Why is the network so slow at work?

[duragezic]

Argh especially lately it all of a sudden gets really slow when I'm entering items or whatever in the computer and it really sucks when you have to do a lot and it runs so slow. It just has big pauses in between stuff.

I don't know much at all about networking but I'm not sure why it all of a sudden gets slow for awhile or maybe the whole day then the next day it can be fine. The computers are all celeron 500mhz, 64mb ram but considering it runs good usually, I don't think it's that. We use Red Wing Business software and uh a Novell server.

The fricking morons who come from the comp places always a) replace the hub (it's been replaced like 3 times in 6 months) b) run virus check c) both. Last time he replaces the hub with a 10/100 one and is all like "this should speed it up nicely". Yah too bad we don't have the other equipment to run at 100mbps. I bet they just charge alot everytime they come and they never fix a damn thing.

But anyway, what would possibly make it suddenly slow at times?

 

[jkukowsk]

Perhaps all the computers on the network all trying to do something on the network at one time. :Q

 

[ojai00]

Like jkukowsk said, it could be that everyone on the network could be doing something on the network at the same time. It could also be that your PCs have so little RAM in them. It could also be a bunch of people not working and surfing the internet and downloading stuff. Hope this helps.

 

[Tallgeese]

<< But anyway, what would possibly make it suddenly slow at times? >>

Not enough specific details to know really.
However, those consultants should be fired...they obviously aren't troubleshooting the problem at all, simply tryin random "bandaids" hoping one will do the trick.

In a Novell environment, major sources of slowdowns could be:

* Low available memory in server (Netware performance relies on caching, which can't happen if no available memory).
* High amount of purgeable space

This one may actually be interesting for everyone. By default, on a Netware server "volume" (logical storage space) files aren't immediately deleted, much like the Recycle Bin in Windows (or the Trash on MacOS for all you zealots out there). At that point, deleted files can be recovered by using the "salvage" command. Depending on settings which can be modified, the server will retain deleted files until a "purge" command is run on that Netware volume. Newer versions help automate this process, but a volume that has been a long time without a purge can exhibit slowdowns (mainly related to the sheer amount of memory required to track "deleted" files).

* Excessive number of files in a single directory

Known issue with Netware. Performance-wise, there is a finite limit to the number of files that can reside in a single directory before server performance is SEVERELY impacted (like, night and day difference).

Still, the possible causes are so vast, that only through careful observation and analysis could the problems with your network be found. And it doesn't sound like you are getting that from your current consultants.

g/l

 

[RagManX]

<< The fricking morons who come from the comp places always a) replace the hub >>


That's a possibility right there. Depends on how many machines are on your network, how much is pulled off the servers, and how good the NICs are in the PCs. Also, could be your server is underpowered and gets bogged down when a lot of requests are coming in. Had a recent upswing in demand at work? More work could mean more server access, could mean slower performance.

RagManX

 

[KerryMunro]

Could drive mapping be a problem? My sister had this problem on her network. Whenever she was working on 1 machine, the files would be copied to every HD on the network.
Slowed things down just a bit....

 

[JustinLerner]

Could be many things. I would hope and assume your company is using a switch, not a hub. Yes, a hub will increase collision problems which a good switch will decrease. (A hub broadcasts every request to every Ethernet port. A switch is smarter and only broadcasts the first time after power up or for the first unique address and 'stores' the info on the switch.)

Of course, the server resources are probably still limiting factors. If many users attempt to access the server simultaneously latency problems could increase unless the Novell server has the capability to process and reply quickly with multiple users (like previously mentioned, enough memory and processing power). Multiple NIC's multilinked together for Etherchannel (802.3ad) bonding would work to increase the server NIC aggregate bandwidth, but not the processing power of the server as previously mentioned. Etherchannel works with Cisco and Intel or other switches incorporating 802.3ad and the NICs must be able to multilink, preferably provided software by the manufacturer. I tend to like Intel, but 3Com NIC's are good also.

Shouldn't your network service company be running a trap, network monitor or diagnostics to find out what is causing the network problems?

 

[cmv]

Suggest to your company that you call a different company! Why deal with the same morons again and again?

 

[ojai00]

<< Suggest to your company that you call a different company! Why deal with the same morons again and again? >>



They can call a bunch of fellow ATers

 

[BIGMACC]

I not that familar with Novel, but what kind of network are you running? Static ip, where each computer has it's own ip address, or DHCP where a server or network appliance is serving the ip address to each computer. Sometimes on a DHCP network, if a computer does not receive an ip address, from the DHCP server, that is in the same network range (i.e. the network range is 192.168.0.0 and your computer is getting an ip address of 169.216.5.0) then the computer will runn very slow connecting and communicating across the network. This is rare, but it does happen. Shutting the computer down and then restarting will usually solve the problem. If you just restart the system, it will not get a new ip from the server, the server will just give the system the same one it had before.

Also, check to make sure that you Ethernet cards in each system are set for auto sense or 100baseT. If they're set for 10baseT, and your hub is 100baseT, then your running slower than you r network is capable of running.

And, a switch is better than a hub. Faster--Smarter!!!

Good Luck!!!

 

[Garion]

I'd be willing to bet that TG hit it right on the head. Novell is not necessarily a RAM hog, but the more you have it in it, the merrier. This is especially true if you're running other NLM's like backup, antivirus, etc. It's been a LONG time since I've run any Novell, but there are some pretty basic mentrics you can look at on your server (Syscon somehwere? TG?) that show you pretty quickly if you need to add RAM or not. If in doubt, add a bit - It's cheap and usually is a pretty good bump in performance.

How many people do you have? Nowadays a 10/100 hub is pretty slow, but if you really have only one server that everyone uses it's probably OK. A switch would be nice, just for better network performance, but it's likely not required.

One danger, however, is that stepping above 10BaseT can cause issues. for example, a lot of NIC's don't do a good job of autonegotiating very well. They end up deciding to run at 100BaseT, full duplex and your hub only works at half duplex. Things function this way, but it causes severe bottlenecks and can result in horrid performance. Only way to really see this is with a sniffer, unless you can get on the server console and look at the network stats to make sure there's no errors.

One last word.. Server and network resources are not infinite. A lot of people seem to think that SERVER is this massive chunk of iron that will do everything and last forever. I've seen far too many "servers" that were home-built using a single old, slow IDE drive. Novell is a GREAT OS, but even it has it's limitations. It might be time to look at your server or your network infrastructure (the hubs) and make sure that it's up to snuff.

If you're not happy with your current support company, post what city you live in here and I bet that you can find at least one reputable reference, if not someone willing to do a bit of consulting on the side to help out a fellow AT'er.

- G

 

[Tallgeese]

On a Netware console, LOAD MONITOR (or just MONITOR in newer versions) will bring up a nice "graphical" (for circa 1986) console that offers BOATLOADS of info, including:

* Installed RAM
* Original Cache Buffers (RAM available for buffering after system loads)
* Available Cache Buffers (RAM left on server currently)
* Long-Term and Short-Term Cache Hits (good indicator of low RAM conditions)

BTW: Novell's Knowledgebase has some excellent documents that offer troubleshooting checklists for resolving server performance problems. If your consultants were worth half a d@mn, they would know that already and have used that resource, rather than taking potshots in the dark (all the while bleeding your company).

 

[Tallgeese]

<< I've seen far too many "servers" that were home-built using a single old, slow IDE drive. >>

You and me both, Big G!
In fact, disk channel bottlenecks can also cause out-of-memory conditions on Netware servers. What happens:

As the server fails to keep up with the read/write requests, it buffers that request in memory (cache, in Netware parlance). If available memory (available cache buffers) drops precipitously, it can hit a critical point where the network starts to fail.

Had this at one of my previous jobs. An engineering group was using some third party application to stream a LARGE amount of data to/from the primary server periodically. This application didn't allow the server to finish a write until the entire operation was complete. Which caused the condition above. Which led to the following:

* CFO unwilling to pay to upgrade server.
* Engineers unwilling to modify their procedure.
* Me puts out fires for a good while, and starts searching for a better company that "gets it" about upgrading.

 

[Garion]

<< * Available Cache Buffers (RAM left on server currently) >>



Ahh. That's the one - I seem to recall that once it's below 50% you've got a problem. Down into the 30's and they are BAD problems. Anything below that - You're lucky the thing still runs. Optimal is 60%+ free. Sounds like a lot, but it's really not.

- G

 

[Cable God]

TallGeese makes some good points. Just breakdown and buy cheap 10/100 NIC's for the LAN. Some are as cheap as $9 a piece. Also, being a hub network, it will be slower than a 10/100 switched network. 10MBPS split among among quite a few PC's will drain down to nothing quick, really quick. Gigabit Ehternet to the rescue...

 

[Garion]

Actually, if your server access is occasionally slow, adding 10/100 NIC's to the workstations won't do anything and, if anything, could have a negative effect.. If one PC is able to hit the server at 100BaseT, it could do a lot more "damage" than a 10BaseT machine could.

I just realized something - You don't ever say how fast the NIC is in your server. Have you verified that you've got a 100BaseT NIC on that box? If not, that's definitely a bottleneck.

- G

 

[skimo]

A bad NIC can be ringing and bogging the network down. Any true network guru would put a sniffer on the network and track traffic Its not a tough job. We had a bad NIC (3com) doing exactly that. the network CRAWLED. A simple check is to unplug one machine at a time and see if thruput improves. Also, if more than one swithc/hub is in the network, a bad switch or hub can do the same thing. We had a hub fail and it did similarly.

 

[Tallgeese]

The main issue in this thread is not what is wrong with the network. We're stabbing in the dark just as your current consultants have been, and in all honesty, we'd be lucky to find it anyway without having physical access to the systems.

The KEY ISSUE here is that your company is getting CRAPPY service from your current consultants, who DO have physical access to the systems and should be troubleshooting these issues PROFESSIONALLY (ie in a logical fashion) and not replacing components willy-nilly without gathering some info and eliminating possible culprits.

[rant]
I am AGHAST at some of the things we discover that new clients have been conditioned by their previous consultants to put up with, such as rebooting servers daily to "fix" server hangs that occur every 26 hours (like clockwork). They are amazed when, after contracting with us, stability becomes the norm, not the exception.

Rather than attack and resolve the problem, some consultants prefer to bandage the symptoms, which nearly always come back.
And which they will gladly "fix" again, for more $$$.

I get more pissed the more I read this thread.
It's doofuses like your current consultants that give those of us who care half a damn about the job we do a bad rap.
[/rant]