Home » Core switching network upgrade

Core switching network upgrade

Serverius is renewing its switching network to cutting-edge VxLAN with EVPN overlay powered network, ready for all the exciting plans we have for the coming years.

Our NOC department will replace all the current Cisco network switches with Huawei which will result in a huge capacity upgrade. This large core switching network upgrade started with 250 times 100G ports.

Sergey Petukhin, head connectivity infrastructure

From the very beginning of the Serverius datacenters, we have relied on various iterations of Cisco Catalyst for our switching and routing. In later years most of the BGP routing moved to much more capable Huawei routers but the Ciscos kept playing an invaluable role as core switches. Now the time has come to say goodbye to our trusty green friends and take a big step into the future.

In the last few months we have investigated and tested a number of solutions from different vendors with various levels of success. From the beginning it was clear that we would standardize on VxLAN/EVPN to support our future growth and all the exciting plans we have. It seems to become the industry standard for large datacenter networks and is widely supported by most major equipment brands.

Another problem we would like to solve is the cabling nightmare that comes with a growing datacenter and a small number of very large switches. All connections to the Serverius network end up in just a few racks that slowly start looking more like a bowl of spaghetti. In our new design we build a fabric of smaller switches which will be distributed throughout the datacenters. This keeps the number of connections per switch down and the cables much shorter, allowing for faster delivery and happy support engineers.

These and many more challenges were easily met by our partner Huawei and there CloudEngine switches. Together with Huawei we designed our new network around a 100 Gbps leaf-spine architecture, a routed VxLAN overlay network and an EVPN control plane. The expertise of their engineers and technological excellence of their hardware gives us the confidence to migrate from our proven Catalysts to their solution and keep on expanding for the foreseeable future.

In the coming months we will start installing and configuring the new switches and thoroughly testing the technology at scale. The parts of the network that are currently at their maximum capacity will be the first to migrate. These are mostly internal links and will have minimal noticeable impact. Eventually we will work with our customers to move their links to the new network as well with minimal or no downtime.

Old network design

New network design

Gijs van Gemert, Managing director of Severius IT infrastructure

Using Huawei as hardware supplier

As most people know, Serverius is build on a multi vendor infrastructure principle. The redundant fibers to our datacenters are from different suppliers, the DWDM equipment on it is Adva and Infinera (one brand per fiber path). With a routing and switching network it’s not wise to use multiple networks brands together. Although it’s redundant by many hardware devices in any possible way, it’s still just one network where every piece of hardware should exact work as one. Using multiple brands will higher the chance of miscommunication between hardware, where during debugging situations may arise in which the different suppliers will blame each other for the fault cause. To prevent this kind of situations it’s always better to use just one vendor. 

Which network vendor is best to use? In the past 20 years I have seen many brands dominate market segments. Where 15 years ago the average Amsterdam datacenter was totally Cisco green, it was clear the Cisco was the dominant player in our industry. But in the past few years we have seen this changing to other brands like Huawei and many others. Partly because the known brands like Cisco/Juniper/Brocade(Extreme) where beaten by others in functionality, support and sometimes in price. And also because vendors starting to offer hardware where engineers could add their own software layer on to it, because these days it’s more and more about the software which is running on boxes and less about the hardware itself. If I look to our clients in general I see that in most cases private networks are choosing this kind of hardware/software setups. But most larger non global carrier networks like us where its infrastructure continuously anticipate on external factors like different clients, cyber security, changing layer 7 hosting hardware and software under it, etc. are choosing a commercial brand for switching and routing. Simply to lower the risk of running into problems and if problems do arise by for example bugs, to get vendor support which can solve this kind of problems faster because hardware and software is from the same vendor. This is one of the reasons that Serverius choose Huawei.

Politics about Huawei

Like Cisco in the past, it are the Chinese brands today which are used in every Dutch IP network (mainly direct or indirect). Currently all Dutch commercial datacenter networks, financial sector networks and even government institutes are using Huawei for their data communication and/or data storage. All of them have a different reason of course but in general its the technology level which is the deciding factor. Not pricing like some will think, because during this project its upfront pricing comparison we have seen that other known suppliers offered us the same or even cheaper hardware. Secondly, as a mentioned it’s no politics where companies like us have their decision based on, our industry is based on technical functionality and facts. In our case it’s our good experience with the amount hardware and software failures, the speed of product development to beat the competition and respectful personal behavior of the past years what concludes to go with Huawei again. Also because with our current core routers we are using Huawei for years now, and we only can tell that they are very stable, without hardware failures, we always got good updates and during our visits with the R&D people of Huawei we got always an open communication with a high level human respect to both sides. This is something other brands can learn allot… We understand that currently there is allot of media attention about the possibility of spying from the Chinese government via Huawei hardware. Based on the fact that till today there is no proof of spying of Chinese government by Huawei hardware/software back-doors and only allot of political suggestions without proof, Serverius will as a neutral Dutch datacenter not make his choice dependent on external political motives. Especially when this spy suggestions are sent by those who do the same or even more. We are engineers and we measure on facts. That’s it. This does not alter the fact that we are very aware that there is always a possibility of espionage in our network, because in this age of internet addiction, larger networks such as Serverius are not interested in foreign governments. Yes, we understand that it could be possible because it’s known that European, Russian, American and Chinese governments spy on each other for ages and always will, so it could be possible from every other foreign government by every other brand. So if it turns out that there is nevertheless a backdoor, leak, etc, then we will take the necessary steps at that moment, whereby a broad risk comparison of the past and future of alternative suppliers will be made.

The European discussion about China is about safety, but in reality it’s about China’s rise as a technological superpower. If we in the Netherlands believe that there should be a counterbalance with western technology, then our government should create an long term environment for local companies which are capable to produce innovative technology (much more than only supporting the educational system and development tax discounts). But as a small Dutch country we should always be realistic and embrace the technology of others, even if it’s Chinese, American or else.

Better government rules

My personal opinion is that if Dutch politic parties really want to exclude non western European brands in our Dutch infrastructures, they should offer an European alternative first. Secondly, they should push way more effort in to the Dutch (European) product development, improve the international business climate for Dutch companies and they should not allow the sale of Dutch companies to foreign parties. As a result there will be running infrastructure projects with real technology alternatives where our IT engineering industry will use its building blocks from. But do not prohibit the import of foreign technology, because that is ultimately only harmful to our self. Our government should arrange good regulation rules which are widely supported by young people from our datacenter, carrier and hosting sector, where it should not be a reflection of a hand full of people from some Dutch IT sector organization or American government its Den Haag lobby. Politicians should make a better strategy based on information from our Dutch industry, whereby international companies which are established in the Netherlands (with less ethically bonding to our society) are excluded.

And even though the Dutch citizens nowadays only wants have the convenience to perform all kinds of things online, our government should not allow the really critical things of our country are be connected to the internet. It will save us from possible data breaches with state interests and as a result a lot of (non) discussion from those who do not understand how the internet is really working. Hopefully it will keep the remains of the free Dutch/European internet we currently still have.