.com
Wed Oct 8 19:10:05 EDT 2003


On Tue, 07 Oct 2003 19:13:38 +0100, Alan Kennedy
<alanmk at hotmail.com> wrote:

[Lots of valid points about maintainability snipped]

> Also, your management team have to understand that CPU cycles are
> vastly cheaper than programmer cycles. 

BUt here I disagree, it depends entirely on the deployment scale.
If an app requires a PXC upgrade to, say 5000 usrs thats easily
$2.5 million US, which buys a lot of programmer cycles. Most of
the projects I work on tend to require dedicated server resources
and with a Sun F15K coming in at around $1million the hardware
and network links nearly always outweigh the programmer cost.

But on small scale projects utilising cheap PC based servers 
( inc Linux etc) then of course the programmer costs start to
come to the fore, but its not always the case.

> PC? Just spend the money on a 3GHz PC instead, and save thousands of
> currency.units in programmer costs.

Yes, but if it needs a 90 x 1GHz Sun box... Or even a blade
computing solution with 50-100 blade servers...

[ more good stuff on maintenance issues]


> And as for the projected number of users (50), I'm hard pressed to
> think of many business data-processing applications where a single CPU
> running a python app would not be up to the task. 

I'm inclined to agree except we don't know how much background
batch work is involved. Some of our apps have few users but they
simply set up the data which then gets crunched for 30-40 hours
of solid CPU crunching...

> surprised if the performance, in a well designed python system, would
> not match or exceed that of a FoxPro system. 

And here we really get to it. Yes, I agree, in this particular
case, if it runs OK on Foxpro then Python shouldn't be an issue.

> A day or two spent optimising a relational database (e.g.
> normalisation, indexing, precompiling stored procs, etc) can often
> achieve far greater performance improvements than spending that same
> time optimising other code.

Here too I heartily concur. The two biggest performance issues in
any distributed system will likely be the RDBMS and the network
connectivity.

Alan G.
Author of the Learn to Program website
http://www.freenetpages.co.uk/hp/alan.gauld




More information about the Python-list mailing list