Comware

Dear colleagues, 
I have a question about IRF Stacking and bandwidth. 
For example, I have the topology below.
How will in that case IRF stack link work?
Will the switches try to route the traffic between the servers basically inside the switches and use IRF for some synchronization? Or IRF link will transfer all the servers traffic randomly?
I'm afraid to use LACP because of a possible bottleneck in the IRF link of 20G.
Of course, 20G is a great speed, and not all hard drives and storages can provide it, but nevertheless, it is interesting to understand the work of the IRF links better.
I hope I explained my question well and you will point me in the right direction...

Hi, I believe that (I mean, I recall I've read about how traffic is managed by IRF members) all depends what IRF member the incoming traffic from dual homed Server is hitting and where (and how) the destination is connected to.

Example: source SRV1 sends a file to the destination Store, LACP of source SRV1 decides (due to hashing mechanism) which LAG link will be used for SRV1 egressing traffic  so it could be that SW1 or SW2 will be hit by this outgoing traffic (incoming on SW1 or SW2 via their 10G legs). Since each HPE 1950 is dual connected to downstream peers (SRV1 and Store) then the Switch receiving that incoming traffic will be the one preferred to be used to send traffic to destination if that destination is directly connected (this is due to IRF internal logic)...so I believe that traffic traversing (and impacting) the IRF Link will be marginal during normal operations. Totally different will be the case when Source and Destination have a single leg to their upstream switch and those legs are connected to different Switches (say SRV1 only connected to SW1 and Store only connected to SW2).

------------------------------
Davide Poletto
------------------------------

Original Message:
Sent: Dec 15, 2021 04:17 PM
From: Mikhail Solovey
Subject: HPE OfficeConnect 1950 12XGT IRF stack Bandwith

Dear colleagues, 
I have a question about IRF Stacking and bandwidth. 
For example, I have the topology below.
How will in that case IFR stack link work?
Will the switches try to route the traffic between the servers basically inside the switches and use IRF for some synchronization? Or IRF link will transfer all the servers traffic randomly?
I'm afraid to use LACP because of a possible bottleneck in the IRF link of 20G.
Of course, 20G is a great speed, and not all hard drives and storages can provide it, but nevertheless, it is interesting to understand the work of the IRF links better.
I hope I explained my question well and you will point me in the right direction...

Thank you very much for such a detailed answer, Davide! 

@parnassus is absolutely right. Each IRF member device has complete layer 2 and layer 3 forwarding capabilities. Below are the packet forwarding scenarios explaining the same.​​​​​


------------------------------
Brian Dsouza
------------------------------

Original Message:
Sent: Dec 15, 2021 04:17 PM
From: Mikhail Solovey
Subject: HPE OfficeConnect 1950 12XGT IRF stack Bandwith

Dear colleagues, 
I have a question about IRF Stacking and bandwidth. 
For example, I have the topology below.
How will in that case IRF stack link work?
Will the switches try to route the traffic between the servers basically inside the switches and use IRF for some synchronization? Or IRF link will transfer all the servers traffic randomly?
I'm afraid to use LACP because of a possible bottleneck in the IRF link of 20G.
Of course, 20G is a great speed, and not all hard drives and storages can provide it, but nevertheless, it is interesting to understand the work of the IRF links better.
I hope I explained my question well and you will point me in the right direction...

The HPE OfficeConnect 1950 Switch Series includes five switches: New to the series is a 16-port
10-Gigabit aggregation switch that has 12 10GBASE-T and 4 SFP+ ports, which is ideal as the
core of a high performance workgroup or small business network. Additional series models have
Gigabit access ports with 10-Gigabit uplinks, including two standard and two PoE+ models in
24- and 48-port configurations. The access switches each have two 10GBASE-T ports supporting
copper- based Category 6A-based cabling, and two 10G SFP+ ports for fiber connectivity.
The PoE+ models both ha

------------------------------
Terry Hayward
------------------------------

Original Message:
Sent: Dec 15, 2021 04:17 PM
From: Mikhail Solovey
Subject: HPE OfficeConnect 1950 12XGT IRF stack Bandwith

Dear colleagues, 
I have a question about IRF Stacking and bandwidth. 
For example, I have the topology below.
How will in that case IRF stack link work?
Will the switches try to route the traffic between the servers basically inside the switches and use IRF for some synchronization? Or IRF link will transfer all the servers traffic randomly?
I'm afraid to use LACP because of a possible bottleneck in the IRF link of 20G.
Of course, 20G is a great speed, and not all hard drives and storages can provide it, but nevertheless, it is interesting to understand the work of the IRF links better.
I hope I explained my question well and you will point me in the right direction...