cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1691
Views
0
Helpful
12
Replies

planning new network

flokki123
Level 3
Level 3

hi,

at  the moment iam planning a new network for a customer. the network is  split in three departments: developement(190 user), sales(100 user) and  data center(around 25 server).

now iam thinking about the design i should take.

i guess a collapsed distribution/core layer should be alright here.

i  thought i will configure 3 vlan´s(local vlans) for each department one  and build a stack within each vlan with the appropriate number of  switches for the number of users.

for  the core i would take two chassis switches like the 4500-R series. the  access switches get connected to the each core switch via a 1gbit fibre  cable, the switches which connect the servers get a 10gbit connection  for each core switch and the core switches itself i would connect via a  etherchannel consisting of 2-4 fibre gbit connections.

now i was wondering if this is a good approach to this scenario. never had to configure such a big network before.

what iam not sure about is how to configure the redundant connections.

because  if i had only one distri/core switch i would create 3 vlans on it,  create a trunk connection to each stack, create for each stack a vlan  and the distri/core would do the routing for the vlans.(think thats what  you would do with local vlans and routing)

but  if i have two distri/core switches i would connect each stack with both  distri/core switches. how would i then configure it? guess i would need  spanning tree, or could i accomplish it through routing?

would i  configure the connections from the stacks to the distri/core switches as  trunk and the connection between the two 4500 also as a trunk, then  configure each vlan on each distiri/core switch and also implement e.g.  hsrp? guess stp would be need then to block one of the connections to  the core.

would you do that?

any help and ideas would be really appreciated!

2 Accepted Solutions

Accepted Solutions

Leo Laohoo
Hall of Fame
Hall of Fame
for  the core i would take two chassis switches like the 4500-R series. the  access switches get connected to the each core switch via a 1gbit fibre  cable, the switches which connect the servers get a 10gbit connection  for each core switch and the core switches itself i would connect via a  etherchannel consisting of 2-4 fibre gbit connections.

Whao!  This doesn't sound right.

Let me get this straight.  The two core switch are connected to each other via an etherchannel of four 1 Gbps links.  The switches connected to the servers have an uplink to the pair of core switches using 10 Gbps?  Hmmm ... that defeats the purpose of the core switch.

By the way, 4500R is so "yesterday".  Look at the 4500R+E.  To get full potential of 10Gbps then you're looking at the Sup7E.

View solution in original post

"if i would use the L3 scenario, every connection between the access and the core would be a own subnet, right?"

Yes

Users

(3750 stack option)

Sales (100 users)- 3750 stack (3 or 4 switches) with one - layer3 uplink (/30) to each core- 1 data and 1 voice VLAN (/23)

Developement(190 user)- 3750 stack (5 switches) with one - layer3 uplink (/30) to each core- 1 data and 1 voice VLAN (/23)

If both sales and dev users are going to be serverd from the same IDF then you can have one big stack of 3750 switches with 2 VLANS one for sales and one for Dev users.

(3560X and 2960 Access Switches)

Try to get two 3560X-24T-E or 3560X-48T-E switches and use them as your User distribution layer.

In this scenario the two 3560X switches will be connected to the core switches with L3 links (/30) and act as a core for your users.

Run STP and HSRP on 3560X switches with etherchannel (4 links ) between them.

All the 2960 access switches will be connected to 3560X switches

(2960 Access Switches)--Layer2

Connect 2960 switches directly to your core with STP and HSRP

Data Center-

Option-1

Create an aggretion layer for datacenter with layer 3 uplinks to core.Create VLANS on the aggreagation layer switches.These will act as a core for the data center.

You can connect servers to the aggragation layer switches or you can add another switch which connects to both aggregation switches (STP will block on link) and add servers to that switch.

Option-2

Connect the server switches directly to core, one of the disadvantage in this design is STP will block one of your 10G uplinks.

"for the voip phones i would configure a own voice vlan. thus QoS would be implemented. anything else?"

Thts all you need config wise from the network side.

Siddhartha

View solution in original post

12 Replies 12

siddhartham
Level 4
Level 4

If you go with the collapsed core distribution model and you have 2 core swithes, limited budget then you have to configure the network as you stated above-

1. You would need spanning tree-one of the core switches as the root and other one as the backup root.

2. Trunk link between the user switches connected to both cores-one will be blocked by STP

3.Trunk link between the 2 core switches ( port-channel with 4 links) and also HSRP between them

The above design works good in a smaller company. Which switches are you considering for the users? if you have budget to buy 3750 switches with ip services image (gives you full rouing capability) then you can avoid spanning tree between the core and user distribution and go with routing between them, helps you for future expansion.

Siddhartha

hi sid,

thanks a lot for your answer.

i thought about taking 2960´s as access switches. i think 3750 would be to expensive here.

so i would have a own instance of STP and HSRP for each vlan right?  One link between the access and the two core swichtes would get blocked by STP.

How would you do it, if the access switches could do L3?

Leo Laohoo
Hall of Fame
Hall of Fame
for  the core i would take two chassis switches like the 4500-R series. the  access switches get connected to the each core switch via a 1gbit fibre  cable, the switches which connect the servers get a 10gbit connection  for each core switch and the core switches itself i would connect via a  etherchannel consisting of 2-4 fibre gbit connections.

Whao!  This doesn't sound right.

Let me get this straight.  The two core switch are connected to each other via an etherchannel of four 1 Gbps links.  The switches connected to the servers have an uplink to the pair of core switches using 10 Gbps?  Hmmm ... that defeats the purpose of the core switch.

By the way, 4500R is so "yesterday".  Look at the 4500R+E.  To get full potential of 10Gbps then you're looking at the Sup7E.

hi leo,

thanks for your help. you are actually totaly right. the etherchannel doesnt make sense here.

in the meantime i switched my plans and think about getting 2x 3750x-12s-e switches for the core, as the access switches will get connected via fibre only and through the optional modul i could create a 10gbit uplink between the switches.

hello everyone,

Can I ask a totally unrelated question please!!! I've been looking everywhere but can't seem to find the answer.

I am taking my ccna exams in a month's time and I am currently learning about frame-relay.

I have been looking everywhere but how do I know the DLCI that has been assigned to my local router? I know about using several show command e.g. show frame map, show fram lmi, show frame pvc , show run etc. but these only show the DLCI of the REMOTE routers. 

I guess my question stems from if there is already a preconfigured and fully operating frame-relay router being used and I as a new engineer wants to know the DLCI that is assigned to it, how do I know. I mean are there some other show commands that I have missed or is it a case of me contacting my service provider to know?

Once again guys am sorry for posting this in this particular forum but I am baffled!!

Thanks.

hi walker,

its no problem that you posted your question here, but its better to open an own thread, as here nobody will find your question and thus you wont get an answer.

as far as i remember from my ccna studies, try the command: show interfaces type slot#/port#, for the interface the frame relay is configured on.

maybe this shows the DCLI as well, but iam not sure.

in the meantime i switched my plans and think about getting 2x 3750x-12s-e switches for the core, as the access switches will get connected via fibre only and through the optional modul i could create a 10gbit uplink between the switches.

Will you be pushing MPLS (now and later)?

hi leo,

as far as i know, no mpls implementation is planned.

what is the reason for considering 3750X switches for core instead of 4500, if its budget I will agree but if its for 10G backbone I won't recommend them.Are you going to have any applications that would require 10G bandwidth?

As Leo suggested 4507R+E with sup 7-E would be a good recommendation for the core, if its too expensive then you can go with 4506E, the only difference I found between R+E and E is R+E supports redundant supervisor modules. Both R+E and E support 10G (http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps4324/product_data_sheet0900aecd801792b1.html).With this you can buy a 10G line if you need 10G capacity in future.

"so  i would have a own instance of STP and HSRP for each vlan right?  One  link between the access and the two core swichtes would get blocked by  STP."

Yes you are right

"How would you do it, if the access switches could do L3?"

I would suggest either stack switches or a layer 3 distribution block in each MDF for this design, you can configure a layer3 link between the stack and each core, benifits- no STP in this scenario between core and distribution, convergence will be very fast in case of failure scenario  and VLAN localization

Siddhartha

ît is acutally about the money. on the other hand, in the near future(next 2-3 years) i guess the company may expand the network structure with 2-3 new switches, if that. so i think the 3750x should be enough.

regarding the 10gbit uplinks, the customer is using a lot of VM´s and 20 server are quite a lot in my opinion, so i think 10gbit uplinks are a good idea.

if i would use the L3 scenario, every connection between the access and the core would be a own subnet, right?

as mentioned above, the customer will use VM´s and voip. do i have to be aware of something regarding voice and virtualization?

for the voip phones i would configure a own voice vlan. thus QoS would be implemented. anything else?

"if i would use the L3 scenario, every connection between the access and the core would be a own subnet, right?"

Yes

Users

(3750 stack option)

Sales (100 users)- 3750 stack (3 or 4 switches) with one - layer3 uplink (/30) to each core- 1 data and 1 voice VLAN (/23)

Developement(190 user)- 3750 stack (5 switches) with one - layer3 uplink (/30) to each core- 1 data and 1 voice VLAN (/23)

If both sales and dev users are going to be serverd from the same IDF then you can have one big stack of 3750 switches with 2 VLANS one for sales and one for Dev users.

(3560X and 2960 Access Switches)

Try to get two 3560X-24T-E or 3560X-48T-E switches and use them as your User distribution layer.

In this scenario the two 3560X switches will be connected to the core switches with L3 links (/30) and act as a core for your users.

Run STP and HSRP on 3560X switches with etherchannel (4 links ) between them.

All the 2960 access switches will be connected to 3560X switches

(2960 Access Switches)--Layer2

Connect 2960 switches directly to your core with STP and HSRP

Data Center-

Option-1

Create an aggretion layer for datacenter with layer 3 uplinks to core.Create VLANS on the aggreagation layer switches.These will act as a core for the data center.

You can connect servers to the aggragation layer switches or you can add another switch which connects to both aggregation switches (STP will block on link) and add servers to that switch.

Option-2

Connect the server switches directly to core, one of the disadvantage in this design is STP will block one of your 10G uplinks.

"for the voip phones i would configure a own voice vlan. thus QoS would be implemented. anything else?"

Thts all you need config wise from the network side.

Siddhartha

hi sid and leo,

sorry for the late reply.

thanks a lot for your effort, really appreciate your help.

@sid:

your last answer was really helpful, as i now know much more on how to configure different designs!

thanks!

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Innovations in Cisco Full Stack Observability - A new webinar from Cisco