Networking

Current Focus, an update

So much happening since my last blog entry.  I’m currently working on two topics at the moment,

  • IPv6 transition architecture
  • SDN/NFV architecture

These are interesting topics respectively.

With IPv6 it seems the more one progresses with this protocol the clearer it becomes that the challenges which exist are not with IPv6 natively 1 so much as the concern with enabling IPv6 and IPv4 to co-exist, with an emphasis on intercommunications between them.  Seems a bit silly I guess, but from what I’ve observed and seen so far, this is the challenge (IPv4 <-> IPv6).  It’s not impossible, however it does require careful consideration; there are many approaches that attempt to solve for the co-existance between IP protocols, each with their own strengths and weaknesses.  So the good news is there are solutions ….and with these solutions the concern becomes choosing wisely.  Everyone will have a uniq perspective on this given their line of business and network needs.  One size does not fit all.

With SDN/NFV this is like a snow-ball rolling down a mountain, it’s gaining momentum and gaining in size.  Managed service providers and hosting providers have capitalized on how to operate with SDN/NFV.  Network service providers are looking to capitalize, with some Tier 1 services providers already implementing SDN/NFV.  I see programming as a corner stone and I see networking know-how has a corner stone.  Evolving with the industry is a requirement for success.

So, having written the above, I’m off now to delve into cloud image creation.

~ Jeff


  1. I will say the immediate exception I see here is the residential gateway (RG) devices.  Manufactures in the market space appear willing to implement IPv6 solutions into their products; however they want a to know how may new orders can be placed.  If it is not immediately obvious what the problem is with this statement (new orders) then please consider there are millions of RG devices already deployed.  Does the existing deployed hardware of these RG devices support IPv6?  If do support IPv6, which transition protocols are supported? 

Is it RIB or FIB capacity?

When reading or hearing about IP route capacity of routing gear it is often stated that a piece of equipment will support x-number of routes.  This number is a large number typically – and can sound impressive.

So if you hear or read of one capacity number mentioned keep in mind there is more than one capacity number to know.  RIB (or routing table) and FIB (or forwarding table) are two different tables within an IP networking platform.  They share common information, like routes, but perform two distinctly different purposes.  They also each have a different degree of resource capacity to perform their respective roles.

Technically RIB is an acronym for Routing Information Base and FIB is an acronym for Forwarding Information Base.  A RIB is the same thing as a routing-table.  A FIB is the same thing as a forwarding-table.

The RIB stores all routing input collected from routing protocols and holds that information for processing by the routing protocol(s).  Once a routing protocol has preformed it’s analysis on the RIB it determines what within the RIB should belong in the FIB.  If it sounds like I’m over simplifying the process, I am.  There are many different routing protocols (distance vector, link state).  Each of the routing protocols will employ an algorithm(s) to invoke functions and achieve its purpose.

The details of the routing protocols and methods they use are unimportant for making this simple point of awareness – there is more than one performance number to know when it comes to the number of routes a router can process.

Beyond the how many routes question it’s worth mentioning that performance is a broad discussion that includes several topics.  Vendors implement different hardware architectures to achieve several objectives like, performance, redundancy, capacity, services, green/environmental-initiative etc..  For example on higher-end gear it is standard to find a distributed hardware architecture whereby the FIB is located on the physical line cards.  So if you have a chassis with eight slots, there would be 8 FIBs, one FIB per slot.  Keep in mind I’m over simplifying it and skipping a lot of detail – some of which I probably can’t explain anyway.

Other performance metrics include knowing how well a vendor’s operating system and hardware ASICs perform at processing IP flows.  The overall point of awareness here is that simply reading a vendor’s datasheet should be a starting point for conversation(s) and not considered the end-all-be-all to make purchase recommendations and/or decisions.

If you’ve heard the terminology, “marketing math” then you know that datasheets can mislead.