8

The setting: Two locations I'm trying to connect. I have a /24 address block from ARIN Both location have the Same ISP.

The Question:

  • Can I announce the same /24 address block at the HQ location to Lumen?
  • Can I announce the same /24 address block at the School location to Lumen?

Since /24 is the smallest I can announce. At the end of the day, I'm also trying to connect these two locations. Thanks in advance for any advise.

enter image description here

Improve this question

3 Answers 3

Reset to default
7

You can announce it from both locations, but the result will be that traffic to the IP block will be delivered to whichever location provides the "best" route for the traffic, which may or may not be the site you want the traffic delivered to.

Assuming you want to split the IP block between the two locations then I see two basic approaches.

  1. Speak to your ISP. See if they are prepared to announce the /24 block on your behalf and route sub-blocks of it to your two sites.

  2. Establish a private link (this may be a VPN, but if it is a VPN then the VPN endpoints must not use IP addresses from the split block) between the two sites. Traffic arriving at the "wrong" site would be sent over the private link to the "right" site.

Improve this answer
1
  • The VPN solution would also cause a reduction in MTU. This might cause strange network problems later on, and I'd strongly suggest avoiding it if you can.
    – Manawyrm
    May 5, 2022 at 11:46
5

Yes, you can announce the same prefix in multiple location on the Internet. It is called anycast, and you see examples all the time, e.g. Google's 8.8.8.8 DNS server. Google has the network with that host address in many places around the world, and when you try to connect to that host, BGP routes your traffic to the nearest (from a BGP perspective) location.

The real problem you have is trying to connect between the two locations. You cannot connect from the /24 network at one site to the same /24 network at a different site. You could do something like having half the addresses at each location, but you will need to have something like a separate link (real or a tunnel) between the two sites, and one site receiving traffic for the other site needs to forward the traffic over the link to the other site.

You really have not provided enough information in order to give a more specific answer.

Improve this answer
4
  • For some more context. I'm trying to use my allocated IP block address since I'm at 0% IP utilization. I understand I could do an IPsec for the two locations, but if I wanted to use my own IP block, if possible. Hope this gives more information. I don't know what other info i could give please advise thanks
    – 13months
    May 5, 2022 at 3:37
  • ... or you could use another set of IP addresses to route between sites.
    – Zac67 ♦
    May 5, 2022 at 5:53
  • 1
    Accurate, I could, but I'm worried that if I do that, I'll be wasting a 254 IPs when at the school site; I'll only use like 2 or 3 ips
    – 13months
    May 5, 2022 at 7:05
  • I wonder what Google uses the rest of 8.8.8.x for
    – user253751
    May 5, 2022 at 10:28
-1

In general, NO, you cannot.

As others mention, you could announce same /24 from different locations that some things like anycasting, but that requires stateless protocols like UDP (so no using "fancy" TCP stuff like WWW, e-mail etc.). But it would not work for the purpose you seem to want.

Depending on your ISP goodwill (and your exact needs) however, you might be able to:

  • use IPv6 (which you can easily get in abundance) instead of scarce IPv4 (provided your use case is happy with that, of course).
  • subnet that /24 into two /25, and use different /25 on each location (and have each location only announce its /25). Sure, wider internet will would not accept smaller than /24 announcements, but BGP as a protocol allows it, and as both your connection travel via same ISP, the ISP could happily aggregate them back to /24 for announcing to the wider world. But it requires goodwill (and some technical competence) at the ISP.
Improve this answer
7
  • Anycast works just fine with TCP as well as with UDP. BGP directs the packets to the nearest (as BGP sees it) host with the anycast address. That is done at layer-3. TCP and UDP are layer-4 protocols that have no idea about anycast. In fact, many Internet sites use TCP with anycast.
    – Ron Maupin ♦
    May 5, 2022 at 22:19
  • @RonMaupin remember that BGP topology is not fixed and changes over time (how often depends on network stability, flapping actions, amount of multihoming, routing preferences and their stickiness/load balancing etc). When the topology changes, the middle packed of your TCP connection might end on different anycast node, which will have no idea of that that TCP stream (as it never did TCP 3way handshake), and will thus send a RST to terminate connection. So, no, it wouldn't work reliably.
    – Matija Nalis
    May 6, 2022 at 22:11
  • Do note that you could work around that issue by having a TCP-state sharing between different anycast nodes (e.g. via VRRP), but that in itself requires that you have separate non-overlapping IP address space for them to communicate, which brings up back to square one (as this is being unapplyable to case given in this question). Or you could accept that unreliability if you were desperate enough to accept such recurring problems, I guess.
    – Matija Nalis
    May 6, 2022 at 22:19
  • Anycast works with TCP on the public Internet at many sites all the time. Major web sites use it. The BGP routing is actually pretty stable, which is why it works. You are assuming a worst-case scenarios is common, but it is not. Remember that the metric BGP uses for the closest site is the number of ASes through which the traffic must pass, and that typically does not change despite various changes internally to an AS.
    – Ron Maupin ♦
    May 6, 2022 at 22:53
  • 1
    Anycast TCP works fine at major sites because they likely do TCP-state sharing via separate (non-anycast) IP addresses (which the OP seems to be unable to accomplish due to lack of address space). Also, while incoming BGP traffic is by default determined mostly by number of BGP hops (and communities/prepending etc which are usually quite stable too), outgoing traffic is often determined by other means (like RR / weighted balancing etc) so packets are routed asymetrically and can change on each trip (unless set to sticky, which depends on where - and how lucky - you are).
    – Matija Nalis
    May 6, 2022 at 23:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.