Conference 7.286::fddi

749.0. "PROBLEMS WITH RSM BACKUPS AFTER FDDI!!" by TPSYS::BRITT (The Node is BOND...JMSBND) Fri Oct 16 1992 12:32

Help!	S.O.S!!	   URGENT PROBLEM WITH RSM BACKUPS AFTER FDDI!!

Has anybody experienced a problem with throughput running RSM after cutting
over to an FDDI ring?  We are having major problems running RSM backups after
implementing FDDI into our network topology. I've researched all the related 
notes in the RSM notes file and also the FDDI notes file and haven't found
anything that can help. I'm also having major problems getting support out
of Colorodo and also from my local Field Service branches. The following is
a description of all the related components. I am entering this note here in
the RSM notes conf. and also the FDDI conference. This is a critical problem
that I need to have resolved asap if possible.


VAX HARDWARE:

We have a 3 node CI cluster configured as an RSM backup management server.
The nodes consist of an 8800 (2 CPUs) and 8550 and a 6420. Also configured
are 4 HSC70s and about 48 disks (RA90s, RA92s). 

VAX SOFTWARE:

The TAYRSM backup cluster runs VMS V5.5 and RSM V2.3. 


BACKUP OVERVIEW BEFORE FDDI:

We normally run between 120 and 130 network backups each night, both fulls and
incrementals. Backups run between 8:00 p.m. to 7:00 a.m. and complete without
any major problems. Our backup cluster sits on the main site network backbone
segment (copper 10mb) and backup clients hang off 4 other copper 10mb segments.


FDDI IMPLEMENTATION:

While running backups we realize 80% - 90% utilization on the site backbone
segment and have decided to upgrade the backbone to fiber.

We installed a DEMFA module into the VAX 6420 and replaced the main site back-
bone segment with an FDDI fiber ring using a DECconcentrator and 2 DECbridge
620's. Now we have 2 CI nodes from the backup cluster running on a copper 10mb
segment and the VAX 6420 connected directly to the DEFCN. This is a known 
workable configuration and has been verified by several network support 
persons.

THE PROBLEM:

After bringing up the MFA-0 100mb ethernet circuit on the VAX 6420 we notice
a 60% - 75% DROP in throughput ACROSS THE WHOLE CLUSTER! The backup entries 
dont fail for any reason they just take forever to complete and the disk IO 
is almost nothing!  

If we switch back to the 10mb ethernet controller things are back to normal. 


WHAT WE'VE CHECKED SO FAR TO DATE:

We've verified hardware revision levels on the recently purchased FDDI network
equipment and firmware revision levels on the recently purchased DEMFA. 
We've tried replacing the FXDRIVER with one from VMS 5.5-2 and did not see 
ANY improvement. The DEMFA is running firmware 1.4.

I have DECps performance, historgram, and analysis reports for several days
before and several days after the FDDI implementation but do not see anything 
that would point it to a software problem. And the reports do verify the drop 
in throughput. I have a directory with all these reports. Also included in 
that directory is system configuration,  XFDRIVER image info, and other 
cluster specific data. We installed the fiber on Saturday the 10th of October
so all reports with a title of 050ct92 are for before FDDI and all reports
with a title of 15oct92 are after FDDI. The directory is not accessible now
but I will make public upon request.



		Larry Britt
		TP East Computing Services
		TAY1-1
		TPSYS::BRITT
		227.4442

    

	If you have any unused Ethernet ports on your Decbridge 620s, make sure
that they have a test connector on them. If any are open the self test that
runs every few seconds will cause the kind of performance loss that you are 
seeing. This has been discussed elsewhere in this conference.

Joe

    
    Yes we do have a loopback connector on the 1 unused port on the 
    DECbridge 620. 
    
    The orange (left) light on that port is blinking constantly. Is
    that normal for a port that is unused?