Noction is pleased to announce the release of Intelligent Routing Platform 3.6. The major feature available in the new product version is
The post Noction announces the release of Intelligent Routing Platform 3.6 appeared first on Noction.
I've been researching overlay network strategies recently. There are plenty of competing implementations available, employing various encapsulations and control plane designs. But every design I've encountered seems ultimately hampered by the same issue: scalability at the edge.
Imagine a scenario where we've got 2,000 physical servers split across 50 racks. Each server functions as a hypervisor housing on average 100 virtual machines, resulting in a total of approximately 200,000 virtual hosts (~4,000 per rack).
In an ideal world, we could allocate a /20 of IPv4 space to each rack. The top-of-rack (ToR) L3 switches in each rack would advertise this /20 northbound toward the network core, resulting in a clean, efficient routing table in the core. This is, of course, how IP was intended to function.
Unfortunately, this approach isn't usually viable in the real world because we need to preserve the ability to move a virtual machine from one hypervisor to another (often residing in a different rack) without changing its assigned IP address. Establishing the L3 boundary at the ToR switch prevents us from doing this efficiently.
I've been researching overlay network strategies recently. There are plenty of competing implementations available, employing various encapsulations and control plane designs. But every design I've encountered seems ultimately hampered by the same issue: scalability at the edge.
Imagine a scenario where we've got 2,000 physical servers split across 50 racks. Each server functions as a hypervisor housing on average 100 virtual machines, resulting in a total of approximately 200,000 virtual hosts (~4,000 per rack).
In an ideal world, we could allocate a /20 of IPv4 space to each rack. The top-of-rack (ToR) L3 switches in each rack would advertise this /20 northbound toward the network core, resulting in a clean, efficient routing table in the core. This is, of course, how IP was intended to function.
Unfortunately, this approach isn't usually viable in the real world because we need to preserve the ability to move a virtual machine from one hypervisor to another (often residing in a different rack) without changing its assigned IP address. Establishing the L3 boundary at the ToR switch prevents us from doing this efficiently.
I've been researching overlay network strategies recently. There are plenty of competing implementations available, employing various encapsulations and control plane designs. But every design I've encountered seems ultimately hampered by the same issue: scalability at the edge.
Imagine a scenario where we've got 2,000 physical servers split across 50 racks. Each server functions as a hypervisor housing on average 100 virtual machines, resulting in a total of approximately 200,000 virtual hosts (~4,000 per rack).
In an ideal world, we could allocate a /20 of IPv4 space to each rack. The top-of-rack (ToR) L3 switches in each rack would advertise this /20 northbound toward the network core, resulting in a clean, efficient routing table in the core. This is, of course, how IP was intended to function.
Unfortunately, this approach isn't usually viable in the real world because we need to preserve the ability to move a virtual machine from one hypervisor to another (often residing in a different rack) without changing its assigned IP address. Establishing the L3 boundary at the ToR switch prevents us from doing this efficiently.
Shifting to cloud infrastructure can help speed an enterprise digital initiative.
In March 2016 my friend Matt Oswalt announced a distributed network testing framework that he used for validation in his network automation / continuous integration projects. Initial tests included ping and DNS probes, and he added HTTP testing in May 2016.
The project continues to grow (and already got its own Github and documentation page) and Matt was kind enough to share the news and future plans in Episode 63 of Software Gone Wild.
To ask questions about the project, join the Todd channel on networktocode Slack team (self-registration at slack.networktocode.com)
When we launched Universal SSL in September 2014 we eliminated the costly and confusing process of securing a website or application with SSL, and replaced it with one free step: sign up for Cloudflare.
When you complete the sign-up process, we batch your domain together with a few dozen other recently signed-up domains, and fire off a request to one of our Certificate Authority (CA) partners. The CA then sends us back a shared certificate covering the root (e.g. example.com) and top-level wildcard (e.g. *.example.com) of your domain, along with the hostnames of the other customers in the request. We then package this shared certificate with its encrypted private key and distribute it to our datacenters around the world, ensuring that your visitors’ HTTPS sessions are speedy no matter where they originate.
Since that process was created, we have used it to secure millions of domains with free Universal SSL certificates and helped make the Internet a faster and more secure place.
But along the way we heard from customers who wanted more control over the certificates used for their domains. They want Continue reading
