Brocade VCS architecture, or what i understood

I assist a presentation on the new Brocade VDX switches that implement their VCS technology.VCS enable a multipath Ethernet fabric using the emerging ietf standard Trill (https://datatracker.ietf.org/wg/trill/charter/) therefore replacing the venerable Spanning-tree protocol. So during the presentation, they built the following architecture and explain how the traffic would switch. All links between the switches are in the etherchannel trunk (Whatever it is called today)

In a nut shell, the VCS architecture will collapse the traditional distribution and Access layers into one, called edge layers.

So base on the layer 2 routing algorithm the traffic from server A to server B will use the links between Switch A and C, the cost is determine by the speed of the physical links, and Link aggregation do not change the overall cost the link. IE (3 links 10G in a LAG will not change the cost of the link)

Source Destination Path BW Cost
Switch A Switch C SwA-SwC 30 Gbps 500
Switch A Switch C SwA-SwB-SwC 30 Gbps 1000

According to Brocade, traffic from server 1 to server 2 will be equally load-balanced on all 3 links between Switch A and Switch C. So far this is looking good.

Now what if one link go down,

After recalculating the layer 2 routing algorithm the traffic from server A to server B will still use the links between Switch A and C. The total amount of bandwidth available is not taken into account in the path determination

Source Destination Path BW Cost
Switch A Switch C SwA-SwC 20 Gbps 500
Switch A Switch C SwA-SwB-SwC 30 Gbps 1000

According to Brocade, traffic from server 1 to server 2 will be equally load-balanced on the 2 remaining links between Switch A and Switch C.

May be it’s me in my perfect world scenario but I would have expect the traffic to switch to the fastest aggregate BW witch is via SWB.

When we ask about the fact that more BW was available on the other path, the answer was that the layer 2 algorithms is based on a Cost and therefore the total aggregated bandwidth is not considered, but we can adjust the cost on each link to manually do traffic engineering.

I do not know nor I saw how the other vendor did implement their layer 2 routing algorithm, nor if my expected behaviour would made sense and scale.

About Mario

Consultant en télécommunication et CCIE #7704 depuis plus de 15 ans et totalisant plus de 30 ans d'expérience en TI. Ce blog sera un reflet de mes opinions sur les nouvelles technologies et l'impact de ces dernières sur nos vies Network consultant and CCIE #7704 for more than 15 year with over 30 years in IT, This blog will be my take on technology and the impact on our life suivez moi / Follow me on Twitter , Google Plus , Linkedin
This entry was posted in networking. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s