DB support: “The test that you did on the server versus the client proves that something is slowing down the response time from the client to the server which is not a DB issue, but a network issue.” Furthermore…. “Please send a reply back if it is OK to close the incident since this is a network issue and not a DB issue.”
We (system admin and I) took a laptop to the server room (which I wanted to do a couple of MONTHS ago[sub]“no reason to since the network is fine” :dubious: [/sub]) and hooked it right to the main switch (I’m not a network guy). In any case, this bypassed some other switches, hubs and routers.
Response time was what clients where getting untill the system fell on it’s face a few months ago. This really started about 10 months ago. Data that clients used to fetch in 2 seconds, now takes 10-18 seconds.
It could be network trafic I suppose, but we have not added any new users, DB’s or apps in that time.
Security/Scanning? Disabled it to test it (server and client). No change.
We have a 100mb pipe at our local facility. We have a 10mb pipe to another facility.
Here’s my WAG. An accidental change was made in the network configuration where our data is now being shiped out our 100mb line, down the 10mb line and then back to us. I’ll ask that Monday (might be a dumb question, but what the heck).
It’s doing just what it’s being told to do. So nothing is ‘wrong’ with the network. No lost packets, whatever. Did network really trace where the data is going?
Like I said, a WAG. If all the equipment checks out, why are our speeds a tenth of what they should/used to be (just a Rhetorical question)?
At least I finally have some hard numbers that I can replicate. Not just a “seems real slow”. And the other IS folks are wondering what the heck is going on too.
I’ll WAG back, admitting that I don’t know the first thing about your network. If you have hubs on the network, you must seek them out and destroy them. Maybe a switch failed and they replaced it with a hub, either as a quick repair or because someone wasn’t paying attention. (I can neither confirm nor deny the allegations that I have done that myself). That could account for performance suddenly tanking on one or more segments. Just a thought. Of course, if you imply that the network admin doesn’t know the difference between a switch and a hub, you better get used to the performance, 'cause he will never fix the problem or talk to you again.
eeehhhh. My memory is that the network admin at the time replaced the switch with a hub to make things better. Or so he said. He no longer works with us
I suspect the hub (whatever the heck it is) is still there. I have no idea. I just cut code and deal with data and users. The programmers are always the first line of defense when something is not working as it should.
What the heck is a hub compared to a switch?
eta - If I have to do every ones job I should at least have a clue.
A switch is a smart hub. On a hub, when traffic comes in one port is gets sent to all the other ports. A switch is smart enough to learn where the traffic is supposed to go and will only send it on that port. This leads to less traffic congestion.
Edit: By the way, I’m not even sure if anybody makes hubs anymore…
That sort of makes sense. Especially if we have a hub instead of a switch in front of the server. (I’m not a net guy). The last net admin guy said he replaced a switch with a hub. Sounded like it would be better to me.
:another shrug: What the hell do I know. That’s why I ask you guys.
We often have great gobs of data going on one way, (GIS info) and little bits like an email going another.
“hub” isn’t an acronym. It is the noun “hub” like hub of a hub & spoke arrangement.
Each machine (client or server) is a spoke & they each connect directly to the hub. The hub takes incoming traffic on one wire from one machine & retransmits out all the other wires to all the other machines. This is true for each machine on each connection. None is special and there is no idea of a front- or back- side. it’s just an amplifier for a party line with all the machines yakking away & ethernet’s collision resolution on each machine’s NIC trying to handle any collisions.
Hence the effect that if one machine is particularly busy (ie a server), then all the other legs get saturated. Or that traffic from many clients out to the internet can also saturate the link to an internal server.
In a true peer-to-peer environment where nobody is using much bandwidth, hubs work OK. Think 10 desktop PCs in a Windows Workgroup. Just plug and play baby!
Their chief virtue 15 years ago was because they were utterly stupid (just a handful of transistors per connection) they could be very cheap.
One other neat effect of a hub is that everyone can monitor everyone’s traffic. All of the packets are transmitted, so you just need to have a card that’ll record every packet instead of just the ones that are meant for it.
I can’t think of any cases in which a hub would be preferable to a switch, although I am not a hardcore networking guy. As others have noted, hubs are an older and cheaper device, but switches do pretty much everything better. I doubt anyone makes hubs anymore, but I’m sure lots of them are sitting around in network closets or storerooms as “spares”. Keep in mind, though, that some people get sloppy and use the terms “hub” and “switch” interchangably. If your previous network guy said, “I replaced the switch with a hub,” odds are he was using the terms deliberately. But if he said on one occasion, “the switch is broken”, and on another, “I put in a new hub”, he might have just misspoken. I’ve heard several network guys use these terms lazily.
I’m really curious now – let us know what you find out!
Real good info everyone. I think our previous network guy said he replaced a switch with a hub. But that was months ago. And I also think that the guy was smart enough to recognize if it was a bad move.
Sounds like hubs are out. And the closer to the server on the network they are, the worse it is.