Slow Performance on NW51

[dvannoy]

I just put in a new NW51 box. clients have all been updated to the current version. comming from a NW42 environment I have noticed a serious delay when asking for services from the server.

Server side: I have increased record locking. have also increased packet receive buffers. I have followed all the rules from novell on how to tune up the server.

Client Side: I have done traces with packet burst OFF and ON..When doing a trace with Packet Burst ON, I show a serious amount of bursting across the LAN. The client will Burst up to 8 times to the server. the server will then only burst once back to the client. If I do the trace with Packet Burst OFF, I see no bursting, but performance is much slower. Also If I have Packet Bursting OFF and File Caching OFF I notice that It's much slower then If they are both set to ON.

At this point I cannot figure out if there is a memory leak in the server or if this is a client problem..

Any help would be appreciated

Thanks

 

[Beaghler]

Let's look at a couple of things: In Monitor, what are the "Original Cache Buffers" and "Total Cache Buffers"? Your Total Cache Buffers should be about 40% or better of the Original Cache Buffer setting. if less, you have a memory problem.

What are the "Long Term Cache Hits"? They should be about 85% or better. A lower # indicates a lot of disk reads.

From Monitor>Disk Cache Utilization, Short term and long term cache hits should all be 85% or better (I get worried when mine drop below 90%). LRU sitting time should be at least 30 minutes. Less indicates a memory problem.

If something looks like a memory problem, look at Monitor>Loaded Modules. This will list the NLMs in order of memory usage. DS.NLM will probably be 1st or 2nd, along with Groupwise if you are using it. I have found that in some cases, DS leaks like a sieve. Novell won't admit it, though. The later versions of DS dynamically use memory and are supposed to give it back. This doesn't always happen. If your DS is using in the neighborhood of 20-50,000,000 Bytes you're probably O.K.(My main replica server that minds around 200 servers is using 90,000,000.) Otherwise You'll need to reboot or SET DSTRACE=*. to reset the ds.

There are also a number of set commands that will optimize the server, but do not noticably affect the user speed.

Also, look at your replication and licensing schemes. If you do not have a local replica, response will be slower as it has to look somewhere else for authentication information. Same with licensing-logins will be slower if it has to walk the tree for licensing info. If the license is across the wire somewhere and the line goes down, your users are down as well.

Look at the backbone for bottlenecks. Look for sync problems in DSREPAIR.

 

[dvannoy]

OK,

Original cache buffers = 523,643
total cache buffers = 454,365

Long term cache hits = 93%

DS is the fourth NLM to load. Alloc Memory = 4,759,552

MM is the first using 12,619,776

DSRepair comes back clean. No Sync problems..

 

[Beaghler]

All that looks good. Let's look at replication. Do you have several partitiions, or are all servers in one big partition? You can use NDSMGR32 to check this. You want to have at least a Read/Write replica of the partition residing on the server.

Also, is your NIC set to autoselect, or at a fixed speed? Autoselect doesn't work all that well. Set the NIC to whatever your backbone will support- (10 1/2 duplex, 10 full, 100 1/2 duplex, 100 full duplex). Use INETCFG>BOARDS for this. Some NICs require certain packet sizes. Check this with the card's mfr.

What kind of server is it? Some mfrs (like Compaq and HP) put out new drivers and BIOS periodically. Does your current BIOS support NW 5.X?

What service pack are you at? SP5 and Compaq do not always work the best. There may be issues with other makes as well.

Is Client file caching set to off? Also, set set file and record locks to the max.

Another thought- Are your workstations connecting IP or IPX? When possible try to use IP. The clients will support both and connect IPX if they can't connect IP. For that matter, is IP bound to the NIC? If it is, do you go across a router or some other gateway? If so, is the next hop (gateway) set in INETCFG?

I kind of rambled on a bit, but wrote down ideas as they came to me. These are all things I have run into that contribute to network slowness.

 

[dvannoy]

Right now there is only one server on my LAN. Once this gets ironed out I will add two more.

Iam using a Compaq ML530 dual Xeon processors and 2GB ram.
I am running NW51 SP4..

Client side - File Cache is set to off.

Tryed both IP and IPX no improvement..
IP is bond to the NIC..I have two segments..Ethernet and Token Ring..The token ring segment travels over my Cisco Router..The server has one ethernet card.

I did try setting the 100 1/2 duplex forcing that on the switch as well...no change.

I am about to shut down the server and put in another NIC..right now I am using the onboard NIC from Compaq. I am going to put in a 3com and see if that makes a difference..

I think it's the NIC,, Even my printing gets delayed..dont know what else it could be..

Thanks

 

[Beaghler]

Some more suggestions- set the NIC to 100/full duplex- it sounds like your system will support it.

If that fails, swap the NIC. Make sure that you have the right drivers and get the loads and binds right.

Can you get the Token Ring out of the mix for a while? could be a hardware conflict.

If that doesn't help, call compaq. They have pretty good support. It sounds like that box is still in warranty. If it is the NIC, they'll arrange to have someone come out and fix it.

 

[dvannoy]

Iset the NIC to 100/full..there is no token ring NIC in the server..token ring segment gets accessed over the router via IPX just like if they were on the WAN..

I think it's the NIC..even my printing is delayed and opening excel and word docs takes forever..

Thanks

 

[dvannoy]

Beaghler,

Let me ask you, do you Jetdirect cards at all??

I just run a config page on a few of my printers and saw that there was over 600 burst errors and 90 lost frames.
do you think this is caused from a bad NIC?? it seems my lan is bursting all over..