nanog at ics-il.net
Mon Oct 2 18:49:11 CST 2017
*nods* I'm obviously a bit biased, but I have some reasoning behind it. Also, this is excluding large scale networks like Comcast, AT&T, Charter, etc. that do have the scale to have on-net solutions. It's also excluding remote locations where there may not be another node for hundreds or thousands of miles (Alaska, Hawaii, Caribbean, etc.)
There are some CDNs that are pushing out on-net boxes and PNIs to networks with relatively low usage. We've had networks that were in the process of joining the IX get approached by one of the CDNs for a box right on their network. That network had under 1G of usage (IIRC) and was joining an IX where the CDN was already present. The network declined the box because they didn't want something else to manage and would already get the service anyway once on the IX.
Also, from my understanding of multiple CDNs (nothing really proprietary, just whatever's been made public) is that different nodes have different content. The more traffic a node gets, the more likely they are to have the content the end user seeks. Some CDNs also have different platforms on different boxes. If an ISP has one box, they may not even have access to all of the platforms that would be available in a several box deployment. In these cases, an IX makes more sense.
We've also had CDNs go the PNI route to a network. Sure, a PNI beats an IX in that it cuts out a middle man (fiscally and point of failure). However, if the networks wouldn't use a substantial portion of the IX port in the first place, it's an extra cost that small networks may not have room for or have to choose between a PNI and an IX.
Per the last message, Cloudflare seems to have a similar philosophy. Join existing infrastructure where it makes sense, deploy additional nodes otherwise.
Intelligent Computing Solutions
Midwest Internet Exchange
The Brothers WISP
----- Original Message -----
From: "Tom Paseka via NANOG" <nanog at nanog.org>
To: "Marco Slater" <marco at marcoslater.com>
Cc: "NANOG list" <Nanog at nanog.org>
Sent: Monday, October 2, 2017 1:21:10 PM
Subject: Re: isp/cdn caching
Cloudflare does deploy caches, however we usually look to do so in unique
locations, ie. where an ISPs network isn't already in reach of one of our
existing deployments/peering points.
You can email peering at cloudflare.com directly if seeking this.
On Fri, Sep 29, 2017 at 7:22 AM, Marco Slater <marco at marcoslater.com> wrote:
> Do they publicly have any more info on this?
> I thought CloudFlare didn’t consider doing that because of their vast
> coverage and peering arrangements provided by their PoPs.
> Marco Slater
> > On 29 Sep 2017, at 14:38, <michalis.bersimis at hq.cyta.gr> <
> michalis.bersimis at hq.cyta.gr> wrote:
> > I think that Cloudflare has a caching solution, but I think they have
> strict requirements towards the isp in order to install them on their
> > Best Regards,
> > Michalis Bersimis
> > -----Original Message-----
> > From: NANOG [mailto:nanog-bounces at nanog.org] On Behalf Of Aaron Gould
> > Sent: Thursday, September 28, 2017 6:25 PM
> > To: Nanog at nanog.org
> > Subject: isp/cdn caching
> > Hi, I've been aware of a few caching providers for a few years now, but
> I'm learning of others as time goes on. which makes me curious if there are
> more springing up and gaining popularity. I'm speaking of ISP-type caching
> whereas the cache provider sends hardware servers and perhaps a switch to
> the local ISP to install locally in their network. Can someone please send
> a simple list of what they know is the current players in this ISP Caching
> space? I'll list the ones I know about and you please let me know of
> others. This seems to be an evolving/growing thing and I'm curious of
> where we are today for significant providers and possibly up-and-coming
> ones that I should know about. (amazon prime has my wondering also.)
> > Google (GGC)
> > Netflix (OCA)
> > Akamai (AANP)
> > Facebook (FNA)
> > Apple (I heard this isn't isp-located like the others, but unsure)
> > ? others ?
> > ? others ?
> > ? others ?
> > -Aaron Gould
More information about the NANOG