Original Design Manufacturers (ODMs) that produce incumbent profit busting white box switching technology could soon be releasing the next wave of programmable networking based on technology from a silicon company best known for it’s encryption products. Cavium have released the XPliant chipset which it acquired from a $90m purchase earlier this year. This chipset comes in four flavours varying from 880 Gbps to 3.2 Tbps. This results in devices having 128×25 Gbps switching lanes allowing switches with 32x100GbE, 64x 50/40GbE, or 128x 25/10GbE ports in a single device. The highest speed Cavium device is currently twice the speed of the next highest merchant silicon offering, however merchant vendors will catch up with the speed aspect before too long. The important part here to remember is this chipset is programmable and is touted to be released with support for Generic Network Virtualisation Encapsulation (GENEVE) out of the box, along with a “simulator” for product designers to test their code against. All designed to increase the speed to market and decrease delay.
Let’s take an ODM switch from the likes of Accton that is currently based on the venerable Trident II chipset. Current merchant silicon chipsets limit the features to those Continue reading
As horizontal scalability demands increase for policy based fabrics such as Cisco’s ACI and performance demands push the development of high speed Ethernet standards like 25G, 50G, 100G and beyond in to 400G, how that data is pushed and pulled on the fabric begins to become a problem. In the words of Scotty, devices of today “canne take any more” due to issues with physics, traditional approaches to ASIC interfaces and currently used materials.
A company that was born to attack this problem from a different direction has recently been acquired by Cisco. Memoir, the said acquired startup, has been inserted in to the Insieme business unit within Cisco which says much about the strategy of the policy capable hardware and the company faith in the direction of the ACI strategy. So why Memoir? They offer a solution for multiple areas of memory to be addressed concurrently, making the operation more akin to a parallel one instead of a fast serial operation. They call this: Algorithmic Memory™ and it can increase memory options ten fold!
Cisco in keeping with their origins will always try and hold the hardware networking space. As commodity devices start threatening profits in bread and butter networking, Continue reading
After watching the Tech Field Day (TFD) events for a while, I decided to fill out the form and apply to be a delegate. With the events being based in the USA, me being based in the UK and my status not being at the power blogger level of the likes of Ethan Banks, Greg Ferro or Ivan Pepelnjak, the perceived chances of actually being selected to go were negligible to none. So how surprised was I when I received an email with an invite? You could have blown me over with a feather, so much so, the whole side of the train carriage I was sitting in at the time all heard the “whoop whoop” I decided to share!
So for any new delegates or those that want to know how it plays out, your travel, accommodation and pretty much all arrangements are taken care of by Gestalt IT and the TFD team. You just have to worry about getting to and from your chosen airport to depart and return.
The week that the event takes place in is northing short of hectic and by my experience was superbly executed by Steve Foskett and Tom Hollingsworth. You can pretty Continue reading
Not really the line anyone wants to hear, especially after watching the Terminator films! This however isn’t what this post is about, so if you’re a bit of a rebel, fear not. No network vendor branded termination thing (maybe other than poor documentation or code) will result in your death.
Since the era of the abacus, little consideration has been given to how software that relies upon a computer network actually interacts with it. Sure, most developers know how to drive a socket library and make things happen at a session level, but almost no consideration is given by a developer on how to deploy an enterprise application to a production environment.
This post represents a set of thoughts that have been maturing over the last few months. They are very much my own thoughts and do not represent those of others. I would be interested to hear if you have the same thoughts or any interesting different takes.
Before smart phones and tablets came along, software for the domestic populous provided a means of typing and printing spell checked letters to your pen pals, figuring out your weekly shopping Continue reading
With the current interest in network automation, it’s imperative that the correct tools are chosen for the right tasks. It should be acknowledged that there isn’t a single golden bullet approach and the end solution will be very much based on customer requirements, customer abilities, customer desire to learn and an often overlooked fact; the abilities of the incumbent or services provider.
The best projects are always delivered with a KISS! Keep It Simple Stupid.
Note – I have used the term ‘playbooks’ as a generic term to define an automation set of tasks. Commonly known as a runbook, playbook and recipe.
Because a services provider may have delivered an automation project using a bulky generic work-flow automation tool in the past, it does not mean it is the correct approach or set of skills for the current and future set of network automation requirements. To exercise this, let’s create a hypothetical task of hanging a picture on the wall. We have many choices when it comes to this task, for example: bluetak, sellotape, gaffer tape, masking tape, duct tape or my favourite, use a glue gun! However, the correct way would be to frame Continue reading
We have mostly all been burnt to a level of severity that we will or will not admit to by prodding and poking networks. Whether by an unexpected bug, lack of understanding of the thing we are poking, or sheer ‘bad luck’, there’s no avoiding it.
Being burnt by a network is almost like being zapped by a cattle prod. It doesn’t take many times before your brain rewires itself to avoid getting burnt, unless you’re a network masochist, in which case, you’re a special breed. This rewiring has resulted in using the CLI as an investigatory and validation tool as well as a configuration access method. What was that keyword again?
show ip bgp neighbor ?
Due to mistrust in the documentation, lack of desire or over trusting the CLI, our brains have become used to this behaviour and complacency has set in.
As we shift from configuring network elements manually to configuring them by automated template generation and structured API calls, will our well understood knowledge of a networking operating system with all of it’s caveats and nuances become redundant along with our bad habits? So do we just trust an amorphous piece of software Continue reading
Too many times the question has been asked “How do I adopt DevOps and will I need to become a programmer?”. My own beliefs are almost flipped upside down and my money is on DevOps adopting you; possibly without you even realising it. If you’re reading this, it’s a sign of it already happening or at least the thought process! The answers lead to the questions. There is no spoon.
So what’s NetDevOps? Networking + DevOps. Simples. It’s a thought process and a movement. Not an intended starting point, but a natural end point. A natural and evolutionary extension to your skill set, not a new one.
Instead of rushing towards a “golden bullet” product, it might prove a valid route to start thinking about the network as data derived from configuration schemas. Even if it’s a set of side line exercises from what you do day to day, thinking about the network in data or ‘code’ is a step towards what most magic products are aimed to obfuscate you against. Remember what it was like learning IOS back in the day? CatOS? Linux? First time you used a bash shell? This movement is no different.
The cultural divide between delegation of network control such as in the case of Cisco ACI, or VMware’s NSX is a cause for questioning. These are off the shelf products designed to solve a problem that exists today but introduce new problems to the organisations that acquire them. Who controls the network and at what stratum(1)? In the case of automation and orchestration products, who creates the templates? Who is allowed to trigger automation and orchestration events and even more importantly, when is that person allowed to do it? As the virtual networks are virtual, does control belong with the virtualisation team? Hrmmm. Lots of questions. Many of these are dependent on the company, customer and situation and might not be solved with the most marketed product. A regimented set of answers doesn’t exist, nor do I think there will ever be out of hyper scale data centre environments.
It has taken years for the DevOps community to understand how to handle the requirement of rapid and agile deployment. We’re not the first ones to go through this pain. Can you imagine a tightly controlled ITIL governed network suddenly being comfortable with partial or fully automated approach to network Continue reading
This post is for anyone who’s thought about deploying VXLAN on their network and who like me thinks deeply about stuff, to the point of utter confusion, which hit me on a very hot sunny afternoon. The good news with confusion is, once you’ve cracked the issue (normally one’s correct understanding), the clouds clear and the birds sing.
Virtual Extensible Local Area Network functionality is beginning to hit a wide range of vendor devices. VXLAN provides 16 million (and then some) segments for Layer 2 networks. Some organisations, especially those providing cloud infrastructure currently have or will have problems with the number of VLANs available in 12 bits, which is 4096. In that 4096 number, some are not usable either are reserved for certain things like token-ring and platform specific internal communication.
VXLAN is a simple encapsulation method or a tunnel. It encapsulates the original payload in to UDP packets for transit across an IP network and adds another 50 bytes on to the header tax. At a very high level, VXLAN can be deployed in multicast mode and with unicast. Virtual Network IDs (VNIDs) represent VXLAN segment identifiers. In order to gain connectivity, a network construct Continue reading