Just need a little Cisco help to find out if my config is correct

[LastKnight]

My client is upgrading his frame from 128K to t1s for 8 locations, and dropping in a shared t1 to the internet at their main site. Their IPs are screwball, so I need to touch all pcs to use the 10 network instead of the non-internic IPs they are currently using. I have preconfigured the router, and just wanted to bounce my config off of someone else that knows ciscos to make sure I don't run into any problems Friday night while I'm working.

Single DLCIs at each location, so I need to use point to multipoint routing. Cisco 1601 routers with T1 cards in the Serial1 slot. I configure s1 on the dlci assigned for that location, and do a frame-relay map on serial1 to each individual network using frame-relay map ip 10.1.5.0 205 where the ip address is the class c of the network I am trying to reach, and 205 is the DLCI of that location. One frame-relay map command for each possible route. I'm also using the physical address, and not using subinterfaces. I believe this is correct as well.

My main question is, in order for the offsite locations to be able to access the internet, I need to set up a default route. I believe I should set up the default route to be the internal port on my firewall so that the data uses my frame relay map command to get to the other network, and then accesses the firewall, where the data shoots out to the world. I, however, haven't ever used point-to-multipoint before, and want verification before I modify all these router configs. Any other pitfalls I may run into that you may be aware of? Any help would be appreciated. Thanks in advance. If you want more info, let me know.

 

[spidey07]

A frame-relay map statement "maps" an IP address to the remote port so you'd use the IP address of the remote DLCI in your map statements.

On the routing side the spoke routers will use a default route with a next hop of the hub router's serial interface. Then the hub router in turn will have a default route to the internet as well as static routes to each of the 8 spoke networks.

Here's a good link for you to read up on....pay attention to the hub/spoke section and sub-interfaces. I always use point-to-point subinterfaces instead of doing multi-point. It will make your life much easier, especially when you start dealing with routing protocols (which you don't need right now).

You might run into some congestion issues if you have 8 T1 circuits that all home back to a single T1 port. Depends on your traffic patterns really, but it is something to look out for. You'll be solely limited by that single hub T1 in terms of performance.

hope this helps. OH - and make sure you put modems on every router so you or whoever is supporting this can dial into the router when the circuit goes down.

 

[LastKnight]

Thanks for the response.

correct, that is the way i set up my frame relay map, I guess I wasn't clear.

My clients send large (40MB or so) files between locations, so my expectation was the point to multipoint solution would not use the hub and spoke method. Since I put the entire map in the single serial interface on the router, the network would act as a mesh network, rather than a hub and spoke config, but correct me if I'm wrong, and I need to get a bunch more dlcis per location to send point to point.

 

[spidey07]

Now I'm even more confused.

So every location needs to transfer large files to any other location? In that case you should create a full mesh network, with each site having 7 PVCs to every site. If that is the case then I'd also suggest running a routing protocol (EIGRP or OSPF - take that back a 1600 will have trouble with OSPF, stick to EIGRP).

This might get more complicated than you think if you want good performance and reachability between every site, but what really matters is what do you want the WAN to do? What kind of traffic and response time expectations are there? Are the sites close geographically or very sparse (coast to coast or even global)? Who is your provider and what kind of burst and excess burst guidelines do they have? The answers to those questions will guide you. Sizing the CIR correctly is critical and depends alot on your provider.

 

[LastKnight]

all good questions, sorry for the confusion, the network's more complicated than I should have tried to explain in a couple of sentences.

This might get more complicated than you think if you want good performance and reachability between every site, but what really matters is what do you want the WAN to do?
Purposes behind the WAN are many. They need communication back to their main site to talk to their scheduling system, but this is a base telnet session and shouldn't take up too much bandwidth. They will also be having a file/email server at the main location. Very little use is likely for saving files, email will be downloaded to the local client, so shouldn't take up considerable bandwidth, but will take up some. Only other true need is the ability to transfer these large files between locations at different times during the day, but not often. I'd say 10-20 times a day.

What kind of traffic and response time expectations are there?
currently they are running a 128K frame in a hub and spoke topology. They have no expectations, but I have told them the response will improve. Currently sending one of these large files compressed can take as long as 40 minutes, and slows down the network to a crawl, as you can imagine. Plus compressing the file has consequences that are undesirable.

Are the sites close geographically or very sparse (coast to coast or even global)?
All in one city, spread out pretty well over the city, but all local

Who is your provider and what kind of burst and excess burst guidelines do they have?
BellSouth is the provider. no burstability, but with the lines where they are now, sending the file in ten minutes would stun them. Don't know the CIR offhand, but could look it up on my SLA.


I recommended multiple dlcis per location, so that I could split the serial port into multiple interfaces, and give it a direct path with the interface (I was also more used to this configuration). I was told that I could do this with a single dlci per location using point to multipoint configuration. My frame relay map shows each router ip and dlci for routing. so when the data hits the single s1 interface, it looks at the map, sees the ip address destination, gets the dlci from the map, and sends to the router on the other end, which distributes through the ethernet port. At least that's the way I believe it will work, I have yet to see it in production, so it's quite possible I am wrong. Thanks a great deal for the responses.

 

[LastKnight]

I appreciate the help. Decided through your guidance I needed some additional help, so contacted a friend of mine that knows these routers backwards and forwards, going another way with the config and setting up subinterfaces. Thanks.

 

[spidey07]

cool.

Could you keep us posted on the final design so we all can learn? Being that it is local your latency is going to be very low (sub 20 ms) and you should get pretty decent performance. If large file transfers are causing grief then use some QoS features/Frame-relay traffic shaping to limit the bandwidth.

I'd probably do a full-mesh with sub interfaces and run EIGRP with the router closest to the firewall injecting a default route. You could do it with statics but 8 sites meshed would start to be a pain that way.

 

[FFC]

Just a quick note on the file transfers. IOS has a few QoS mechanisms available to allow you to prioritise the telnet traffic over the file transfers which as you say is not delay sensitive. Also look at using WRED for congestion avoideance and to maximise use of the links you have. Details can be found here WRED in IOS