|
|
| GeoCommunity Mailing List |
| |
| Mailing List Archives |
| Subject: | RE: [gislist] GML too Bulky? (was google maps and OGC Testing) |
| Date: |
02/27/2005 10:55:00 PM |
| From: |
Anthony Quartararo |
|
|
Carl,
Point taken about the specifications, pretty dry read of the UL testing of
toasters and the specifications for heating elements. I'll cry Uncle on
that one, however, perhaps it's a matter of semantics. Whatever it may be
called, what I and Gregory and likely others are looking for are the
real-life conditions of testing environments and what all is involved:
essentially the entire metadata for the testing, NOT the conformance tests
used by OGC to give a thumbs-up, thumbs-down on any given component by any
given vendor, that's academic. You cite the example below, and that's an
interesting start, but how does 35,000 responses per minute fit into the
overall test: what was being tested, what were the expected results, what
data was involved, network connections, what other "parts" were involved.
This was a server test no, well, that may be ok, but what about the client
side metrics as well, I mean, that's really what matters isn't it ? If a
server processes a request in 5 ms, but the client (and user sitting at the
client) has to wait 100 ms for that single request, is that good ? Compound
that and really dump real-world conditions into that test and it could be a
recipe for "churn". The nuggets of performance metrics are appreciated,
would be great if OGC as an organization were to publish some complete and
unedited tests for review by the community at large. This would go a long
way towards greater acceptability and credibility. Thanks.
Anthony
-----Original Message-----
From: gislist-bounces@lists.geocomm.com
[mailto:gislist-bounces@lists.geocomm.com] On Behalf Of Carl Reed
Sent: Sunday, February 27, 2005 7:50 PM
To: ROSS,GREGORY K: gislist@lists.thinkburst.com
Subject: Re: [gislist] GML too Bulky? (was google maps and OGC Testing)
Ahh, but not for the standards that make these appliances all work :-) What
folks want the performance statistics for is the appliance - not for the
standards . . .
That said, I happen to know of a WMS implementation (server side) that has
been stress tested to 35,000 responses per minute. Not bad. Uses Enterprise
Java beans to wrap the WMS response interfaced to a Tibco bus into a small
server farm.
Cheers
Carl
----- Original Message -----
From: "ROSS,GREGORY K" <GKROSS@mail.ifas.ufl.edu>
To: <gislist@lists.thinkburst.com>
Sent: Wednesday, February 23, 2005 7:46 AM
Subject: RE: [gislist] GML too Bulky? (was google maps and OGC Testing)
I think Anthony's point here is that "plenty fast enough for many jobs"
will not cut it in the long term. Those who invest their money in such
things are going to want "performance statistics." Whether that is a
valid concern is not important. That investors WILL want them is
important.
And to answer Carl's question about asking my appliance makers about
"performance statistics", it happens all the time. What is the gas
mileage of my new car? How many pieces of toast can my toaster cook and
how fast? What is the speed of my computer? What wattage is this light
bulb? All are performance statistics. And the people whose money is at
risk in business want that information printed ON THE BOX.
Gregg Ross
-----Original Message-----
From: gislist-bounces@lists.geocomm.com
[mailto:gislist-bounces@lists.geocomm.com] On Behalf Of
JeffreyGHarrison@aol.com
Sent: Wednesday, February 23, 2005 9:37 AM
To: bthoen@gisnet.com: gislist@lists.thinkburst.com
Subject: Re: [gislist] GML too Bulky? (was google maps and OGC Testing)
Bill,
My position is that GML is plenty fast enough for many jobs. One size
doesn't fit all though.
I would also point out that the community needs to look at things like
Binary XML and such because handling text strings is not always optimal.
That said, GML is plenty fast enough for many jobs.
Regards,
Jeff
In a message dated 2/23/2005 8:24:31 AM Eastern Standard Time,
bthoen@gisnet.com writes:
On Wed, 23 Feb 2005, Jeff Harrison wrote:
> Interoperability Testing has demonstrated the network performance of
> the specifications, so let's get beyond that. Oh, somebody will
> probably say that WFS doesn't work and that GML is too "bulky" to
> "perform" well-- Sorry, that's not true either, it's works just fine
> for me, and it's getting even better with compression, etc.
I'm very curious about this. How can geometries stored as plain text
strings possibly be faster than the same objects stored as binary data?
Even with compression, you have the cost of decompressing on top of
string parsing. Or is the point more that GML is "plenty fast enough"
and the simplicity and interoperability of plain text is well worth the
small performance hit?
- Bill Thoen
_______________________________________________
gislist mailing list
gislist@lists.geocomm.com
http://lists.geocomm.com/mailman/listinfo/gislist
_
|
|
Sponsored by:
For information
regarding
advertising rates
Click Here!
|
