Service Providers

The Global Internet IPv4 forwarding table has now breached the key 512k threshold. Why should you care?

by pmoyer on ‎09-30-2014 11:15 AM (5,497 Views)

The Internet IPv4 forwarding table continues to grow unabated. Is anyone surprised? Earlier thinking by industry pundits believed that as networks migrated to the new promised land of IPv6, the need for additional IPv4 address space would diminish; leading to a flattening of growth for IPv4 addresses. On top of that, it was expected that as IANA becomes more stringent on IPv4 address space allocation (due to the imminent inevitability of running out of IPv4 address space), this must surely result in a dimished appetite for IPv4 addresses – and both of these would lead to a flattening of the growth trend for the global IPv4 Internet table.


However, the expected flat lining of the IPv4 growth trend line is not occurring. The below diagram clearly depicts this.





A double-click view at the IPv4 table size from the CIDR report; which is the industry de-facto resource on Internet routing tables, is below.





So, why is the IPv4 forwarding table continuing to grow at such a rapid rate and what’s so special about the 512k threshold? I’ll answer both of those questions in this blog.


First, there are several reasons the IPv4 table continues to grow even while the IANA address allocations have pretty much maxed out. As seen below, the IPv4 address allocations have flat lined.




Source: []


While the address allocations have flat lined, other factors such as prefix length distribution and unique prefix advertisements are driving the growth trend. In other words; de-aggregation of the IPv4 address space is driving this. In a nutshell, the IPv4 address space in the Internet routing table is becoming more fragmented. This is due to the lack of available large address block allocations; which leads to more unique and specific prefix lengths being advertised. This directly contributes to the continuing growth of the IPv4 table.


Now, back to the question of what’s so special about the 512k threshold? The simple answer is there are tons of routers deployed today that can only store 512k routes in their forwarding planes (ie., TCAM, FIB, etc). Some of these routers have been deployed many, many years ago. Some of these routers have been deployed more recently; however, they have been deployed or re-deployed into BGP peering roles without the necessary foresight regarding the growth of the Internet routing table. As the IPv4 table breaches this 512k threshold, these legacy routers can no longer add any additional prefixes and the result is a black-hole routing effect. In fact, some vendor gear handles this in a more disastrous manner – by crashing. Of course, I’m not talking about Brocade gear here! There have already been reports of such crashes occurring in the Internet.


Now the good news!


Brocade has been shipping line cards that support 1M IPv4 entries for many years now. Here is a data sheet of the industry’s first 2-port 100GbE module on the MLXe Internet scale routing platform that has been shipping for over 3 years. If the router is not intended to be involved with full BGP peering, hence there is no need for a module that can carry 1M IPv4 route entries, there are modules available that store 512k route entries. Customers must be sure that these modules will not be deployed in a scenario now or in the future where full BGP routing is required. A good example of this is in the data center as a core router. The standard data center architecture does not put the burden of full BGP routing (ie., Internet peering) on the data center core routers. The common practice is to deploy an Internet border router for this peering role, such as a Brocade MLXe with modules that support the 1M route entry capacity or a Brocade CER platform (that happens to support 1.5M IPv4 entries). The morale of that story is that the intended application of a router must be well planned and defined before any deployments. An operator should not assume that any router can be deployed into an Internet BGP peering role.


Now the better news!


Brocade is now shipping advanced Internet scale modules for the MLXe platform. We have doubled the IPv4 route table size to 2M IPv4 entries and practically quadrupled the IPv6 capacity to 800k entries. These modules are the 20x10GbE and 2x100GbE line cards and both are half-slot, line-rate modules with full feature support; including data plane programmability which is so important with the emerging SDN architecture.



Brocade MLXe 20x10GbE Module with SFP+ Optics: Supports 2M IPv4 Entries!


The new 20x10GbE module will also support 128-bit MACSEC encryption, in addition to the feature support of IPv4, IPv6, MPLS and OpenFlow/SDN.



Brocade MLXe 2x100GbE Module with CFP2 Optics: Supports 2M IPv4 Entries!


With the capacity for 2M IPv4 route entries these modules will scale well into the future and continue to provide the investment protection you’ve come to expect from Brocade. The 800k IPv6 route table capacity also provides unparalleled investment protection, as the IPv6 route table also continues to grow. But more on IPv6 in a future blog …


I hope this clears up the question with the IPv4 route table growth trend and provides you with a clear path towards which products to deploy in critical BGP peering roles. As usual, please provide any comments or questions to this blog!


by Martin Mune
on ‎10-03-2014 03:47 PM

Thanks for the great info on IPv4 route table growth.  I was wondering about your views on future trends with the IPv4 space.

by pmoyer
‎10-07-2014 08:47 AM - edited ‎10-07-2014 09:17 AM

Thanks Martin for the comment and question. The future trend in IPv4 address space is an interesting question and one that a customer asked me about just last week. The customer's concern was that if the IPv4 table grows exponentially in the next few years, will a 2M IPv4 table become a reality sooner rather than later? My short answer is that we won't see a 2M IPv4 table for many, many years.


The longer answer is that this estimate on future growth depends upon a few factors; some which are mentioned in my blog. One factor is the growth of de-aggregated IPv4 space and the other factor is the lack of available IPv4 address space from IANA and the Regional Internet Registries. If de-aggregation continues at its current trend, this shouldn't be a cause of great alarm. The current trend line in my blog includes de-aggregation and the trend line is not going up exponentially; it is basically on the same trajectory. This implies that de-aggregation won't "explode" the IPv4 table size. Then when you add in the lack of available and unique IPv4 address space, then you get a clearer picture of what the future trend may hold for IPv4 address space.


Interesting enough, Akamai recently released its Q2/2014 report on the "State of the Internet" and in it they show that the global number of unique IPv4 addresses declined on a quarterly basis for the first time. While they didn't go into the gory details that led to this decline, the lack of available IPv4 space is clearly one factor. The growth of IPv6 addresss space is becoming more apparent as well, so it appears we are finally at the point where networks are embracing IPv6.


So, I hope this provides some clarity on where we may be going in terms of future IPv4 address space growth. Thanks for the question!