[NNagain] The Whys of the Wichita IXP Project
Dave Cohen
craetdave at gmail.com
Thu Feb 22 20:47:34 EST 2024
It sounds then like the sample of organizations I’ve worked with may not be truly representative of the full sample. Nonetheless, it does appear to me that the expansion rate is accelerating, and that datacenter operators are making efforts to attract IXPs into their facilities (with a couple of notable exceptions where the operators themselves are attempting to be their own IX) in a way that didn’t occur years ago.
Dave Cohen
craetdave at gmail.com
> On Feb 22, 2024, at 7:51 PM, Bill Woodcock <woody at pch.net> wrote:
>
>
>>
>> On Feb 23, 2024, at 01:09, Dave Cohen <craetdave at gmail.com> wrote:
>> Prior to roughly 2016-17, most IXs in the US that we worked with were in a single site
>
> The vast majority of IXPs are still in a single site, but all the ones with normal switch-fabric extension policies are on a growth path to multiple sites, provided the markets they’re in are of sufficient size to support it. Both of those things have been the case for more than thirty years.
>
>> …that site was often not a third-party facility, often a private R&E oriented datacenter.
>
> Can you explain in more words what you’re saying here? IXPs spring up in all kinds of sites… Many of them are in closets of buildings which happen to have multiple ISPs already present because all of the ISPs already have customers in the building. Some of them are in datacenters, and some of those are “neutral” datacenters, while others are owned or operated by one of the participating networks, and others are governmental or academic. There’s a very wide diversity of IXP siting arrangements, and often individual IXPs are spread across several different kinds of sites; that’s always been the case.
>
> Over time, existing IXPs, as they grow, tend to spread across more sites; but also, new IXPs form, which almost always form in a single location before growing to multiple sites. IXPs that _start_ organically in multiple sites simultaneously certainly aren’t common.
>
> -Bill
>
More information about the Nnagain
mailing list