3750 switch for intervlan communication?

[cpals]

I was just thinking tonight while studying for my CCENT exam and the guy was talking about the ip routing function and realized that I wasn't even sure our 3750s had that feature turned on.

My question is if, the 3750 does not have that feature turned on and on computer on port 3, but in a different vlan tries to talk to a different computer on port 5 then the packet will go out that port, out of the switch, into our 6509 (routing) and then back to the switch and to the computer on the different vlan?

Or will the switch automatically match the mac-address from the computers and forward the frame regardless if ip routing is turned on or not?

My setup at work really wouldn't have too many places where this could actually happen - it was more of a self-learning question.

Thanks.

 

[ScottMac]

It'll go out the trunk to the 6500 and back.

A VLAN is a logically separate entity. essentially a totally separate switch within a switch. It takes a Layer three device to pass traffic from one network or subnetwork to another.

If the trunk didn't exist, you'd need to dedicate a port from every VLAN port group to go to a layer 3 device ... i.e., 4 VLANs on the (L2) switch, then there'd be four connections to a L3 device.

 

[cpals]

So if we're not even turning on the 'ip routing' function of the 3750s in our IDFs, was it a waste of money to purchase them? Would a 2960 worked just as well if they're just being used at the Access level? The only thing special that I can see that we do is stacking the switches, which I'm not sure the 2960s can do.

I know it's a broad question, but it's more of an answer just for me and not my work.

Also, I guess another question in a different form would be is there ever a valid reason to purchase a layer 2/3 switch (like us), but only use it at the layer 2 level? I suppose the 3750s probably have more processing power and backplane speed?

 

[Pheran]

ScottMac is correct, the 3750 is not going to do any routing unless you have set up the appropriate SVI interfaces to be gateways for those networks. For the record, all 3750 switches support layer 3 routing, though only the ones with the IP services feature set will support the "good" dynamic routing protocols (OSPF/EIGRP).

 

[Pheran]

Originally posted by: cpals
So if we're not even turning on the 'ip routing' function of the 3750s in our IDFs, was it a waste of money to purchase them? Would a 2960 worked just as well if they're just being used at the Access level? The only thing special that I can see that we do is stacking the switches, which I'm not sure the 2960s can do.

I know it's a broad question, but it's more of an answer just for me and not my work.

Also, I guess another question in a different form would be is there ever a valid reason to purchase a layer 2/3 switch (like us), but only use it at the layer 2 level? I suppose the 3750s probably have more processing power and backplane speed?

It depends on what you need. If you need stacking (StackWise), you need the 3750 as the 2960/3560 do not support it. If you only have 1 switch operating a layer 2, then the 3750 is probably a waste of money unless you have lots of other ones and are trying to keep a standard switch, or you need some other esoteric layer 2 feature like full private VLAN support. I'm pretty sure that the 2960 backplane is the same speed as the equivalent 3750, since the 2960/3560/3750 all use the same core technology.

 

[cpals]

Gotcha... I guess technically we wouldn't need stacking, but it sure is nice to only have to have one management IP per IDF. We currently have four IDFs in our building with three 3750s stacked in each IDF. They're just doing layer two switching with one vlan, possibly a couple more depending on if a host needs a specific IP for some reason.

Thanks!

 

[Pheran]

Originally posted by: cpals
Gotcha... I guess technically we wouldn't need stacking, but it sure is nice to only have to have one management IP per IDF. We currently have four IDFs in our building with three 3750s stacked in each IDF. They're just doing layer two switching with one vlan, possibly a couple more depending on if a host needs a specific IP for some reason.

Thanks!

Yup, this is the same reason I use 3750s in our access layer. They aren't doing routing, but StackWise makes management extremely convenient. Never underestimate the cost savings of ongoing labor as opposed to capital expense.

 

[Jamsan]

Agreed on the ease of management. We just installed 3750s across all of our IDFs, and life is much easier now. The other thing 3750s can do over 2960s is port density with PoE support. The max a 2960 comes with on PoE is 24 ports. There is no 2960 that can do PoE and Gig on all parts as well, so one of those 2 factors may be a reason why 3750s were chosen for the IDFs.

Also, the 3750-E's are the ones that have the advanced routing services (OSPF/EIGRP, etc.). The regular ones can only do static and RIP if I'm not mistaken. We actually have a pair of 3750-E's in our core doing all of our inter-vlan routing, and no issues with throughput so far.

 

[Pheran]

Originally posted by: Jamsan
Agreed on the ease of management. We just installed 3750s across all of our IDFs, and life is much easier now. The other thing 3750s can do over 2960s is port density with PoE support. The max a 2960 comes with on PoE is 24 ports. There is no 2960 that can do PoE and Gig on all parts as well, so one of those 2 factors may be a reason why 3750s were chosen for the IDFs.

Also, the 3750-E's are the ones that have the advanced routing services (OSPF/EIGRP, etc.). The regular ones can only do static and RIP if I'm not mistaken. We actually have a pair of 3750-E's in our core doing all of our inter-vlan routing, and no issues with throughput so far.

You are slightly confused about the 3750-E. The 3750-E is new hardware that has support for 10-gigabit uplinks and an improved backplane and stack protocol (StackWise+). You can still get advanced routing protocols on the regular 3750 with the IP services code; I believe that the 3750-E code includes this by default.

 

[spidey07]

Originally posted by: Pheran

Originally posted by: Jamsan
Agreed on the ease of management. We just installed 3750s across all of our IDFs, and life is much easier now. The other thing 3750s can do over 2960s is port density with PoE support. The max a 2960 comes with on PoE is 24 ports. There is no 2960 that can do PoE and Gig on all parts as well, so one of those 2 factors may be a reason why 3750s were chosen for the IDFs.

Also, the 3750-E's are the ones that have the advanced routing services (OSPF/EIGRP, etc.). The regular ones can only do static and RIP if I'm not mistaken. We actually have a pair of 3750-E's in our core doing all of our inter-vlan routing, and no issues with throughput so far.

You are slightly confused about the 3750-E. The 3750-E is new hardware that has support for 10-gigabit uplinks and an improved backplane and stack protocol (StackWise+). You can still get advanced routing protocols on the regular 3750 with the IP services code; I believe that the 3750-E code includes this by default.

Yep.

BUT!!!!

There are IOS features different from the 3750 and 3750-e. Specifically a regular 3750 can't be a eigrp stub router, 3750-e can. Learned that the hard way. 🙁 There are very minor differences in terms of features so check the differences before making a decision. Probably higher TCAM resources on the 3750-e as well.

 

[spidey07]

Originally posted by: cpals
Gotcha... I guess technically we wouldn't need stacking, but it sure is nice to only have to have one management IP per IDF. We currently have four IDFs in our building with three 3750s stacked in each IDF. They're just doing layer two switching with one vlan, possibly a couple more depending on if a host needs a specific IP for some reason.

Thanks!

The biggest advantage of 3750 is the stacking. It allows you to do cross chassis link aggregation/etherchannel and elminate spanning-tree convergence all together resulting in sub second convergence times if a switch or link fails. It really all depends on the design but this feature alone is why a 3750 is so powerful.

 

[Jamsan]

Originally posted by: Pheran

Originally posted by: Jamsan
Agreed on the ease of management. We just installed 3750s across all of our IDFs, and life is much easier now. The other thing 3750s can do over 2960s is port density with PoE support. The max a 2960 comes with on PoE is 24 ports. There is no 2960 that can do PoE and Gig on all parts as well, so one of those 2 factors may be a reason why 3750s were chosen for the IDFs.

Also, the 3750-E's are the ones that have the advanced routing services (OSPF/EIGRP, etc.). The regular ones can only do static and RIP if I'm not mistaken. We actually have a pair of 3750-E's in our core doing all of our inter-vlan routing, and no issues with throughput so far.

You are slightly confused about the 3750-E. The 3750-E is new hardware that has support for 10-gigabit uplinks and an improved backplane and stack protocol (StackWise+). You can still get advanced routing protocols on the regular 3750 with the IP services code; I believe that the 3750-E code includes this by default.

You're absolutely right. I was thinking of the part number difference between the Standard and Enhanced versions (SMI/EMI) on the normal 10/100/1000 models. The standard is WS-C3750G-24TS-S whereas the Enhanced is the WS-C3750G-24TS-E (for the particular 24 port model switch we ordered).

The 10 gig models all start with part number 3750E.

 

[Pheran]

Originally posted by: spidey07

Originally posted by: Pheran

Originally posted by: Jamsan
Agreed on the ease of management. We just installed 3750s across all of our IDFs, and life is much easier now. The other thing 3750s can do over 2960s is port density with PoE support. The max a 2960 comes with on PoE is 24 ports. There is no 2960 that can do PoE and Gig on all parts as well, so one of those 2 factors may be a reason why 3750s were chosen for the IDFs.

Also, the 3750-E's are the ones that have the advanced routing services (OSPF/EIGRP, etc.). The regular ones can only do static and RIP if I'm not mistaken. We actually have a pair of 3750-E's in our core doing all of our inter-vlan routing, and no issues with throughput so far.

You are slightly confused about the 3750-E. The 3750-E is new hardware that has support for 10-gigabit uplinks and an improved backplane and stack protocol (StackWise+). You can still get advanced routing protocols on the regular 3750 with the IP services code; I believe that the 3750-E code includes this by default.

Yep.

BUT!!!!

There are IOS features different from the 3750 and 3750-e. Specifically a regular 3750 can't be a eigrp stub router, 3750-e can. Learned that the hard way. 🙁 There are very minor differences in terms of features so check the differences before making a decision. Probably higher TCAM resources on the 3750-e as well.

You sure about this? I haven't used the feature, but the 3750 is supposed to support EIGRP stub routing as of 12.2(25)SEE, even in the base image.

 

[mcmilljb]

Originally posted by: Pheran

Originally posted by: spidey07

Originally posted by: Pheran

Originally posted by: Jamsan
Agreed on the ease of management. We just installed 3750s across all of our IDFs, and life is much easier now. The other thing 3750s can do over 2960s is port density with PoE support. The max a 2960 comes with on PoE is 24 ports. There is no 2960 that can do PoE and Gig on all parts as well, so one of those 2 factors may be a reason why 3750s were chosen for the IDFs.

Also, the 3750-E's are the ones that have the advanced routing services (OSPF/EIGRP, etc.). The regular ones can only do static and RIP if I'm not mistaken. We actually have a pair of 3750-E's in our core doing all of our inter-vlan routing, and no issues with throughput so far.

You are slightly confused about the 3750-E. The 3750-E is new hardware that has support for 10-gigabit uplinks and an improved backplane and stack protocol (StackWise+). You can still get advanced routing protocols on the regular 3750 with the IP services code; I believe that the 3750-E code includes this by default.

Yep.

BUT!!!!

There are IOS features different from the 3750 and 3750-e. Specifically a regular 3750 can't be a eigrp stub router, 3750-e can. Learned that the hard way. 🙁 There are very minor differences in terms of features so check the differences before making a decision. Probably higher TCAM resources on the 3750-e as well.

You sure about this? I haven't used the feature, but the 3750 is supposed to support EIGRP stub routing as of 12.2(25)SEE, even in the base image.

And you would be right. Just checked the feature navigator.

 

[rathsach]

I've always failed to see the big advantage from 3750 gigabit switch stacking.. The cost is SO close to a 4506 chassis solution, that I don't think it makes sense going for a stack.. The stacking on the 3750 100mbit switches are a good cheap high port density access layer solution..
Gigabit ones.. Don't see it.. If you don't need the port density, then a 4948 is a WAY better switch, if you need high port density I'd pick a 4506 any day..

 

[drebo]

Well, as Spidey has iterated a few times in the past, which is easy to agree, with stacked switches, you can do cross-chassis EtherChannel and link aggregation. This reduces your failover time in the event of a failed switch to near instant, rather than the minute-plus you'd have with STP.

If you don't care about the time it takes for STP to choose its new path, then it doesn't really matter.

Basically, I don't think that he's advocating using the 3750 for user access (depending on numbers, you're right, a chassis solution is far better for user access). For server access, though, if you have two links to your distribution and two links to each server, one from each switch, switch or NIC failure recovery is nearly instant.

No more spanning tree!

 

[Pheran]

Originally posted by: rathsach
I've always failed to see the big advantage from 3750 gigabit switch stacking.. The cost is SO close to a 4506 chassis solution, that I don't think it makes sense going for a stack.. The stacking on the 3750 100mbit switches are a good cheap high port density access layer solution..
Gigabit ones.. Don't see it.. If you don't need the port density, then a 4948 is a WAY better switch, if you need high port density I'd pick a 4506 any day..

I use 100M 3750s for the access layer. I agree that the 48-port gigabit ones are extremely pricey, I've not been doing high-density gigabit stacks. The 4500 series was practically useless for a long time due to the crappy 6G slot bandwidth, but the newer E-series improved that to 24G so they are not too bad. The 4948s are nice, I even use them for small core applications. Of course they are just 4500 Sup V modules in a 1U box.