[twsocket] Re: Application freezing when getting data
Thu, 29 Jun 2000 14:33:17 +0300
He didn't say it was bad but I would :)
If you stick everything in memory you could grind your
application down to a crawling pace because the more
memory that is consumed the more hits will occur on
the swap memory. That's pretty important on NT's as
the page files can really take a beating if you hit
some limits. You can see it with NT PerfMon.
As a memory killer example you try reading TQueries
with large datasets on BDE and you'll kill NT.
(You can avoid it by setting the undirectional
property to false.)
Also if you try to send all your data in one chunk
you will find that you have no idea of how much data
would have been received until the end of the transmission.
That's good for your application and for your protocol.
> -----Original Message-----
> From: Wilfried Mestdagh [mailto:firstname.lastname@example.org]
> Sent: Πέμπτη, 29 Ιουνίου 2000 2:15 μμ
> To: email@example.com
> Subject: RE: [twsocket] Re: Application freezing when getting data
> Hi Rik,
> > I was curious as to why you think my method of sending the
> stream is bad.
> > It has all worked absolutely fine for me from the day I wrote it.
> Sorry, I did not mean it is bad. It is just that if you send
> the stream at
> once, TWSocket will buffer it all, and if it is a stream of several
> megabytes then it will allocate that memory. So depending on
> the size of the
> stream and how many streams are sended the same time it can
> be a lot memory
> > something I'm getting lucky and avoiding that I should be
> aware of, or is
> > it supposed to send faster in 4K blocks?
> I don't know if it's faster than sending all at once, but I
> did had long
> time ago some experiments sending blocks of different size
> and it takes out
> that 4K was best performance as sending buffer.
> rgds, Wilfried
> TWSocket mailing list [http://elists.org/mailman/listinfo/twsocket]
> Visit our website at http://www.rtfm.be/fpiette/indexuk.htm