Programs become more important than iron

After reading the news about staff redundancy at Cisco, I decided to phone my good friend, who has been working for more than one year for the benefit of this network equipment manufacturer. A friend confirmed that they were really going to fire some of the employees. True, he voiced a more modest figure than the one that appeared in the article of one Russian Internet edition.

This is not the first cut. And as it was in past years, the main reason is the reorientation of the company to software development. He himself moved to another department, where he is now engaged in the support of cloud products.

At the new job the best friend of my friend, like many other network engineers, became a python. Not one that creeps, but one that is a programming language. Python is in demand for all sorts of projects to improve the internal systems, procedures and automation in the company. In general, this programming language has become quite often featured in various documents, presentations and vendor solutions. For example, support for this programming language has appeared in Nexus switches. In general, allowing users to create the necessary logic for the behavior of network equipment is correct. At this rate, soon we will have to write various programs at Cisco courses. And instead of show run we will output Hello world.
')
Exchanging a few words about the main trends, it turned out that everyone is full of thoughts about neural networks, Big Data, machine learning and other transcendental technologies. This is probably correct. After all, switches, routers, access points make a revolutionary, difficult. But while there is a persistent feeling that these transcendental decisions are a little about us. Anyway, at the moment. Although in terms of the same security, Cisco already offers cloud-based security solutions based on implementations like artificial intelligence.


Production of software products are becoming more interesting for business than hardware solutions. Firstly, it is no secret that iron sales are slowly falling. This also affects the clouds, the transition to which leads to the fact that the company's own infrastructure, moving into the clouds, becomes poorer. And serious competition, which does not go anywhere, but only intensifies, fueled by the constant entry of new companies into the market. And at the same time, iron becomes relatively standard (sometimes it seems to me that many products from different manufacturers are made at the same Foxconn factory), competition is only growing.

Secondly, in terms of finance, software production seems more profitable than hardware. There is no need to solve problems with the construction of factories and their maintenance, remnants of warehouses, components, defective equipment, etc. To sell licenses is a more pleasant and marginal occupation. Indeed, in fact, in this case air is sold practically. However, if this air constantly podglyuchivaet, it becomes expensive to maintain it.

If we are talking about products that are traditionally software, a deeper reorientation to them does not cause any questions. The same can be said about software-defined networks (SDN). This concept opens up vast spaces for creating new software solutions that are closely related to network equipment. The scope of SDN is huge. Starting from automating the routine processes of setting up network equipment. And ending with the management of all elements of the IT infrastructure in order to automate various processes and obtain completely new services.

But what about the switches, routers and other devices that we used to see in the iron version? Here comes to the fore virtualization technology. The recipe is relatively simple: virtualize network services and move away from the iron part. If we look at Cisco, we will notice that the vendor has been following this path for quite some time. There is a systematic transition from the uplines scheme to the virtual machine scheme on anything. Of course, for Cisco, it’s perfect if it’s a UCS server. But if not him, then the vendor, I think, will not be very offended.

At the moment in the Cisco product line, we see that many solutions are offered in the form of virtual implementations. Modern processors, coupled with the support of optimized processing of network traffic, allow implementing many solutions without resorting to the use of specialized hardware (ASICs, network processors, etc.). We mentioned this partially in the CBS company's blog about control and data plane in network equipment .

Now in the form of a virtual solution, we can get a router (for example, Cisco ISRv), a firewall (for example, Cisco ASAv), a wireless network controller (for example, Cisco vWLC) and other solutions. Other market players offer similar solutions: Juniper, Huawei, etc. And since in most cases we have Ethernet at the entrance, all these solutions can be run on the basis of a regular server.


It all comes together under the umbrella of the Network Functions Virtualization (NFV) architecture. It should be noted that the virtualization of services allows significant progress in the direction of the clouds, and SDN. Automatic deployment of services here and now is a simpler task when these services are virtualized. Therefore, for data centers, NFV solutions are extremely interesting. But the enterprise is also slowly catching up. We have already got used to something in virtual performance. To something else not. For example, we still want to see a router or a firewall in our office as a separate box. It seems both safer and safer. But the market moves us further by offering fully virtualized office solutions. For example, Cisco, complementing the NFV solution with orchestration, monitoring, and hardware resources (the vendor does not want to get rid of the hardware itself), offers an Enterprise NFV architecture. Its main components are:

• Enterprise Service Automation (ESA) orchestration solution based on APIC-EM (SDN controller) and Prime Infrastructure monitoring;
• various NFV solutions;
• software based on which all services are launched: a Linux-based hypervisor with additional monitoring, switching, etc .;
• hardware resources: UCS or UCS-E server.

Enterprise NFV should allow to reduce the time of implementation of various services in the office, making the process as automated as possible. Somewhere I have already heard. In my opinion, this approach does not seem too vital. Although time will tell. Whatever it was, but the motion vector is set.

With most network devices, everything is clear. But what about the usual iron switches? With them harder. Still, patch cords need to be connected somewhere, and obviously the server is not very suitable for this. But here on the market there are products that allow to separate the software implementation from the hardware. These are the so-called bare metal switches. Such a switch goes without an operating system that the user chooses and installs independently. Such solutions are in HPE, Edgecore, Dell, etc.

Nowadays, you can safely concentrate on developing software solutions without significantly losing presence in various market segments. Of course, the production of specialized iron solutions, where the software is practically inseparable from the hardware, will continue as long as the market is interested in it. These are powerful cabinets and cabinets for handling a huge amount of traffic. And, for example, wireless access points. Although no, with access points the question is interesting. If you look at the hardware of these devices, you can see that many of them are built on the Broadcom chipset. By and large, only the software and the device case are changing.

Thus, there is a tendency that many solutions can be offered to the market in isolation from the iron component. I, as an engineer, like this opportunity, as it allows you to test almost any solution without buying expensive equipment.