Net Work design

[lynneshri]

Dear All,

I have a question about selecting Cisco switches for my network. I recently joined this company. We are moving to a new location, and we have 5 floors. We have two IDF rooms and one MDF room. All cables are terminated in the IDF and MDF rooms. For each IDF room, I am going to install 4 switches (192 ports) in a stack. In the MDF room, I am also planning to install 4 switches in a stack. These switches are access switches. From these switches, trunk cables will connect to my core switch and the core switch will connect to the firewall and then ISP.

Each floor will have 100 to 150 clients (laptops, printers, and phones) connected at any given time, including Wi-Fi.

I am thinking that for access switches I should go for Cisco Catalyst C9200 with uplink port 40G. I am not sure about the core switch, but I think it should have for my core switch three 40G ports that will connect from my access switches.

Please I need help for selecting cisco access and core switch. Let me know if you need anything to more clarification. 

Thanks in advance.

LynneShri.

[Leo Laohoo]

---

@lynneshri wrote:
Please I need help for selecting cisco access and core switch.

---

What is the requirement of the access switch?  Is it purely Layer 2 (flat) network?  Is there going to be Dot1X involved? 

Without knowing the most basic requirement for the switch, there is no difference to a very expensive Catalyst 9200 against a Catalyst 1000, 1200 or 1300.

[Kasun Bandara]

Hi @lynneshri according to the your explanation, i suggest you to check cisco 1000 series also (cannot stack). you can use 10G uplinks with LACP (can bundle 2 links for 20G) which gives redundant and active links towards the core. also you need to do some bandwidth calculation to understand the usage of users. this depends on the applications/systems they are using. if you want to use stacking, you can go with 9200 series and check the estimated bandwidth before selecting 40G or 10G. you have many alternatives. for core switch, you can go to 9500 series. also you need to care about wifi plan. make sure to conduct a heatmap plan and design proper coverage and density to office area. 

from access switches to core switches can be trunk according to your plan. and internal routing can be done at core switch. (if you need to sperate internal networks, and do inspection, better terminate network gateways at firewall)

first plan the exact requirement, then calculate estimated usage.

[lynneshri]

At a given time 75 to 100 users are connected on each floor, that's why I am thinking uplink port 40 Gig. let me know if I am wrong?

In forum user suggested to check C9500 for Core switch, but I would like to see whether I can (3) connect 40 Gig trunk port? to core switch?

let me know what your opinion is.

[Joseph W. Doherty]

"I am thinking that for access switches I should go for Cisco Catalyst C9200 with uplink port 40G. I am not sure about the core switch, but I think it should have for my core switch three 40G ports that will connect from my access switches."

To recap, I, and apparently Leo too, believe using 40g links, for what you described is much more than needed.

That said, if you want to use 40g links, it's your budget fight - not ours.  ; )

If you want to use stackable switches (especially not exceed 4 per stack) for user access edge, that's likely fine too.

For distribution and/or core switches, and often for server edge switches, you need to be a bit more cautious in what you're selecting, as there's more to switch architecture than its port bandwidths.  Other hardware resources/features become more important and as might also software features.

I haven't used them, but on paper, all the Catalyst 9K switches appear to have more inherent capabilities and capacity than most of the prior gen Catalysts they replace.  The only possible downside, with all the new gee-wiz hardware and software, newly released Cisco devices seem to also have more "teething" pains than ever before.  Unsure what the situation is with 9Ks, across all them, families series, or particular models.  Possibly @Leo Laohoo had direct experience in this regard.

[Leo Laohoo]

40 Gbps uplinks is way too overkill. 

Catalyst 9200 is an overkill for a Layer-2 network without any DNAC or automation.  I'd go with Catalyst 1000 instead -- Cheaper and less prone to "undocumented features" brought to you by IOS-XE.

[Joseph W. Doherty]

Although Leo suggests the Catalyst 1000, it's not stackable, so if you want to stack, you'll need something like a 9200.

As the non-"L", i.e. modular uplink 9200s can have their uplink capacity changed, so you can decide to use a lower speed port bandwidth than one for 40g, and upgrade when, and if, needed.  However, again, really think it unlikely you'll need 40g, or even 25g.  Possibly to reduce costs, you might go with a 9200L supporting SFP+ ports, and "hedge" your uplink bandwidth needs using 10g.  Again, keep in mind, for user access, often an uplink to downlink ratio is fine up to about 50:1.  With 192 edge ports, and 2 (?) 10g uplinks, your ratio would be less than 10:1.

As to selecting a core switch, you really don't provide enough information beyond connecting your 3 access edge switches, and connecting to your FW.  You mention the possibility of using a C9300.  However, I would debate whether your MDF switch stack might not also be your site core too (heresy!).  If you use your MDF stack for both roles, than you might also consider using the non-"L" C9200s or any variant of the stackable C9300s.  Again, even a 9200L stack, for dual role (i.e. access/core) in the MDF would very likely be perfectly fine.

What you might be overlooking, higher bandwidth equipment is more about capacity, then raw "speed".  If I drive from point A to point B, an "18-wheeler" tractor/trailer doesn't deliver a single package any faster than a small compact delivery VAN.  Ditto uplinks with more bandwidth than the edge ports.  I.e. throughput for a user, with a gig port, is limited to gig, regardless of higher bandwidth transit paths.

BTW:

Heresy refers to any belief or theory that is strongly at variance with established beliefs or customs, particularly the accepted beliefs or religious law of a religious organization.

In regard to the classic three tier network model (i.e. core/distribution/access), many seem to overlook it came into being with "slow" routers and shared media hubs.  It was designed to scale performance.  It can still be used for that, but the capacity of hardware has increased exponentially.  For up to 576 user edge ports, you really don't need to scale device "performance".  (Heck, years ago, even had a 6513 with 1056 (eleven, classic bus, 96 port PoE [VoIP and data devices] line cards) 100 Mbps user edge ports.  Worked just fine - and it had to, as users were call agents [considered business critical].)

[Leo Laohoo]

Catalyst 1000 Single IP Management Configuration Guide, Cisco IOS Release 15.2(7)Ex (Catalyst 1000 Switches)

[Joseph W. Doherty]

The current variation of a switch cluster management IP, correct?

If so, redundant links now need something like STP vs. Etherchannel, correct?

Each physical switch cannot stack, if so, would need its own uplink(s), correct?

Further, if you cannot stack, you cannot take advantage of the stack's stack bandwidth capacity (depending on model; 80 or 160 Gbps for 9200s and 320, 480 Gbps or 1 Tbps for 9300s), correct?

BTW, the above is not meant to imply that using Catalyst 1000 isn't also a "good" approach too, just discreet switches vs. switch stacks have different pros and cons.

(NB: Another consideration, might be number of fiber runs between IDFs and MDF.)

[Leo Laohoo]

---

@Joseph W. Doherty wrote:
The current variation of a switch cluster management IP, correct?

---

What is old is new, the saying goes.  This is exactly "switch clustering" but without using the word "clustering".  This is just like calling a 6500E with a "Sup10T" supervisor card a 9600 with a Sup1 or calling a 9400 with Sup2 a "9400X".  Nothing but marketing gimmick.  

---

@Joseph W. Doherty wrote:
Further, if you cannot stack, you cannot take advantage of the stack's stack bandwidth capacity (depending on model; 80 or 160 Gbps for 9200s and 320, 480 Gbps or 1 Tbps for 9300s), correct?

---

Correct.  Most definitely not.  If anyone wants that rate of stacking bandwidth then that "somebody" can afford to pay for a Catalyst 9200/9300.  But the Catalyst 1000, 1200 and 1300 are targeted at a different market with "limited" funding budget.  

---

@Joseph W. Doherty wrote:
BTW, the above is not meant to imply that using Catalyst 1000 isn't also a "good" approach too, just discreet switches vs. switch stacks have different pros and cons.

---

I am putting options on the table.  Catalyst 1000 is, most likely to be, the last Cisco switch to run on classic IOS.  Because it is classic IOS (and not IOS-XE or IOX-Linux), the OS can stay up for years.  Another thing I want to point out to @lynneshri is the description of what is required is still very vague and none of it says "9200/9300" as a solution.