Welcome to the Cisco Networking Professionals Ask the Expert conversation. This is an opportunity to discuss Fixed Configuration Stackable ArchitectureCisco Catalyst 3750 with Cisco expert Hari Krish. Hari is a Technical Marketing Engineer of Cisco's Desktop Switching Business Unit. Hari joined Cisco in 2000. During his tenure, he also has served as a Team Lead and Customer Support Engineer in the High Touch Technical Support (HTTS) Organization of Cisco's Technical Assistance Center (TAC). Feel free to post any questions relating to Fixed Configuration Stackable ArchitectureCisco Catalyst 3750. Remember to use the rating system to let Hari know if youve received an adequate response.
Hari might not be able to answer each question due to the volume expected during this event. Our moderators will post many of the unanswered questions in other discussion forums shortly after the event. This event lasts through June 20. Visit this forum often to view responses to your questions and the questions of other community members.
congratulations on the release of the Cat3750. Here are couple of questions that i have not been able to find in your collateral:
1. Performance. What is the performance of a standalone Cat3750 (different models). There is mention of the 32G max switching fabric and 38Mpps per stack, but how are these numbers derived ? Is this accummulative over 9 switches in a stack ? is it standalone ? I would like to know the blocking ratio per unit and per stack. Is IP forwarding at the same performance ?
2. Architecture. The feature set is similar to the Cat3550s. Are they based on an enhanced version of the satelite architecture of the 3550s ?
3. Stackwise. Is all forwarding using stackwise distributed ? Including L3 ? Or are the routing tables distributed among the units, yet L3 forwarding handled only via the master until there is a failure
thats all for now
Let me answer your questions, one at a time:
1. Performance: The forwarding rates for the individual boxes are as follows:
6.5 mpps (Cisco Catalyst 3750-24-TS),
13.1 mpps (Cisco Catalyst 3750-48TS),
35.7 mpps (Cisco Catalyst 3750G-24T),
38.7 mpps (Cisco Catalyst 3750G-24TS).
These are based on 64-byte packets, and all the boxes are wire-rate at these rates. These rates are for stand-alone units. Obviously, if you exceed the stacking throughput of 32Gbps, the switches will no longer forward at wire-rate across the stack.
This is a brand new architecture known as StackWise technology. It has new ASICs and a completely new design.
Please refer to the great multimedia training presentation that we have here:
We also have a great whitepaper that can introduce you to StackWise:
The Layer 2 forwarding is completely distributed.
For layer 3, the master switch in a stack builds the forwarding table, but then shares the FIB with all slave switches using distributed CEF (dCEF), and thus they all have this information and L3 forwarding becomes distrbuted.
I hope that helps
thanks for the information....unfortunately the presentation is very limited in the real technical information that i need and the demo looks slick but probably targetting end users rather than technical people like myself.
A couple of more questions for you though:
1. Cross stack Etherchannel: Are the switches intelligent enough to ensure that packets are hashed efficiently across the cross stack trunk and that minimal traffic goes via the Stackwise bus ? (ie hash packets on local trunks before sending via the bus)
2. Does the stackwise bus see every packet seen by the local front panel ports ? or does each switch build its own database and send packets to the bus if the destination is on another switch in the stack ?
Let me answer your questions-
1. We do offer a number of ways a user can configure to hash traffic across the Etherchannel. It can be dest-mac, src-dest-IP etc. Here are all the possibilities:
Now, traffic will be sent over the Etherchannel ports based on this hashing. However, that has nothing to do with traffic on the stackwise ring. The second answer should answer this question.
2. Every single packet sent on the front panel ports traverses the stack ring. This is to allow for the unified table build-up across the stack and to aid things like multicast. Each switch builds a database and this database is shared. So, all switches in the stack have the same database.
However, even though every frame traverses the stack ring, the stack bandwidth is 32Gbps and the latency is in nanoseconds. So, you dont have to worry about inefficient use of the stack ring.
I hope this helps.
The 3750 will wait for the IEEE 802.3af standard to get ratified before introducing any inline power switches.
My ?stion is abt the HSRP technic of using the flaoting address.let us take it as we have two switch acting in a redudant passion and cascaded each other.we are planning to use HSRP for redudancy.so each of the switch will be having a layer 3 info and a flaoting layer 3 address which will be used by the network a sdefault gateway.Now my ?stion is if we design a network in this style and we are using MAC address based dynamic VLAN which of the switchs MAC address will be available as for the default gateway interface or will it gets changed when ever it replaces the active or primary gateway for all machines in teh network provided there will be redundant ports reserved in each switch for each machine
The 3750 stack acts as a single router, and can act as one HSRP peer. The other peer can be another stack of 3750's or another router. However, you cannot do HSRP within the stack. For more information about stacking the 3750 and HSRP, please see here:
I do not fully understand your follow-up question, but I will answer what I think you are asking. The 3750 can act as a VMPS client (not server) and details about setting up VMPS are located here:
Does this answer your question?
we are planing to put four 3550 with stackable. And we need to config 4 Vlans in these 3550. Should we config the 4 GBIC GS ports to trunk mode with 802.1q ? Where we can learn these configruation?