Wired Intelligent Edge

Hello,

i am currently in the process of finding 2 new Switches for our Core Infrastructure. I am very interested in the 5400R zl2 Series and the VSF Feature. The two 5400R Devices would not be located in the same Room, so DAC is not an option. Instead we would choose a 40Gb Fiber connection with a supported QSFP+ Transceiver for the VSF Link.

The VSF Configuration Guide lists four supported port and media types

•  10Gig fiber
•  10Gig DAC
•  10Gig copper (SR)
•  40Gig DAC

It seems to me that 40 Gb VSF Link over Fiber ist not supported and so the above descripted scenario would not be possible.
Is this correct or is the guide maybe outdated?

Thanks in advance

Christian

 

Hello Christian, I'm quite sure that QSFP+ Transceivers can be used instead of DAC Cables to setup VSF Link(s) at 40Gbps (VSF Configuration Guide was published in 2016 IIRC...better to read the Aruba 3810/5400R Management and Configuration Guide for ArubaOS-Switch 16.06 (Edition 3 August 2018) <- VSF Virtual Switching Framework dedicated Chapter 24 at page 634 about VSF actual requirements and restrictions on Aruba 5400R platform).

Things to consider:

• to start with Aruba 5400R zl2 fully updated (actually KB.16.06.0006)
• to aggregate at least two (up to 8 IIRC) 10Gbps or 40Gbps interfaces on each logical VSF interface (so, as example, 2x40Gbps on each VSF member) and, that's important, try to differentiate modules on which those links start/arrive...so - on each Aruba 5400R zl2 - use two different modules with QSFP+ port(s) and use one port from each one...so if a module goes down...you will not lose the entire logical VSF link.
• to use OOBM-MAD or LLDP-MAD as MAD mechanism to help your VSF to sustain a Split Brain scenario. 

Hello parnassus,

thank you for the explanation.

I didn't consider the failure of a single module would crash the VSF Link and i can't mix 10 Gb and 40 Gb in a VSF Link. The 40 GB Modules are quite expensive for only two Ports, so i'm thinking of droping them and use a 4x 10Gb VSF Link distributed over 2 SFP+ Modules since we need them anyway.

 

 

 

Yes, a best practice would be to use two (or more) different modules providing some ports at 10 Gbps or some ports at 40 Gbps and then use one (or more) port(s) of each module to build - this on each VSF Member - the required logical VSF Link port(s) (example: VSF Link made of 2 or more aggregated 10 Gbps ports, or made of 2 or more 40 Gbps ports).

Yes 40 Gbps scenario is expensive...as example we deployed an Aruba 5406R zl2 VSF stack using 2 x 10 Gbps per VSF Member and it is working flawlessly but it all boils down to understand how much inter VSF Members traffic there would be (next step will be to add some DAC Cables on actual 20Gbps VSF Link to enhance resiliency agains cable(s) failure).

Parnassus,

We have two brand new buildings... togther the size is around 600,000 SQ Feet.  The building are split and one is 20% larger.  We want to put a 5400R (latest model) in each buidling.  We are going to run fiber strands between the two buildings.  We want to setup VSF but worried if there is a distance limit? The two 5400's will be around 1,000+ feet away from each other.  We are having our wiring company run 24 strands between the two 5400's.  The goal is to connect every IDF back to the two 5400's which will be our core.  Thoughts?

Each IDF is going to have two links - one for each 5400.  

Open to any suggestions, comments.  :-)

---

@BYGROB wrote: We want to setup VSF but worried if there is a distance limit?

---

A distance of 1000 feet between your two buildings equals to about 305 meters so - first of all - SX/SR SFP/SFP+ Transceivers could be used for your inter-building fiber optic links, respectively:

• SFP SX - Short Reach - Transceivers (up to 550 meters / 1804 feet using OM2 quality 50/125 um Multi-Mode fiber optic cable) like J4858C or J4858D.
• SFP+ SR - Short Reach - Transceivers (up to 400 m / 1312 feet using OM4 quality 50/125 um Multi-Mode fiber optic cable) like J9150A or J9150D.

avoiding much more expensive LX/LR - Long Reach - ones. That's for SFP/SFP+ Transceivers selection.

40Gbps links require, if deployed, QSFP+ Transceivers...and things are a little bit differents.

VSF deployment requires VFS links at 10G or 40G between two Aruba 5400R zl2 (I'm pretty much sure that you can't use 40G link splitted into 4x10G links in order to form the VSF Logical Port's members...in other words you need native 10G interfaces or native 40G interfaces not a breakout of 40G into four 10G links typically obtained using DACs on very short distances)...said so...if you are going to use two 40Gbps links to estabilish your VSF Logical Ports instead of using two or more 10Gbps links (I wrote two 40Gbps links to consider a minimum of resiliency and aggregated bandwidth even with 40Gbps) then you need to be very careful about QSFP+ Transceiver selection because you need to overcome the 300 meters threshold.

Probably (if MPO connectors are going to be preferred against more usual LC connectors) the HPE X142 40G QSFP+ MPO eSR4 300M Transceiver (JH233A) should be the right choice (Up to 400 meters / 1312 feet with OM4 quality 50/125 um Multi-Mode fiber optic cable)...using LC connectors on QSFP+ will force you to select LR4 (JH232A) QSFP+ Transceiver on Single-Mode fiber optic cable (Up to 10 km / 6.2 miles)...since the equivalent for Multi-Mode fiber optic cables (the JL308A) works with 50/125 um on OM4 quality only up to 150 meters / 492 feet.

The above is valid, considering your sites "cabling" distance, both for VSF on QSFP+ 40Gbps links and also for any other SFP 1Gbps / SFP+ 10Gbps links.

Then you need to think how to implement MAD (Multi-Active Detection): you're going to be forced in using LLDP-MAD because OoBM-MAD (possible between two Aruba 5400R zl2 in a VSF) will be limited - if we speak about Cat. 5E/6 - copper links...to 90 meters. LLDP-MAD (which uses a LACP to VSF from the MAD Device) will require instead an intermediate device (placed on Building A or on Building B...or in between Building A and Building B)...it's a Switch...and probably you will be forced to use at least another fiber optic link between Building A and B in order to provide one leg of the required LACP link (that's the best case if you plan to deploy the MAD Device on A or B very near the related Aruba 5400R zl2 instead of placing it in between).

Last...with remaining fiber optics links (24 strands mean 12 Tx/Rx pairs...isn't it?) you let devices - generally Switches or Servers connected via LACP against the VSF - on Building A already connected to VSF 1 Member to be also concurrently connected on VSF 2 Member on Building B...and vice-versa.

If you consider the case of an access Switch on Building A uplinked to VSF with 2x10Gbps (or 2x1Gbps) links...it will mean that one link of its aggregated port (LACP) will go to VSF 1 Member on the very same Building A and the other link will go to VSF 2 Member via inter-Building fiber optic links...this is approach is symmetrically valid applied on each side...the more access switches - or servers - you're going to connect resiliently to VSF the more fiber optic links you need between your two buildings...that's to say that 12 pairs could be easily not enough once you start considering VSF, MAD and access links that should use them.

Thank you for your response.  Very helpful and appreciate the detail.  What do you think?

https://glenair.itglue.com/2942071/docs/3274866

 

I think you asked about Aruba 5400R zl2 and VSF...then the scenario shows two pairs of Aruba 8325 (so VSX instead of VSF...).

Restrictions/Requirements about fiber links between buildings are somewhat still valid...but scenarios are a little bit differents.

VSX can be deployed only between two (and not four) Aruba 8320 (or 8325 or 8400X), not using two pairs...at least if the drawing is wrong...and VSX are within each building and between two VSX there is just a LACP aggregated port trunk (in any case wrong too since two links aren't enough to create a resilient VSX to VSX interconnection).

Sooooo...what's going on?