|
|
| GeoCommunity Mailing List |
| |
| Mailing List Archives |
| Subject: | RE: [gislist] GIS and computer configurations |
| Date: |
10/14/2004 06:35:01 AM |
| From: |
=?iso-8859-1?Q?Luis_Gon=E7alves_Seco?= |
|
|
Hello,
I'm completely agreed about your statements (Bill Huber and Richard =
Nicoll)
That should be the perfect solution for multi-users. I have a Gigabit
network, because we have 25 GIS persons working (server-client).
That was the doubt I had initially, but he said: "I am the GIS person" =
so I
interpret that he is the only person o uses the server. And if is that =
so,
in my opinion he doesn't need a gigabit structure.=20
The memory suggestion it was only in the local user perspective. In =
contrary
of free software, proprietary software consumes many resources and if he
works with many applications at the same time, maybe can help a little.
PS: thanks Richard for the whitepaper, it's very interesting.
Kind regards,
Luis Gon=E7alves Seco
-----Mensagem original-----
De: gislist-bounces@lists.thinkburst.com
[mailto:gislist-bounces@lists.thinkburst.com] Em nome de Richard Nicoll
Enviada: quinta-feira, 14 de Outubro de 2004 11:02
Para: Quantitative Decisions: gislist@lists.thinkburst.com
Assunto: RE: [gislist] GIS and computer configurations
I completely agree with Bill's calculations, it is common for hardware
manufacturers to overstate the performance of their products (or perhaps
the source of the problem is in how standards are defined?). Other
examples include the 802.11x wifi and bluetooth speed ratings where
again the typical performance is less than half the officially stated
speed.
You do not mention whether the ArcGIS software is installed on your
server (ArcGIS Server) with your workstation acting as a thin client, or
whether you have the ESRI package on your local machine (perhaps with
the license manager on the server). Either way the spec of your
workstation should be enough to run with the applications, and I'd agree
with Bill that it is likely the network link which is slowing things
down (IP packet transmission).
The most obvious upgrade that you could perform to enhance performance
(other than chucking more RAM at the situation which may not help much)
is to upgrade your network connection between the server and workstation
to a gigabit LAN (depending upon your office environment and company
preferances this may not be possible). This may require purchasing new
NIC's or your current motherboard platform may support this standard.
There is an excellent whitepaper regarding system design strategies at
http://www.esri.com/library/whitepapers/pdfs/sysdesig.pdf (a bit lengthy
but section 3 could be of use).
In the long term I'd suggest you conduct a comprehensive review to
establish what is required from your GI roll-out, whether datasets are
being stored in the most effective formats and locations, could a linux
platform be of benefit, etc.
Richard Nicoll
-Casella Stanger, UK
-----Original Message-----
From: gislist-bounces@lists.geocomm.com
[mailto:gislist-bounces@lists.geocomm.com]On Behalf Of Quantitative
Decisions
Sent: 13 October 2004 20:32
To: gislist@lists.thinkburst.com
Subject: RE: [gislist] GIS and computer configurations
At 08:57 PM 10/13/2004 +0200, Luis Gon=E7alves Seco wrote:
>Depending of your network traffic if you have a 100 Megabits network
>connexion that means you can transfer, more and less 12.5 Mbytes (data)
per
>second. So I think it's enough for the size of your images.
That's the right calculation, and correctly performed, too, but to get a
more realistic result, one has to accommodate differences between
nominal=20
values and actual values. A 100 Megabit/sec network requires not 8, but
about 9.5, bits to transmit one byte (there are check bits and some=20
overhead involved). The actual rate of this network will top out at
half=20
its nominal rate, but typical speeds are more like 20 megabits per
second=20
(even for an unloaded network). If you are running through a hub rather
than a switch, this bandwidth must be shared by all traffic passing
through=20
the hub, further decreasing the network's capabilities for fast
transmission.
Accounting for these realities indicates the actual data throughput on
the=20
network is unlikely to exceed 5 MB/sec, requiring four seconds at least=20
(and more like 10 seconds) to transfer just one 20 MB image.
For whatever reason, ArcView also tends to perform slowly--more slowly
than=20
these kinds of calculations usually suggest--when accessing data over
networks.
My basis for these statements consists of having monitored system and=20
network usage and performed formal timing studies on AV 3.1, 3.2, and
3.3=20
on machines running Win NT, 98, 2000, and XP using Intel and AMD=20
architectures with chip speeds ranging from 200 MHz through 2.4 GHz on a
100 Mb/sec network.
In response to the original query, consider maintaining local copies of=20
large geographic datasets (such as your images). That might not
completely=20
solve the problem
|
|
Sponsored by:
For information
regarding
advertising rates
Click Here!
|
