Originally Posted by Anthony A.
thanks for the response. to be more specific, i have about 20 drops of cat6 that will be used solely for accessing the internet. these will go into the 24 port switch which, through some reasearch, looking into the netgear managed switches with jumbo packets, etc. (ie. GSM7224). the other 4 ports on the switch will be used for home automation panel and touchscreen panels throughout the house for music control. this is all located in the electrical room. i then have a media closet which houses all my audio components for whole-home control and home theater. since it is in another room (and i don't want to run 8 cat6 cables to the electrical room), i was hoping i could get away with a 8 or 16 port managed switch, again with jumbo packets, etc. in the media closet will be whole-house music controller, htpc's, network media streamers with HD/bluray content streaming to different tv's in the house. i am concerned that the HD stuff will stutter with this setup.
so im really wondering what i can do to get away with using 2 switches as outlined above. perhaps getting better switches? where will the bottle neck be.... at the 24 port switch or the 8 port in the media closet.... or will all ports be bogged down?
thanks for the help.
So here's the problem with going with jumbo frames. Every device connected into a jumbo frame infrastructure has to support jumbo frames. If one device does not support jumbo frames, this device will not be able to talk over the network. So what can you do in a mixed mode environment? You can isolate components which can support jumbo frames by creating a specific VLAN for them. Other devices which don't support jumbo frames can then be put on a separate VLAN. Now the problem comes where you need the devices in the jumbo frame VLAN to talk to the devices in the non-jumbo frame VLAN. In this case, you'll need a router. Layer 3 switches can bridge this deficiency as they support routing. Otherwise, you'll have to add a separate router which will then route traffic between the VLANs. You will also have to run two different subnets to get routing to work. One subnet per VLAN. The assumption with doing VLAN segregation is if the managed switch in question supports isolation of jumbo frame enabled ports via VLANs. The last gotcha is with performance. When you have traffic crossing through areas of the network where there is a difference in MTU sizes, the device which has to either assemble frames or fragment them takes a performance hit. How huge depends on the amount of traffic we are talking about. In this case, the device would either be your switch (if you decide to go with a layer 3 switch which the GSM7224 is not) or the router you add to route traffic between the two VLANs.
As you can see, the use of jumbo frames is not trivial. It requires proper planning and many times just isn't worth the headache.
An alternative which you've alluded to is to run multiple uplink connections between switches. Since you are planning on using managed switches at both ends, you should have no problems getting this to work. You don't need to run 8 links. You can run 2, 3, or 4 links. It's really up to you. I personally haven't gone past 2 links between switches in my home environment. Once you have the physical drops pulled, you need to configure both switches to utilize the links as one virtual uplink. Doing this is called link aggregation. You will some times run into manufacturers who call this trunking. I prefer not to call it trunking as trunking in the Cisco world is used to describe a link which carries more than one VLAN traffic through it. Link aggregation supports up to 8 physical uplinks grouped into one virtual link. The IEEE standard for this is 802.3ad. At the basic level, a standard 802.3ad aggregated link operates in a static mode. The logic in choosing which link a frame is pushed down is very basic and takes on a round robin type behavior. The protocol enhancement to this is LACP (link aggregation control protocol.) LACP allows switches to pass loading information between each other to better determine which physical link in the link aggregation group to use. Cisco has their own proprietary protocol which they are phasing out called PAgP.
Notice I've been careful about not stating a link aggregated group operates at the sum of all the links participating in the aggregation. What this means is if I use 8 GigE links, the link aggregation does not operate at 8 Gig. All traffic going through a link aggregation is still limited to 1 Gig. But because there are multiple pipes, the chances of multiple sessions saturating a single GigE link is decreased. There are other reasons why link aggregation is used such as link redundancy. I have done tests where I've pulled links out of a aggregation group without dropping a single frame...same applies when I re-establish the link.
The above is probably more than you or anyone would care to know. But since we are discussing more advanced layer 2 switching, I felt it appropriate to talk about this to ensure that people understand there are some complexities and concepts that need to be understood.
So how does this fit in with your scenario? You have to determine how much traffic you think will traverse between switches. As I've said in other posts, it is always best to keep high bandwidth heavy traffic generating devices on the same physical switch. Having these type of devices on the same physical switches allows maximum performance as the communication never leaves the switch but stays within the backplane which always has a at a minimum the performance/throughput of all the physical puts added together. Many times the backplane will have more by a certain margin. Solid, well performing networks are always designed properly before any hardware gets thrown into the mix.
PS....Mr. Bobb is going to love this post......