bgp-and-beyond.com

Netflix – Comcast deal: why I believe that Netflix has not sold its soul

nospam@example.com (Fredy Künzler) — Mon, 24 Feb 2014 11:52:19 +0100

by Fredy Kuenzler

Most news sources report that Netflix is paying to access the Comcast user base for its streaming content. I believe this is wrong, because it would remove the strongest argument for Netflix when negotiating with other large eyeball networks such as AT&T or Liberty Global (UPC). „If you pay once, you will pay forever“. This would basically mean that Netflix would have sold its soul. As the deal between Netflix and Comcast is under NDA, every report is pure speculation, and if WSJ or other general media claim „Netflix is paying Comcast“ it doesn't mean that „Netflix signed for Paid Peering“.

I'm involved in the peering community for several years and I'm attending industry events such as Global Peering Forum or European Peering Forum on a regular basis, representing my company Init7. I know the tricks and deals of the industry, and also the legal issues. It's common knowledge that large eyeball provider and incumbents such as Deutsche Telecom, Liberty Global (UPC) or Verizon give content operators such as Netflix, Akamai or Google a hard time. Comcast is not an exception. Init7 is not a content network, rather an independent backbone operator, but nevertheless on the content side. That's why Init7 filed at the Swiss regulator against Swisscom (German text), the Swiss incumbent.

Here are the three points what I believe the deal between Netflix and Comcast include:

1. Netflix gets „free“ (zero-settlement) peering capacity for local routes of Comcast. That means Netflix is only allowed to use metro routes, i.E. Netflix can only deliver for free to Comcasts Philadelphia customer base from servers located in Philadelphia, via the local peering. The same applies for all other US areas, where Comcast is offering services. This to avoid the costly long-haul transport network of Comcast. Likely Netflix agreed to pay all third party cost, such as cross connect fees in carrier neutral datacenters. Possibly Netflix also pays CAPEX cost of Comcast (i.E. additional router linecards or optics). But in general, Netflix achieves „Zero Settlement Peering“ for local Comcast routes. If this wouldn't be the case, I would not consider the deal „mutually benefical“.

2. Netflix also purchases a couple of hundred gigabit of transit or partial transit capacity from Comcast. That's clearly a paid agreement, but not a paid peering agreement, because it offers more than just the (local or national) customer base of Comcast. Netflix can either deliver to a remote location customer base of Comcast, but they can also deliver to 3rd party networks such as Verizon, using the interconnection capacity between Comcast and Verizon. This resolves some capacity issues Netflix faces towards the Verizon customer base, and Netflix won't be forced to sign for paid peering from Verizon (see above - if you pay once, you will pay forever).

These two points are commonly called „Dual Peering / Transit“ strategy – see Dr. Peering whitepaper for reference.

3. Netflix promises not to oppose against the merger of Comcast and Time Warner Cable. Once this merger is finalized, the Comcast-Netflix deal is valid and extended also between TWC and Netflix, too. I think this last point is a very strong argument for Comcast to give in and finally agree to the deal.

Customer Experience at Telehouse Chelsea, 85 10th Avenue, New York

nospam@example.com (Fredy Künzler) — Tue, 04 Feb 2014 18:13:00 +0100

by Fredy Kuenzler

In January 2014 Init7 built a 2nd pop in New York City, to achieve redundancy for it's transatlantic transport termination. We've chosen Telehouse Chelsea due to customer demand. Here is our experience with this collocation supplier (I have to note that Init7 is also Telehouse customer in London and Paris).

While paperwork was a bit difficult to handle for various reasons (some were our fault), the installation was surprisingly easy going. We shipped a large router (~150kg) beforehand. The arrival of the gear was confirmed by Telehouse.

The entrance of the building is rather nice: a lot of artwork, the landlord seems to have a sense for nice decorations. Telehouse itself is a tenant on the 7th floor; I was told that the datacenter was formerly operated by Lehman Brothers but upgraded by Telehouse.

Upon arrival, the shipped gear was delivered to our cabinet within ten minutes without even asking for. The cabinet and cross-connects were delivered and ready as ordered. The router itself (Brocade XMR-16000) is really heavy but a Telehouse technician helped to lift it into the cabinet.

Soon I figured that our power cables did not fit, we got standard US power cables, but the cabinet was equipped with APC power strips. Telehouse staff quickly tried to resolve the issue: they provided other APC power strips, which, unfortunately did not fit with the ordered power supply plugs above the cabinet (2* 208V, 20amps); but after all they could borrow the right power cables for our equipment and within a few hours everything was up and running. We will order and ship new cables to the site in order to return them.

Our new NYIIX connection turned up immediately – and on top, we got a temporary VLAN via NYIIX to interconnect with our NYC-1 pop in TELX 111 8th avenue too bring our gear online, as the ordered dark-fiber from 3rd party vendor was expected to be delivered late.

The new dark-fiber itself had an issue: one fiber strand had too much loss. Telehouse staff was really helpful to get everything measured and checked. I also could borrow a short SC-SC fiber cable within minutes to plug a loop for debugging. At the time of writing it's not yet resolved as the issue is most likely with the 3rd party supplier but I'm positive this will be fixed soon.

All in all I have a very good impression of the service the Telehouse team provided to us, it's smooth and easy going, really helpful and friendly, and the team seems to have a relaxed and welcoming atmosphere. For customers from abroad I think this is a value-add which has no price tag. Engineers from Europe are usually not too familiar with the habits of North America. From my experience I can recommend to collocate at Telehouse Chelsea (85 10 Avenue NYC), there are plenty of cabinets available, too. And there are helpful technicians in short distance available, if needed (unlike other sites in NYC). Also to mention that Telehouse offers WiFi to customers in the whole site, ask at reception for UserID and Password.

Init7 NYC-2 pop at Telehouse Chelsea is now operational. We are ready to take orders for IP transit and MPLS VLL services – please send inquiry to sales at init7 dot net. And yes: we can deploy fast (as Telehouse can do cross-connects quickly).

A visionary has gone: Carsten Schloter

nospam@example.com (Fredy Künzler) — Tue, 23 Jul 2013 14:18:00 +0200

by Fredy Kuenzler

The message is shocking: Carsten Schloter, CEO of Swisscom died at the young age of 49. Our condolences to his family.

I'm known to share the same opinion with Swisscom rarely, nevertheless I think that Carsten Schloter was a visionary who spoke clear words to his European counterparts, the incumbents. His speech at the ETNO conference in October 2012 has not been forgotten.

Cisco Flexlink and Brocade

nospam@example.com (Fredy Künzler) — Sat, 29 Jun 2013 22:41:00 +0200

by Fredy Künzler

Nice article by Tobias Brunner about Cisco Flexlink configuration and how to interconnect with Brocade gear.

STM-1 link with MiRICi-155 Gigabit Ethernet over STM-1 SFP Converter

nospam@example.com (Lino Moragon) — Wed, 14 Dec 2011 15:55:53 +0100

by Lino Moragon

Some time ago we were given a bundle of two new MiRICi-155, Gigabit Ethernet over STM-1/OC-3 SFP Converter as test equipment.
The goal was to simulate a SDH link with STM-1 framing with a maximum bandwidth of 155 Mbit/s.

In order to get the MiRICi SFPs to operate functional, it is essential to disable Autonegotiation on the SFP Interface, as well as setting Speed to 1000 and Duplex to Full. For Cisco devices it is also necessary to input the following commands, because otherwise the devices would reject the SFPs and error disabling the port:

Switch(config)#service unsupported-transceiver
Switch(config)#no errdisable detect cause gbic-invalid
Switch(config-int)#speed 1000
Switch(config-int)#no negotiation auto

After these steps, the interface will still be shown as unsupported, but it won't go to “err-disabled” state when plugging in.
The interface will also show connected state, even if there isn't any optical cable plugged in.
For the SMT-1 configuration, you won't have any specific configuration options on the switch, such as encapsulation, framing type, etc..
On the other hand you can configure a lot of options on the MiRICi itself with a special USB SFP configuration adapter from the manufacturer RAD (www.rad.com). The settings will be saved on the SFP.

In the first attempt we used a Cisco 3560G switch with a MiRICi plugged into the SFP Slot and on the opposite side we used a Cisco 7301 with a PA-POS-OC3SMI port adapter.
After some testing and troubleshooting we came to conclusion that it isn't possible to create a layer 2 link between these two SDH components. The PA-POS-OC3SMI supports hdlc, ppp and frame-relay encapsulation, whereas the MiRICi SFP only GFP encapsulation.

For our second test we exchanged the PA-POS-OC3SMI with another MiRICi so to create a STM-1 link between the two SFPs. The link came up without any troubles and we could perform some iperf speedtests. The results were about 142 Mbit/s.

All in all it seems the MiRICis are working well in a lot of devices as they support the multi-source agreement (MSA) product identification codes. But you only can use them together with another GFP encapsulation compatible product. For a full list of features and more information, take a look at the manufacturer page: www.rad.com

10Gig interconnect between Brocade and Juniper - phy-mode 28k

nospam@example.com (Fredy Künzler) — Fri, 18 Jun 2010 11:17:00 +0200

by Fredy Künzler

This blog has nearly been abandoned - sorry about that. Time is running, and spare time is very limited...

Anyway, yesterday I came across a nasty issue which seems not to be documented anywhere in the web, and therefore I thought to make a note here for further reference, maybe it's helpful for someone.

We (Init7, AS13030) were about to bring up a new 10gig interconnect with a peer. They happen to run Juniper gear, while we operate Broacade / Foundry XMR routers. The cable was plugged, the light was on, rx and tx values good. The Juniper showed the port up, IP address was pingable, but the port on the Brocade router remained down. We already thought about a broken interface, before I started to play with the phy-mode parameter.

The Juniper can do only WAN-phy and LAN-phy (I don't have Juniper experience though), but the Brocade has three modes. It offers LAN-phy by default (to revert back to LAN-phy, use the 'no' parameter), WAN-phy and a mode called "28k":

(config-if-e10000-4/3)#phy-mode ?
28k Bay 28000
wan 10G WAN PHY mode

It was worth a try, and after setting phy-mode to 28k, the interface on the Brocade came up immediately. I don't know whether there is any drawback, traffic is now flowing normal... (WAN-phy mode has less capacity than LAN-phy, though).

My config looks now like this (anonymised and abridged):

#sh run int eth 4/3
interface ethernet 4/3
port-name Foo-Bar
enable
route-only
ip address x.x.x.x/30
phy-mode 28k
!

Threats to the Internet Routing and Global Connectivity

nospam@example.com (Fredy Künzler) — Wed, 16 Jul 2008 08:31:00 +0200

on blogg.ch (German, based on an English document): Threats to the Internet Routing and Global Connectivity

Sweden is monitoring IP traffic from / to Russia

nospam@example.com (Fredy Künzler) — Tue, 15 Jul 2008 22:23:51 +0200

on blogg.ch (German): Schweden überwacht russischen IP Verkehr

BOGONs should be updated every now and then ...

nospam@example.com (Fredy Künzler) — Thu, 10 Jul 2008 21:57:00 +0200

by Fredy Künzler

We recently received an email saying

Please remove 174.0.0.0/8 from your bogon filter...

and indeed, as time flies by, we had an outdated BOGON filter list implemented. IANA keeps allocating /8 to RIRs and if network administrators don't pay attention, parts of the internet could silently become unreachable.

In the past few month these blocks should have become routable:

112.0.0.0/8
113.0.0.0/8
114.0.0.0/8
115.0.0.0/8
173.0.0.0/8
174.0.0.0/8
186.0.0.0/8
187.0.0.0/8

Therefore it's probably wise to check if you see any more specifics of these prefixes (on Cisco / Foundry / Quagga type "sh ip bgp 113.0.0.0/8 longer").

I know that BOGON filtering could be automated, but we never found the time to get it done properly, and, as the available IPv4 space is running out, the Bogon Route Server Project will become obsolete anytime soon.

IPv6 now! (says OECD)

nospam@example.com (Fredy Künzler) — Sat, 17 May 2008 13:51:00 +0200

by Fredy Künzler

Even the OECD (Organisation for Economic Co-operation and Development) preaches now IPv6! ... and I guess when official institutions start to worry, it's serious. Functionaries usually don't tend to be fast movers.

I recently asked a fellow service provider about his IPv6 plans, and the answer was a bit ignorant: "noone is paying me to implement IPv6, therefore I don't have a project". Looking back to the early days of the Internet, a lot of research and test-implementations have been done withouth payment, and it became the foundation of a whole industry. Money should not be the focus when evaluating IPv6, at least for another year or so.

I gave a presentation at the recent SwiNOG #16 meeting about implementing IPv6 in a providers backbone. In fact, IPv6 is much easier than people generally think. If you are a network professional, everything you know about BGP, subnetting, OSPF etc. is still valid, at least more or less.

If you plan to deploy IPv6 anytime soon, click through the slides, they might be helpful (or download the IPv6 Now! presentation):

| View | Upload your own

IPv6-only hour at RIPE #56

nospam@example.com (Fredy Künzler) — Wed, 07 May 2008 10:49:00 +0200

by Fredy Künzler

Notes about the IPv6-only hour over there at blogg.ch. Yes, it works (more or less).

P.S. Peter, sorry. You have to bother translate.google.com again.

Less than 1000 days left to IPv4 exhaustion

nospam@example.com (Fredy Künzler) — Sat, 03 May 2008 00:56:00 +0200

by Fredy Künzler

Quoting Mike Leber of he.net:

Since nobody mentioned it yet, there are now less than 1000 days projected until IPv4 exhaustion:

http://www.potaroo.net/tools/ipv4/

Do you have an IPv6 plan?

How long do you think it will be until Sarbanes Oxley and SAS 70 auditors start requiring disclosure of IPv4 exhaustion as a business continuity risk, as well as the presence or lack thereof of an IPv6 plan?

When do you plan on telling your customers? (afterwards?)

Ahhh, you don't have any customers that have to plan to buy equipment 2 years in advance. Ok, I understand.

Mike.
ps. 1000 days assumes no rush, speculation, or hoarding. Do people do that?

pps. Of course these are provocative comments for amusement.

ppps. Or not if you don't have any kind of IPv6 plan. Sorry, sorry...

Mike, you're right. Please tell us again in 500 days ...at least I included the IPv4 exhaustion counter here ...

PS: we can deliver IPv6 feed, too.

Net Neutrality

nospam@example.com (Fredy Künzler) — Sat, 26 Apr 2008 23:06:00 +0200

by Fredy Künzler

I had an interesting discussion recently with "eyeballs" (being myself content-heavy with both Init7 and Zattoo) about Net Neutrality. It ended up with the unanswered question "what is more valuable, content or users?"

From my point of view, both are equal. Even more: content and users are symbiotic. If no content would exist, noone would buy a broadband connection and still rely on old dial-up modems to send and receive their emails. On the other hand, without a decent number of users, all nice drawn business models of content owners would be maculature.

Not new thoughts, though. Still there are many network people out there which consider (heavy bandwidth-consuming) content the root of all evil, completly ignoring that their residential broadband customers provide their paycheck.

Of course "eyeballs" are in trouble if the market price for an avarage DSL | Cable connection is below cost. This happens quite often, in some European countries due the fact that the incumbent is squeezing margins, and the regulator is weak | deaf | ignorant. What to do in such a case?

Cost optimization ... reconsider everything. Especially old-worn IP transit contracts. After 12 to 15 month the latest one should re-negotiate price with the current vendor, and ask other vendors, too. (Independent) eyeball networks should also consider buying from content networks, as they tend to have a lot of excess capacity and may be willing to sell ridiculously cheap.

On factor remains: the user. Commonly they have no clue about the political aspect of Net Neutrality, and they don't care either. Therefore I'll point to an old Youtube-Video, in case someone typing "Net Neutrality" into a search engine and gets pointed to this blog entry ...

AS1299 and AS5511 to merge?

nospam@example.com (Fredy Künzler) — Thu, 17 Apr 2008 23:58:00 +0200

by Fredy Künzler

AS1299 + AS5511 = ? Not sure how the Swedish / Finnish and the French can cope in general ... but when reading news about France Telecoms intention to buy TeliaSonera, I remembered that Anabelle (AS5511) and Christian (AS1299) have been whispering alot recently at the global peering forum. Now we all know why ... obviously they have been preparing this plot.

The history of AS1

nospam@example.com (Fredy Künzler) — Wed, 16 Apr 2008 11:36:00 +0200

von Fredy Künzler

I stumbled across a document about the history of AS1 today in a blog entry of Renesys and I thought it's worth to share:

[...] AS1 was originally assigned to Bolt, Beranek & Newman (BBN), a pioneering company in the field of network packet technology. BBN engineers are credited with developing the earliest routers, with sending the world’s first e-mail and with establishing the “@” convention used in all e-mail sent today. Through a series of subsequent corporate transactions, ownership of AS1 passed to Genuity, Inc., a communications company whose assets were acquired by Level 3 in February 2003. [...]

How open vs. restrictive peering policy affects the AS-Path length

nospam@example.com (Fredy Künzler) — Mon, 14 Apr 2008 17:50:00 +0200

by Fredy Künzler

I got the nice graph below from Oli Hitz, network engineer of senseLAN (AS31736). senseLAN is a regional cable ISP in western Switzerland, and could be classified as a Tier-3 provider. They are peering at SwissIX and are purchasing transit from Level(3) AS3356 and UPC AS6830, and added now a partial feed from Init7 AS13030 (appx. 125k routes).

While Level(3) and UPC both are following a restrictive peering policy (according to the respective entries in the PeeringDB), Init7 peers open. The average AS-Path length went down from 3,18 to 3,13 by adding AS13030. Considering that AS31736 is inbound-heavy, this view may be not too relevant for the actual traffic flow, but it's still a nice statistic (don't we all love statistics?!)

BTW I tried to find a link to UPC's Backbone website ... no luck.

Peering Community

nospam@example.com (Fredy Künzler) — Sun, 13 Apr 2008 16:55:00 +0200

by Fredy Künzler

Originally shown at the Peering BOF at NANOG #42 (I'm in the movie too!):

... about bgp-and-beyond.com

nospam@example.com (Fredy Künzler) — Sun, 13 Apr 2008 15:31:00 +0200

by Fredy Künzler I'm a blogger since July 2005, writing the blog CH Internet Szene. My very first blog entry was Cyberlink übernimmt magnet.ch, about a merger of two ISPs in Switzerland. At this time I was CEO of the independent business and wholesale provider Init7 in Switzerland. Blogg.ch approached topics as internet law, interconnections, Swiss telecommunication politics, more mergers and rumours about and of course, new technology. By the end of 2007, I stepped down as CEO of Init7, remaining to be member of the board and shareholder. I'm now focusing on Init7's strategy, and this is mainly building the international backbone of Init7, operational with the AS-Number 13030. My current employer is Zattoo, a startup company in the IP-TV space. Unlike classical IP-TV Zattoo is a piece of software which allows to watch live television on the computer. No setop box is required, Zattoo doesn't aim to replace the TV set in the living room; Zattoo rather allows to watch TV while surfing the web, chatting, writing emails or working on spreadsheets. I've joined Zattoo in June 2006, first as a contractor, later as an employee, and currently I'm the network architect of the company. The Zattoo P2P supported CDN is my field, among BGP routing, datacenters and other related stuff. Together with the experience of more than a decade now in the ISP and telecommunication industry gives me some insight, which I would like to share here. While Init7 operates it's business in Switzerland with the support of the international backbone, joining Zattoo shifted my focus truely global. That's why I hang out on events as NANOG or the Peering Forum. At the last forum, fellow community member Peter Cohen asked me why my other blog CH Internet Szene - blogg.ch is not written in English. The answer is simple: my mother-tongue is German. Peter suggested a change, because the machine translation of Google is, well, not bad but sometimes funny. Changing the language of a blog is generally not a good idea, but starting a 2nd one seems to be more adequate. And that's the story about bgp-and-beyond.com ... kudos to Init7-Engineer Basile for suggesting the domain name. The older sister blogg.ch remains active with a focus on Switzerlands internet. Expect references to blogg.ch here, and don't expect translations for everything. ;-) More information about me can be found on the about-page of blogg.ch and in my Xing-Profile.