[Fwd: Re: Compression question / problem]

[Chipp Walters]

Alex,

Here's my math regarding the download time for a single 1 megapixel JPG
file * 50. A Single 1024X768 JPEG compressed at Quality 60 is over 100K.
Testing here it takes about 8 seconds on a cable modem to upload it to
an FTP server. This doesn't take into account the handshake stuff as I'm
using libURLSetStatusCallback to get feedback on when the file starts
uploading and then is finished.

So 8 seconds * 50 images = 400 seconds = 6 min 40 seconds. I really
don't think waiting an extra 10 to 20 seconds makes *that* big a
difference here-- and that is why I suggest making the decision based on
other criteria. For instance, what happens if you lose your connection
at 6 min 20 seconds? Do you want your user to start all over? I
personally would rather download each file independantly, keep track of
them then have the user only download files not yet downloaded if the
connection is broken. This is 'other criteria'.

Also, as Dave mentions, libURL can't do 'parallel' downloads, so you
don't get a speed boost you thought.

best,

Chipp

Alex Tweedly wrote:
> Chipp Walters wrote:
> 
>> It'll take only slightly longer to download 50 separate files than one 
>> large file. I think I would make that design decision based on other 
>> criteria than just time.
> 
> 
> I have to disagree with that  -  it may take more than "slightly 
> longer", depending on circumstances.
> 
> Multiple small files (assuming you don't parallelize them) will cost you 
> at least one extra round-trip time per file (and possibly more depending 
> on the server, firewall, proxy, nat box, etc.). So on a "typical" 
> client, open a shell and "ping" the server. Take the average round-trip 
> time reported, and multiply by 50 - that's the approx cost of multiple 
> small files.
> 
> You have a base time of around 10 seconds in this case (50 x 10k over a 
> 512K DSL connection), so you can tell whether or not it would be a 
> noticeable increase.
> 
> my house to my web server        round trip = 48ms, overhead = 2.4 
> secs.  10 sec vs 12 sec ? - not important.
> my house to runrev.com
>   round trip = 140 ms, overhead = 7 seconds.   10 sec vs 17 sec - worth 
> worrying about.
> 
> -- Alex.



-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.6.10 - Release Date: 1/10/2005