Table of Contents

BGP Training

22:46:48 Hours
22 Lectures


MPLS Zero to Hero Training

30:07:04 Hours
52 Lectures


Routing Protocols Design and Deployment Course

47:00:55 Hours
51 Lectures


BGP+SPF for Hyperscale/Massively Scale Datacenter Deployment

BGP+SPF Imagine we replace BGP best path selection decision with SPF. BGP+SPF exactly does that. In this post, I will explain why we are looking for alternative protocols for Massively Scale Datacenter. Although there is no exact answer how many devices should be in the datacenter so datacenter can be considered Massively Scale, but we know 10.000 Racks are not uncommon in these type of datacenter and each rack, when BGP is used as a transport, gets it's own unique AS number.

Before I explain BGP+SPF, let's understand why traditional , very well know, OSPF or IS-IS are not used in these type of datacenter. Answer is scalability. OSPF and IS-IS are chatty protocols and flooding aspect of these protocols are just not suited to very densely meshed connectivities. Yes, these datacenter run on CLOS topology and CLOS is densely meshed topology.

Also, we want to have wide ECMP in these type of datacenter, meaning, between the TOR, Leaf and Spine devices, there are so many equal cost path and we want to utilize them all. OSPF and IS-IS will be limited to number of ways of ECMP. BGP is very well suited protocol which provides very wide ECMP and it is not chatty protocol. BGP neighbors just exchange the changes. But, configuring 32, 64 or even 128 ways ECMP, even with automation is not trivial job, even for BGP.

Thus, we wanted to have, infrequent update(less chatty protocol), scalable protocol which can also support large ECMP (Equal Cost Multipath) connections. BGP + SPF comes to the rescue. BGP is not anymore path vector protocol when it is used as BGP+SPF! Mandatory BGP attribute such as AS-Path is still used but not not used for best path selection decision.

AS-Path attribute is even not used for loop prevention in BGP +SPF. Instead, for both best path selection decision, as well as, loop prevention, SPF (Our famously known Dijkstra Algorithm) is used. BGP + SPF , in fact uses BGP-LS Address family to distribute, link, prefix and node information. New SAFI , BGP -LS -SPF is introduced as a Sub Address Family Identifier.    

Figure - BGP Best Path Selection Decision is replaced with SPF in BGP+SPF Protocol 

BGP +SPF is discussed in the LSVR Working Group in IETF and Co-chair of this working group, Jeff Tantsura and I, discussed this protocol in the below video.  
Note: Today, Large scale datacenter use BGP (Classical BGP4) in their datacenter as it is explained in RFC 7938. Also, there is RIFT Working group, discussing RIFT protocol and applicability of this protocol in Large Scale, CLOS based datacenter.

Video : Orhan Ergun and Jeff Tantsura are talking about BGP+SPF and applicability in Large Scale CLOS based Datacenter Transport protocol 

Created by
Orhan Ergun

Orhan Ergun, CCIE/CCDE Trainer, Author of Many Networking Books, Network Design Advisor, and Cisco Champion 2019/2020/2021

He created OrhanErgun.Net 10 years ago and has been serving the IT industry with his renowned and awarded training.

Wrote many books, mostly on Network Design, joined many IETF RFCs, gave Public talks at many Forums, and mentored thousands of his students.  

Today, with his carefully selected instructors, OrhanErgun.Net is providing IT courses to tens of thousands of IT engineers. 

View profile