Provider aggregated vs Provider Independent IP Addresses - Use case

[nwekechampion]

Hi Guys,

I am looking for some info regarding the usage PI and PA addresses.

What are the use-cases?

When would it be advisable to use one?

How does it affect BGP?

Do ISP clients still need to be on a different AS if they have a PA address?

Appreciate some feedback.. thanks guys

[Flavio Miranda]

Hi

  This table must be usefull.

What are the use-cases?

Servce Provider x Customer

When would it be advisable to use one?

Are you a Service Provider? PA

Are you a customer? PI

How does it affect BGP?

 If PI, can be advertised by customer.

Do ISP clients still need to be on a different AS if they have a PA address?

  They theorically would not have a PA unless the are a Service Privider

PI vs PA : Comparison Table

Below table will help give insight how both differ :

PARAMETER	PA (PROVIDER AGGREGATABLE)	PI (PROVIDER INDEPENDENT)
Ownership	Service Provider	Customer
IP pool announcement	Only can be advertised by Service provider owning the Address Pool	Can be advertised by customer through any service provider through BGP
BGP Based routing	BGP may not be required for providing redundant connectivity	BGP is one of majorly used routing option for providing redundancy
Address Block Size	Can be longer IP block of /25 or /26	At least /24 IP Block size
IP address Assignment	PA addresses are assigned from an LIR’s allocation and are registered in the RIPE Database by the LIR	PI address space is assigned separately and not from an LIR’s PA allocation. All PI assignments are registered in the RIPE Database by the RIPE NCC at the time they are assigned
Management	Easier to implement and control. NAT configuration can do the trick.	Complex BGP based routing required for traffic control and providing redundancy
Route aggregation and CIDR at provider end	Possible and this provides flexibility to service provider.	Not possible
Portability	Needs to be changed when moving from one service provider to another	Don’t need to be changed when moving from one service provider to another
Commercials	Cost cheaper than PI address block.	Costlier than PA address block.

[nwekechampion]

Hi @Flavio Miranda ,

Thanks for your reply, but not sure< follow.

When I get a static IP from an SP is that a PI or PA? Because, when I move or change ISPS my public IP changes, so it is really independent of ISP. Can you please clarify?

[Flavio Miranda]

 It must be PI as you are an ISP´s customer.   Dont matter which ISP you are using you will be classified as Customer. Unless of course you become one ISP.

[Jatinjaipur]

When you get a static IP address from a Service Provider (SP), whether it is a Provider Independent (PI) or Provider Assigned (PA) address depends on the allocation and ownership of that IP address block.

PI addresses are allocated directly to an organization by a Regional Internet Registry (RIR), and the organization can use those addresses independently of their Internet Service Provider (ISP). PA addresses, on the other hand, are allocated by the ISP who owns the address space, and the organization is not able to take those addresses with them if they change to a different ISP.

So, if your static IP address remains the same when you switch ISPs, it means that your IP address is a PI address because it is allocated directly to your organization, and your organization uses it independently of the ISP. If your IP address changes every time you switch ISPs, it means that your IP address is a PA address, and it is assigned to you by the individual ISP.

It's worth noting that the ownership of IP addresses is a complex topic that is geographically dependent. This means that the answer to your question may be slightly different depending on your location and the specific allocation policies of your RIR or ISP

[Jatinjaipur]

In the context of BGP (Border Gateway Protocol), PI and PA addresses refer to the types of IP address assignments that are used in the global routing system.

PI (Provider-Independent) addresses are assigned directly to organizations by a regional Internet registry (RIR) such as ARIN, RIPE NCC, or APNIC. PI addresses are portable, meaning that the organization can take them with them if they switch providers or move to a different region.

PA (Provider-Assigned) addresses are assigned by an Internet service provider (ISP) to their customers. PA addresses are not portable and are associated with the provider who assigned them.

And we know that BGP (Border Gateway Protocol) is a routing protocol used to exchange routing information between routers in different autonomous systems (AS). When an organization connects to the Internet, it is assigned an IP address block, and this IP address block is announced to the Internet using BGP.

PI and PA addresses can affect BGP in a few ways:

1. BGP Prefix Advertisement: The PI and PA addresses are announced in the BGP updates as prefixes to the global routing table. The PI addresses are more flexible since they can be announced to multiple providers with BGP while PA addresses are tied to a specific ISP.

2. Routing Policy: The PI and PA address blocks can have different routing policies based on their nature. For example, since PI addresses can be carried over to different ISPs, the routing policies can be more relaxed. Meanwhile, the routing policies associated with PA addresses can be more stringent and provide better control over the announcement of these address blocks.

3. BGP Peering and Routing Agreements: Depending on the use of PI or PA addresses, organizations might need to enter different BGP peering agreements with their ISPs. For instance, an organization with PI addresses might establish peering with multiple ISPs, while an organization with PA addresses might have to rely on a single ISP.

Ultimately, the differences between PI and PA addresses can have important implications for BGP routing, particularly when it comes to maintaining routing control and portability of IP addresses.

[nwekechampion]

Thanks @Jatinjaipur ,

My point then is when would a client choose to use PA or PI? and how does this affect the usage of BGP?

If I used a PA (Dependent on ISP), would I still need BGP or would default routes suffice?

Thanks

[Jatinjaipur]

If you use a PA (Dependent on ISP), the default routes could suffice as
long as your organization is connected to a single ISP and not multi-homed.
However, if your organization is multi-homed and needs to use BGP to ensure
that the traffic is routed through the correct provider, then using PI
address would be recommended. In general, using BGP for multi-homed
connections is recommended because it allows an organization to control the
traffic flow better and ensure redundancy in case one ISP goes down.

[nwekechampion]

@Jatinjaipur .. so  basically PI --> BGP and PA --> No BGP ?

[MHM Cisco World]

Follow

[Joseph W. Doherty]

The big difference is with your own PIs, you also own your own AS.  Makes you independent of ISPs, for addressing, and you can (relatively) easily use multiple ISPs and/or public Internet exchanges to connect to the Internet.

Depending on how you connect with the Internet, you might use BGP or you might not.  Whether to use BGP is not 100% strictly tied to your IPs, but is much more likely if you're using PIs as you will likely advertise your AS.

[nwekechampion]

Yep.. this makes a lot more sense to me now.. I was along the same line of thought that PI's are associated to use of AS and for ebgp neighborship with any ISP.

Thanks for clarifying..

So @Joseph W. Doherty .. basically PI --> BGP and PA --> No BGP ?

[Joseph W. Doherty]

Keep in mind using BGP between you and a provider can be used for Internet routing and/or routing between you and your ISP.  So, with either a PA or a PI you might use BGP or you might not, although it's much, much more likely you'll use BGP with a PI.

[nwekechampion]

Thanks @Joseph W. Doherty ,

But why would one use BGP with a PA if they do not have an ASN to do ebgp neighborship, as I am sure ISP will not allow client to be in its ASN (ibgp)? Doesn't quite stack up. Can you explain please?

[Joseph W. Doherty]

Most common reason would be to accept (just) a default route, dynamically.

Another possible reason would be drop outbound traffic to unknown Internet destinations (assuming you have complete Internet route table).

For either you could use private ASN.