compare CIP with Token Ring

[mjwo]

Hello,

2600----7507--cip-------------------MainFrame

| |

(dlsw)----TokenRing----3745 --------

Suppose, when remote SNA device want to connect to mainframe through DLSW, CIP reponse should be faster than TokenRing right?? But we found most cir go through TokenRing!! Should we need to tuning MainFrame configuration??

Regards,

Jwo

[emazurek]

Hi,

Your network diagram wasn't very clear. Are you

saying that the remote DLSw peer(where the device is

at) connects to two data center DLSw peers where one

is a CIP and the other is an IBM FEP? Please clarify

your diagram.

But, normally, the circuit will be given to the first

device(CIP or FEP) that responds to the Test poll.

This shouldn't have anything to do with the mainframe

because te CIP responds to the Test poll locally.

That is, it responds to the test without the

mainframe.

You need to verify that the XCA major node that

activates the CIP has enough lines/PUs defined so

that it can accept more incoming connections. Also,

you could check to see how loaded the CPUs are in

both the 7507 and the CIP. You get that information

via the show process cpu command(for the router) and

the show controller cbus command(for the CIP).

If you want to force the connections to the CIP then

you can specify a lower cost in the DLSw remote-peer

command(assuming you do have multiple remote peers.

When you do this you will also need to specify dlsw

timer explorer-wait-time 3. Both of these are

specified in the remote side.

Please give me more information such as a network

diagram, dlsw configs from the routers involved,

sho dlsw peer and show dlsw circuit and I'll be glad

to continue to assist.

Thanks,

Ed Mazurek

[mjwo]

Hello,

here is centrol router and remote router configuration.

centrol router:

source-bridge ring-group 100

source-bridge transparent 100 13 1 1

dlsw local-peer peer-id 10.1.253.235 promiscuous

dlsw icanreach sap 0 4 8

dlsw bridge-group 1

!

interface Loopback0

ip address 10.1.253.235 255.255.255.255

!

interface Channel1/0

description connect to production

no ip address

no keepalive

csna 0110 01

csna 0110 00

!

interface Channel1/2

ip address 10.1.98.1 255.255.255.0

no keepalive

max-llc2-sessions 6000

lan TokenRing 1

source-bridge 98 1 100

adapter 2 4000.3745.0001

name cxp401

adapter 3 4000.3745.0008

name cxp400

!

interface FastEthernet4/0/0

ip address 10.1.99.235 255.255.255.0

bridge-group 1

!

interface TokenRing6/0/0

description connect to NCP

no ip address

early-token-release

ring-speed 16

bridge-group 1

source-bridge 9 1 100

source-bridge spanning

!

remote router configuration:

source-bridge ring-group 2000

dlsw local-peer peer-id 10.1.253.39

dlsw remote-peer 0 tcp 10.1.253.235

dlsw icanreach sap 4 8

!

interface Loopback0

ip address 10.1.253.39 255.255.255.255

!

interface Serial0/1

description Connect to MSD of AT&T /w 6 ports

no ip address

encapsulation sdlc

no ip mroute-cache

no keepalive

nrzi-encoding

clockrate 9600

sdlc role primary

sdlc vmac 4000.2610.3900

sdlc address A2

sdlc xid A2 01703911

sdlc partner 4000.3745.0001 A2

sdlc address B1

sdlc xid B1 01703921

sdlc partner 4000.3745.0001 B1

sdlc address C3

sdlc xid C3 01703901

sdlc partner 4000.3745.0008 C3

sdlc dlsw A2 B1 C3

!

Regards,

Jwo

[jishi]

The reason the most circuits go through Tokenring is because the explorers from TokenRing side get back to the remote dlsw router faster than the explorer from CIP side. That means you mostly likely have a faster WAN connection between remote dlsw router and the the TokenRing side. If you prefer the CIP, you have to config "explorer-wait-time" and give lower dlsw peer cost to the CIP side.

Here is link for controlling dlsw peer selection:

http://www.cisco.com/en/US/tech/tk331/tk336/technologies_design_guide09186a0080237a5e.shtml#wp16269

Thanks

Jing

[mjwo]

Thank you for your reply!!

TokenRing and CIP are in the same 7507 router.Remote dlsw connect to the 7507 router, then go through CIP or TokenRing to Host.

Bus as I know, CIP speed should be faster than TokenRing. Why the explorers from TokenRing side get back to the remote dlsw router faster than CIP side??

Should we need to tuning Host configuration??

Thanks

Jwo

[jishi]

I don't think you can tune the host config since the explorer is answered by the vtok adaptor or the NCP. The reason CIP reponses slower than tokenring is because the NCP is on the physical ring, NCP response it immediately, no software processing in between. But for the CIP, it goes the vtok adaptor, and involves software processing, so the explorer gets back little bit slower.

If you use seperate virtual ring group for CIP and the tokenring, and create a virtual-tokenring interface, bridge into the virtual ring which the tokenring is bridged into, you might be able to use "dlsw ring-list" to restrict the ring number only from the virtual ring which the CIP is bridged into and the ring number which the virtual-tokenring interface is on.

Basically you are adding a virtual ring and a bridge between the NCP and the router, hope it will slow down the explorer coming from the NCP.

Regards,

Jing

[jishi]

There is a parameter on the NCP to delay the response to the explorer. You can modify that.

[mjwo]

Hello,

What's the parameter on the NCP to delay the response to the explorer???

Thanks

Jwo

[jishi]

What exact model of NCP you have? It should be in the line defination.

[jishi]

Hi,

Sorry for the confusion. I was trying to say the delay parameter is in the line defination.

Please take a look at the following link for the "balance" parameter.

http://publibfp.boulder.ibm.com/cgi-bin/bookmgr/BOOKS/cxdg7009/2.3.2.2?ACTION=MATCHES&REQUEST=auto+balance&TYPE=FUZZY&SHELF=&DT=20000620173822&CASE=&searchTopic=TOPIC&searchText=TEXT&searchIndex=INDEX&rank=RANK&ScrollTOP=FIRSTHIT#FIRSTHIT

http://publibfp.boulder.ibm.com/cgi-bin/bookmgr/BOOKS/cxdh1009/2.23.15?DT=20000531143225#SPTGBAL

If you set the factor to 1, the TIC might delay sending a response to the explorer.

Thanks

Jing

[emazurek]

Hi Jwo,

As I indicated before there is really nothing to

configure in the CIP to allow it to respond quicker

than the TR attached FEP.

One thing I noticed in your config that doesn't look

right is that at the central site router you have a

combination of transparent, translational and

source-route bring configured. I doubt you need all of those and it could be affecting you.

You have "bridge-group 1" configured under the

TR 6/0/0 interface. Do you know why you have that?

Since you already have SRB

configured(source-bridge...) you generally don't

need that. What that does is configure transparent

bridging under the TR interface. Since you already

have the TR interface bridged into the Fa4/0/0 via

the "source-bridge transparent 100 13 1 1" command

you don't need that.

Also, are there devices on the Fa4/0/0 interface

that need to be bridged into the CIP and/or TR 6/0/0

interface? If there aren't any then you can remove

the "source-bridge transparent 100 13 1 1" command

as well. Be careful here though, you may need it.

Finally, you can configure "dlsw load-balance" at

the central site router to cause DLSw to load

balance between the available local paths. This

should cause an approximate 50/50 balance between

the FEP and the CIP.

Please let me know.

Thanks,

Ed Mazurek

[mjwo]

Hi Ed,

There are SNA gateway and AS400 in on the Fa4/0/0 and need to be bridged into CIP or TR6/0/0 interface.

So we need translation bridge. By the way, I am not sure why has "bridge-group 1" configured under the TR 6/0/0, I would remove it and try.

Very thanks for your help!!

Regards,

Jwo

[rsissons]

I guess it depends how heavily loaded the total path via the CIP is as compared to the path via the token ring and the FEP since the session will always be established via the first to respond. I am not aware of any parameters that are available on the mainframe to delay the reponse however.