We are comparing switch stacking (available in 3750) and switch clustering (available in 3560). We have the following questions:
1. Can you consider switch clustering a way of switch stacking? In other words, is switch clustering a kind of stacking?
2. Comparing stacking and clustering, what are their differences?
As far as your first question is concerned the answer is NO. Clustering and stacking are two totally different concepts.
Stacking connects two or more (up to 9) switches through a stack cable and uses Cisco's proprietary StackWise protocol, the connected switches though physically separate form one logical switch. You will manage a stack of switches as one and having one management IP address. Moreover stacking connects the backplane of the switches thus having full backplane speed connectivity between the switches.
Whereas clustering is almost piling up one switch on top of one another and connecting them in a chained fashion using any of the Ethernet ports (trunks or access). You do not see those piled up switches as one, they have to be managed separately. Clustering is also known as cascading.
Hope this helpsâ¦rate if found helpful
Rohit Rattan Pandita
Some minor clarifications:
"Moreover stacking connects the backplane of the switches thus having full backplane speed connectivity between the switches."
Not always true. For instance, for the 3750-E, if you look at Table 2 in http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps7077/product_data_sheet0900aecd805bbe67.html, you'll see the "128-Gbps switching fabric", yet "StackWise Plus for ease of use and resiliency with 64 Gbps of throughput" (which is a bit misleading, since it's dual 16 gig duplex links).
"Clustering is also known as cascading."
At least with Cisco switches that support "clustering", a physical daisy-chain or cascade is not always required to define a cluster. For instance, even for the older 2900 XL and 3900 XL series "The switch uses Cisco Discovery Protocol (CDP) to discover and display candidate switches that can be added to a cluster. By using CDP, a switch can automatically discover switches in star or cascaded topologies that are up to three cluster-enabled devices away from the edge of the cluster. You can configure the command switch to discover switches up to seven cluster-enabled devices away. The default is three hops."
Also from Cisco:
Understanding Switch Clusters
A switch cluster is a group of connected Catalyst desktop switches that are managed as a single entity. The switches can be in the same location, or they can be distributed across a contiguous Layer 2 network. All communication with cluster switches is through one IP address.
In a switch cluster, 1 switch must be designated as the command switch and up to 15 switches can be member switches. The command switch is the single point of access used to configure, manage, and monitor the member switches. It identifies and controls all member switches in a cluster, regardless of where they are located and how they are connected. You can designate one or more switches as standby command switches to avoid losing contact with cluster members if the command switch fails.
Although my query has no relation with the original post.
Since you have raised this point, I would like to get some clarity.
Table 2 in the datasheet for 3750E switch states - 128-Gbps switching fabric
Table 8 "ordering info" in the same datasheet - WS-C3750E-24TD-S "68-Gbps, wire rate backplane"
My understanding is that Backplane and swithc fabric means the same. However with 68Gbps cisco is mentioning "Wire Rate" Backplane.
I am confised by these figures. Can you throw some ligh on this.
Good catch! Where I think the "68-Gbps, wire rate backplane" spec means, since the 48 port models also have two 10 gig ports, you have, discounting the stack ports, 68 Gbps of local ports, and the switch has sufficient internal backplane bandwidth it can run all at wire rate.
If you look at the datasheet for the 4948 10 gig, http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps6021/ps6230/product_data_sheet0900aecd80246552.html, it notes "136-Gbps switching fabric", which is exactly 2x 68.
The confusion often is how bandwidth is stated, simplex or duplex. Have you ever noticed how 100 Mbps Ethernet full duplex becomes 200 Mbps Ethernet trunks? (Marketing likes larger numbers.)
Consider a two port gig switch. In we send in full gig on each port such that it comes out the other, there's only 2 gig across the backplane, but 4 gig (in and out) on the ports.
Joseph, I agree with what you are saying. But still can't establish any correlation between 68Gbps Wirerate backplane, and 128Gbps switching bandwidth.
For a truly definite answer, we would need Cisco to comment since it's their datasheet. If could be something as simple as a cut and paste from a 3560-E since they don't have the stack ports.
Thanks Joseph. I am not really that much bothered to dig this info. from Cisco.
Was just curious about the facts.
Appreciate your time Joseph.
As mentioned in the above post, clustering and Stacking is not the same.
To start with Stacking is not supported on all the switches. Its only supported on 3750 and 3750E switches, where you have dedicated stacking interfaces on the switches and you connect these switches using the stacking interfaces. Usually these interfaces are of higher capacity. For example 3750 supports 32Gbps stacking bandwidth while 3750E supports 64Gbps. Thus you have a higher bandwidth between the switches if you are stacking the switches.
If you have any other switch you need to go for clustering. Again, while clustering the swithces the connectivity would depend on the interface available on the switch. THe only advantage of clustering over connecting the switches without clustering is management. When you cluster the switch you can manage multiple switch through single IP address. However connectivity wise its as good as connecting these switches without stacking.
When you stack 3750 switch, they start acting as a switch chassis with line cards. So virtually it becomes one switch with an interconnectivity of 32/64Gbps. You configure the switch as you would be configuing a chassis based switch. You don't have to login in evry switch to configure it. The stack can be configured from any of these switches and the configuration is replicated on all the switches so that if the stack master goes down, you still have other switches in the stack operating intact.