Re: Overall networking issue...
- From: Jacob Tranholm <jacobtranholm@xxxxxxx>
- Date: Sun, 26 Oct 2008 17:21:38 +0100
Allen McIntosh wrote:
It all depends on hour network infrastructure. No single client is ever going to get more than 100 Mbit/s of course, but there might be ways to increase the aggregate throughput.
If you have a single large switch in the middle of the network, then you could upgrade the connection between your server and the switch. This might require new hardware on the switch.
You can do the same thing if your network has several switches at its core. Are the inter-switch links 100 Mbit/s or 1 Gbit/s? If the former and management is planning on upgrading them to 1 Gbit/s, are they planning to upgrade server links as well?
There are several switches (5 or 6 48 port rack-switches (they are some years old, but they still work)), but I have not checked which connection speed they use internally (between them self). Management isn't planning anything... This is not a big company we are walking about, it's a school, and we are working within a limited budget... And most of our students use their own computers and connect through the wireless network, where they also have the possibility of accessing the fileserver. And the stationary computers all use the fileserver for storing/reading user-files. And the fileserver is the major reason why we need a faster connection from the server-room to the switches/hubs/routers/whatevertheyare where every network cable is connected. - For the clients megabit should be enough.
Finally, your cheap 'switch' solution also has the possibility of getting aggregate speeds up over 100 Mbit/s. It would be simplest to use a router, not a switch: put the server on its own subnet, and the other subnets use different links via level 3 routing. If you only use a level 2 switch, you need to have some way of keeping level 2 loop avoidance from dropping all but one link. I'm not sure that your dumb switch can do this.
Thanks... This has given me ideas that I will try tomorrow. As for my "dumb switch" I was actually more worried about that it might bring the intire network down because it (in my mind) gave an alternate connection between the existing switches/hubs. And I am not sure they are configured to ignore this connection and only use their primary internal connection.
In my previous experiences with this I have left the network unchanged, and for instance added a network card with 4 ports, connected all 4 ports and have done a software bridging (from Linux or *BSD). But this was back when gigabit networking was much more expensive... But if I should use this approach this time, we would have to add more CAT-5 cables between the server-room and the switches, and there is a slight distance between the two. And this would only solve the problem for one server...
--
Best regards Jacob Tranholm <http://jtranholm.dk/>
Karl R. Popper: Observation statements and statements of experimental results are always interpretations of the facts observed.
.
- Follow-Ups:
- Re: Overall networking issue...
- From: Günther Schwarz
- Re: Overall networking issue...
- References:
- Overall networking issue...
- From: Jacob Tranholm
- Re: Overall networking issue...
- From: Allen McIntosh
- Overall networking issue...
- Prev by Date: Re: Overall networking issue...
- Next by Date: Re: Overall networking issue...
- Previous by thread: Re: Overall networking issue...
- Next by thread: Re: Overall networking issue...
- Index(es):
Relevant Pages
|