Last week, a very small number of our users who are using IP tunnels (primarily tunneling IPv6 over IPv4) were unable to access our services because a networking change broke "path MTU discovery" on our servers. In this article, I'll explain what path MTU discovery is, how we broke it, how we fixed it and the open source code we used.
When a host on the Internet wants to send some data, it must know how to divide the data into packets. And in particular it needs to know the maximum size of packet. The maximum size of a packet a host can send is called Maximum Transmission Unit: MTU.
The longer the MTU, the better for performance, but the worse for reliability, because a lost packet means more data to be retransmitted and because many routers on the Internet can't deliver very long packets.
The fathers of the Internet assumed that this problem would be solved at the IP layer with IP fragmentation. Unfortunately IP fragmentation has serious disadvantages and it's avoided in practice.
To work around fragmentation problems the IP layer contains a "Don't Fragment" bit on every IP packet. Continue reading
As is our annual tradition, this blog provides a year-end review of how the Internet providers at the top of our IP Transit Intelligence global rankings (formerly, Renesys’ Market Intelligence) fared over the previous year. The structure and performance of the Internet remains a huge blind spot for most enterprises, even those critically dependent on it for business operations. Whether it’s the next 3 billion people coming online, poor performance due to suboptimal routing, or security breaches of a trust-based Internet infrastructure, Dyn provides critical insight into the structure and performance of the Internet, both real-time and historical, via its Internet Intelligence products. More importantly, our services help our customers make the changes necessary to optimize Internet availability, reliability, and reach. This blog reviews a single very small slice of our data related to the sizes of the top global players as it pertains to the markets and customers they serve.
Back in 2008, we chose to look at the 13 providers that spent at least some time in the Top Ten that year, hence the name “Baker’s Dozen“. We repeated that exercise in 2009, 2010, 2011, Continue reading
This post is a Zombie that I'm resurrecting from my drafts. I"m not doing any Java these days, but hopefully this post might be useful to somebody
In my quest to get better code coverage for the OVSDB project in OpenDaylight I started to look at increasing coverage for the REST API. It's pretty difficult to test this in an efficient way (lines of code) and frameworks like Robot would have been easier to use. The disadvantage with using an external test framework is that code coverage (using a plugin like JaCoCo) would not be logged. Therefore I harnessed my Junit-Jitsu and found a solution that lives in the JVM
Lets take a very simple example REST API
GET, PUT: /v2/foo
The solution uses the following components
The parameterized runner will run run a test multiple times given a bunch of parameters. This way we can write one test, specifiy our parameters in YAML and let JUnit do the hard work!
Here's a sample YAML file:
---
- name: testGetAllFoo
operation: GET
uri Continue reading
This post is a Zombie that I'm resurrecting from my drafts. I"m not doing any Java these days, but hopefully this post might be useful to somebody
In my quest to get better code coverage for the OVSDB project in OpenDaylight I started to look at increasing coverage for the REST API. It's pretty difficult to test this in an efficient way (lines of code) and frameworks like Robot would have been easier to use. The disadvantage with using an external test framework is that code coverage (using a plugin like JaCoCo) would not be logged. Therefore I harnessed my Junit-Jitsu and found a solution that lives in the JVM
In the final part of our MPLS-focused discussion, Seamus wanted to know how one could combine MPLS/VPN, MPLS-TE and QoS (for example, sending VoIP traffic for one customer over a different path).
Short answer: don’t even think about doing that. The added complexity is not worth whatever extra money you’ll be charging the customer (or not).
This post is a Zombie that I'm resurrecting from my drafts. I"m not doing any Java these days, but hopefully this post might be useful to somebody
In my quest to get better code coverage for the OVSDB project in OpenDaylight I started to look at increasing coverage for the REST API. It's pretty difficult to test this in an efficient way (lines of code) and frameworks like Robot would have been easier to use. The disadvantage with using an external test framework is that code coverage (using a plugin like JaCoCo) would not be logged. Therefore I harnessed my Junit-Jitsu and found a solution that lives in the JVM
Why Network Engineering Is a Top Job for 2015
According to career marketplace Glassdoor, one of the top 25 best jobs in America for 2015 is “network engineer.” No surprises here. Network engineering is a high growth industry: interesting and challenging work, tons of autonomy in how to solve problems and come up with solutions, well paying, and most importantly, the knowledge that you’re building something and bringing creations to life.
Indeed, the complexities of modern networks make network engineers indispensable, with the mix of virtual and real servers, cloud services and data centers, and of course, the integration of SDN into your network. On a good day, you get that rush of power from doing what is essentially mad science.
It should be noted that network engineering was one of eight IT-based jobs to crack the top 25. Clearly, technical skills are in demand.
You know, it brings me back to Nicholas Carr’s books “Does IT Matter?” and “The Big Switch,” which expanded on a 2003 article he wrote in Harvard Business Review claiming that “IT Doesn’t Matter.” Essentially, Carr predicted Continue reading
Last fall, I attended the Tech Field Day NFD8 event, and one of the presenting companies was Nuage Networks. This was actually the second time I’d seen Nuage present at an NFD event, the first one being NFD6 a year earlier. Upon my return from NFD8, I did a short write-up of each presenting sponsor for my coworkers at H.A. Storage Systems to keep them informed. The following is my recap of Nuage Networks after their presentation in which I explain why I think Nuage is really on-target with their SDN solution and is definitely a solution to keep an eye on.
Thank you Ed for sharing the video the led me to find this…
The post Binary Metal Guitar appeared first on Router Jockey.
Four network programmability certs, two exams for each cert (with matching video courses for each exam) – plus one introductory video. Today’s post discusses that introductory video course, namely:
No muss, no fuss – jump into the post for more details.
Cisco recommends their “Introducing Network Programmability Fundamentals” course to anyone who is:
Basically, before diving into the other exams and courses, this introductory course sets the stage. Do you have to watch it to get the certifications? No. Can you benefit? Sure. But if you do want to use this course, take it first in sequence, before taking the other Cisco network programmability courses.
Covering the bases, this list gives the highlights of the course:
Copper is heavy. I’m not talking about it’s atomic weight of 63 or the fact that bundles of it can sag ceiling joists. I’m talking about the fact that copper has inertia. It’s difficult to install and even more difficult to replace. Significant expense is incurred when people want to run new lines through a building. I never really understood how expensive a proposition that was until I went to work for a company that run copper lines.
According to a presentation that we saw at Tech Field Day Extra at Cisco Live Milan from Peter Jones at Cisco, Category 5e and 6 UTP cabling still has a significant install base in today’s organizations. That makes sense when you consider that 5e and 6 are the minimum for gigabit Ethernet. Once we hit the 1k mark with speeds, desktop bandwidth never really increased. Ten gigabit UTP Ethernet is never going to take off outside the data center. The current limitations of 10Gig over Cat 6 makes it impossible to use in a desktop connectivity situation. With a practical limit of around 50 meters, you practically have to be on top of the IDF closet Continue reading
Four network programmability certs, two exams for each cert (with matching video courses for each exam) – plus one introductory video. Today’s post discusses that introductory video course, namely:
No muss, no fuss – jump into the post for more details.
Cisco recommends their “Introducing Network Programmability Fundamentals” course to anyone who is:
Basically, before diving into the other exams and courses, this introductory course sets the stage. Do you have to watch it to get the certifications? No. Can you benefit? Sure. But if you do want to use this course, take it first in sequence, before taking the other Cisco network programmability courses.
Covering the bases, this list gives the highlights of the course: