Tag: Nikhef

  • Pixelflut: A weekend of pinging on SURFnet Infinity

    Pixelflut: A weekend of pinging on SURFnet Infinity

    This blog post is a collaboration between Max Mudde, Edwin Verheul and Joachim Opdenakker.


    Network stress test 

    A new network – and, in particular, a wonderful collaboration with Nikhef to host one of the core sites – naturally calls for a festive opening. Read more about it here: https://www.surf.nl/nieuws/surf-brengt-hart-van-netwerk-terug-in-eigen-cooperatie

    But as a techie, you naturally want the technology to speak for itself. In collaboration with Nikhef and Studenten Net Twente, we came up with the idea of routing ‘difficult’ traffic through the SURFnet Infinity network. This allowed us to display some impressive graphs during the ceremony, and afterwards, ‘just for fun’, to open this up to the public as a fun internet event.

    But what exactly constitutes ‘heavy’ traffic? Gigabits? Terabits? Although we usually measure network capacity in bits per second, a router does not operate using bits. A router works with packets. Every packet, large or small, takes roughly the same amount of work: the router has to receive it, look up the destination, check whether the packet is allowed through, update counters and then forward the packet via the correct port. Whether that packet is 64 bytes or 1,500 bytes makes hardly any difference.

    Normal internet traffic is a mix of large and small packets. To account for this, we generally use the concept of IMIX, or Internet Mix. This is a standard distribution of the traffic observed in practice on networks. If you take the average of this, you arrive at an average packet size of 350 to 400 bytes. A connection with a capacity of 400 gigabits per second, carrying IMIX traffic, means that the router has to process around 140 million packets per second. 

    But IMIX isn’t enough, of course; we want to show what the network is capable of. Enter: the pixelflut!

    What is a ‘pixel flut’? 

    Imagine a huge digital painting that anyone on the internet can draw on at the same time. There’s no brush or mouse. You draw by sending messages via the internet. Each message colours a single pixel on the screen. You can follow the result live via a livestream. That’s a ‘pixelflut’. The idea comes from the hacker scene and is a regular feature at hacker events. 

    The screen is shared, so anyone can overwrite someone else’s work. If you want your drawing to stay on the screen, you have to keep resending your pixels – and do so faster than all the other ‘pingers’. This creates a fun, chaotic competition. Participants write clever programmes that send as many messages per second as possible. This makes a ‘pixelflut’ an excellent stress test for a network too. 

    What is a ping? 

    A ping packet is just about the smallest packet there is. In this SNT Ping Test, all the information is contained in the destination address, so the contents of the packet can remain empty. To fill that same 400 Gigabit per second connection with such ping packets, the router will have to process around 600 million packets per second. For the same volume of traffic, the router therefore has 4.28 times as much work to do as with IMIX traffic.

    On a ‘normal’ pixel-based chat service, you send text commands to a server. Last weekend’s edition, SNTPings by Studenten Net Twente, works differently: there, you ‘draw’ using pings. 

    A ping is the simplest message you can send over the internet. It works a bit like shouting “Hello, are you there?” in a dark room. Your computer sends a small packet to an address. If that address exists and the device is reachable, a reply comes back. Network administrators have been using ping for decades to check whether a computer or server is reachable and how fast the connection is. Virtually every computer has a built-in ping command. 

    How does a ping become a coloured pixel? 

    The trick lies in the address you send the ping to, not in the contents of the packet. 

    Every device on the internet has an IP address, which is similar to a postal address. The latest version, IPv6, has extremely long addresses. There are so many of them that there is more than enough space to hide information within the address itself. SNTPings makes clever use of this. An address looks like this: 

    2001:610:5ea:221e:<X>:<Y>:<B><G>:<R><A> 

    The numbers are written in hexadecimal, a number system commonly used by computers (using 0–9 and a–f). For example: with 

    ping6 2001:610:5ea:221e:0019:0019:00d1:ffff 

    Colour the pixel at position (25, 25) a bright yellow, in SNT’s corporate colour. 

    With SNTPings, the receiver does not send a reply. It simply reads the address to which the ping was sent, extracts the position and colour from it, and colours the corresponding pixel on the canvas. One ping therefore equals one pixel. The screen is 4K (3840 × 2160 pixels, over 8 million pixels in total). This means that millions of pings are needed for a single complete image, and you have to keep repeating this process as long as others are drawing over it. 

    A whole weekend of gaming on SURFnet Infinity 

    SNTPings was organised to mark the launch of SURFnet Infinity. The event was open to everyone from 6.00 pm on Friday 9 October until Sunday 11 October. It was the biggest edition ever. The receiving servers were located at Nikhef in Amsterdam. SURF and Nikhef together provided 10 × 400 Gbit/s of connectivity, and the receiver itself was capable of processing up to 1.6 Tbit/s of pings. By way of comparison, that is thousands of times faster than a high-speed fibre-optic connection at home. 

    A lot was going on behind the scenes: 

    • SURFnet Infinity router: All traffic first entered via the SURFnet Infinity peering router, an HPE/Juniper PTX10002-36QDD. 
    • Nikhef router: That traffic was forwarded to the Nokia 7750 at Nikhef. 
    • Monitoring: some of the traffic was routed to a monitoring system. This enabled SNT to see who was posting what and, where necessary, to block inappropriate content. 
    • Fair play: anyone with an extremely high ping was artificially assigned a higher packet loss rate. The higher your ping, the greater the proportion of packets that ‘disappeared’ en route. This ensured that even participants with a standard home internet connection had a fair chance. 
    • Drawing: the pings that did get through were sent to the ‘painting server’, which displayed them on the canvas. 
    • Broadcast: a dedicated server streamed the canvas live to the rest of the world. 

    Statistics 

    As you can see from the screenshot below, the maximum parcel throughput peaked in the early hours of Sunday 11 October 2026: no fewer than 2.14 billion parcels per second. This graph shows the 8x400G ports that SURF installed in collaboration with Nikhef.

    Packet rate towards Nikhef/SNT canvas servers.

    Below is the resulting bit-rate:

    Bit rate towards Nikhef/SNT canvas servers.

    For incoming traffic, various routes were made available, including a number of internet exchanges, transit providers and, of course, the SURFnet Infinity production network:

    Result

    The result was a weekend full of pixel art, logos and, above all, a great deal of chaos. The best bits can be seen in the time-lapse, which shows the weekend at 200 times normal speed. Meanwhile, the new network proved that it can handle such enormous volumes of traffic without any problems. 

  • SURF network continues to innovate: intelligent, independent and flexible

    SURF network continues to innovate: intelligent, independent and flexible

    SURF network continues to innovate: intelligent, independent and flexible

    SURF is building a new generation of the SURF network. We are doing this with a new network architecture and the further development of the network automation stack into an intelligent network. The network is based on open standards and procurement of network equipment goes through an intermediary.

    With this further development of SURF’s network, we can respond more flexibly and faster to technological innovations and reduce dependence on suppliers. With the new network architecture and a smart, flexible replacement strategy, we are also working towards a more sustainable network.

    Digital highway in the Netherlands

    With the SURF network, Dutch education and research have their own autonomous network infrastructure. In line with our public values, we determine accessibility, freedom of choice and privacy on our network ourselves. Since 1988, based on these principles, we have been building this own digital highway, intended for independent and reliable research and knowledge sharing.

    The SURF network has several features that allow students, teachers and researchers to collaborate safely, reliably and at lightning speed worldwide.This is due to the high continuity and reliability of the SURF network, which simply must always be available.The SURF network also has high bandwidth for moving large amounts of data. Low latency ensures minimal travel time between sending and receiving data within the network. Moreover, all network services are integrated with each other, and we are working to ensure proper integration with other SURF services.

    Growth in network traffic

    Ever since SURF was founded, Internet traffic has been growing by around 29% annually. For research projects, this percentage is sometimes even higher. This growth is one of the reasons why network technology continues to evolve so rapidly and why SURF’s network has to be renewed every 5 to 8 years. Equipment is ageing rapidly and the demand for capacity continues to grow rapidly.

    New approach for the SURF network

    In 2024, we started the project to set up a new generation of the network after SURFnet8.In previous generations of the SURF network, from SURFnet1 to SURFnet8, the architecture was strongly determined by the choice of a network equipment supplier. With the latest generation of the network, SURF aims to be a supplier of independent architecture. Based on open standards, new technologies can be applied where they are needed and the network can continue to develop indefinitely. Hence this new generation is not called SURFnet9, but SURFnet Infinity.

    Workflow Orchestrator

    The first step for this new network architecture was already taken in 2018. With the Workflow Orchestrator developed by SURF based on open source software, various technology domains within the network can be centrally controlled. The Orchestrator ensures that tasks are performed in the right order and data is passed on correctly so that bottlenecks are prevented and reliability increases. This lays a solid foundation for the next step: further development towards an intelligent network in SURFnet Infinity.

    Purchase of network equipment via intermediary

    The purchase of network equipment is organised via an intermediary. In this way, SURF has greater freedom to make technology choices and can respond more flexibly and efficiently to the network needs of educational and research institutions when purchasing network equipment, without being tied to a single supplier.

    New SURFnet Infinity network architecture

    SURF’s new network architecture has established a clear hierarchy and separation of functions between the transport of network traffic, the provision of services to end users and connectivity to the Internet. This will facilitate capacity management, make the network more predictable, make it easier to detect and solve problems and provide more targeted security.It also moves a significant part of the network core from a commercial data centre to Nikhef’s data centre. The four core locations of SURF’s network are thus located within its own cooperative.

    First equipment purchased

    After an extensive selection process, SURF recently purchased the equipment to build the new core and border functionality. This directly realises the new architecture of the SURFnetwork.

    For the SURFnet Infinity network infrastructure, SURF has chosen the Juniper PTX series, which is even more powerful, compact and energy-efficient than the outdated MX series used by SURF. In addition, SURF is working with Salumanus as an independent supplier of transceivers based on OpenZR+ technology. This combination allows routers to connect directly to DWDM infrastructure, without the use of separate optical equipment. This not only reduces costs and energy consumption, but also increases network flexibility and future-proofing. Moreover, by choosing an independent supplier, the lifetime of the transceivers is not linked to that of the router hardware.

    Start of further development of SURFnet Infinity

    The further development of the SURF network starts with the renewal of the heart of our network: the core and border routers.Together with our management partner Quanza, we started preparations in the summer of 2025 .

    At 55 locations throughout the Netherlands, we will replace SURFnet8 equipment with new equipment.Most of the institutions connected to the SURF network will not be affected. For the work, we will contact our contact persons directly.This migration will run until early 2027.

    For international connections, in this phase we also replace the NetherLight equipment and optimise our Cross Border Fiber infrastructure. In the years after 2027, we plan to replace the optical equipment that facilitates high capacity between network components, as well as the access layer to which all institutions are connected.

    Want to know more about SURFnet Infinity?

    Follow this project on this page.
    We also post regular project updates on the network dashboard.
    Are you a network specialist?Then subscribe to the updates on the SURFnetwork Infinity project.

  • SURF and Nikhef complete test with Ciena’s WaveLogic technology showing multiple 1 Terabit paths for scientific research between Amsterdam and Geneva

    SURF and Nikhef complete test with Ciena’s WaveLogic technology showing multiple 1 Terabit paths for scientific research between Amsterdam and Geneva

    SURF, Nikhef and Ciena have successfully tested multiple 1 Tbps paths over fiber between SURF and Nikhef in Amsterdam and CERN in Geneva. The test, powered by Ciena’s WaveLogic 6 Extreme, shows that technological innovations can significantly improve the bandwidth of existing infrastructure.

    A similar development was seen earlier in connections over copper. Where initially only 2,4 kilobits per second was achievable over a classical phone connection, enhanced coding techniques brought this to 50 megabits (VDSL) per second over the same infrastructure. We are seeing the same trend in fiber technology.

    For SURF, this increase in capacity is of great importance. The scientists that SURF supports are able to measure ever more accurately, which leads to explosive growth in the amount of measurement data that SURF has to transport. An example of this is the Large Hadron Collider at CERN, which will be upgraded to much higher precision in the coming years (https://home.cern/science/accelerators/high-luminosity-lhc). 

    SURF and Nikhef together operate a  Tier-1 site for storage and processing of this LHC measurement data. For this purpose, a special 1648 km long fiber path has been established between Geneva and Amsterdam. This trajectory will need to transport much more data in the near future. However, this path also presents a technical challenge due to its long distance, age, fiber type, and the number of small spans that make up this trajectory.

    Vendors of optical equipment continuously come up with improvements for fiber connections like their colleagues did for copper lines in the past. To see what is possible with the most modern optical equipment, SURF and Nikhef therefore regularly conduct trials with these suppliers. In March 2025, we conducted a trial with Ciena over our existing Amsterdam-Geneva connection. In this trial we were able to achieve speeds exceeding 1 Terabit per second over this challenging fiber path. We thank Ciena for the opportunity to do these tests and for its support. These results give us sufficient insights to be sure that at the technical level we can transport the upcoming LHC-data explosion to SURF and Nikhef in Amsterdam.

  • SURF and Nokia Achieve 800Gbps Data Transmission Speed on a 1648 km Network Connection

    In collaboration with Nokia, Nikhef, and CERN, SURF successfully tested the data transmission speed between CERN in Geneva and Amsterdam. The test achieved an impressive speed of 800Gbps over the 1648-kilometre fibre-optic link. This confirms the network’s readiness to handle future large-scale data flows, such as those generated by CERN’s particle accelerator.

    Upgrade of the Large Hadron Collider
    The test was conducted in preparation for the planned upgrade of CERN’s Large Hadron Collider in 2029. This upgrade will enable more precise measurements, facilitating detailed research into the origins of the universe. The upgrade is expected to result in five to seven and a half times more research data than currently produced. All this data will need to be distributed to universities and research groups worldwide, including those in the Netherlands, via networks such as SURF’s.

    To successfully transport this immense volume of data, SURF has optimised its optical network. Nokia tested its latest-generation network cards to evaluate their performance over the SURF network under such demanding conditions. The successful test demonstrated that this section of SURF’s network between Amsterdam and Geneva can handle significantly higher capacities than previously anticipated.

    Increasing Demand for Bandwidth and Network Speed
    The upgrade of the Large Hadron Collider aligns with the global trend of data-intensive research requiring higher network speeds, increased bandwidth, and greater storage capacity. Research projects such as the Square Kilometre Array, the Einstein Telescope, the Low-Frequency Array, and the International Thermonuclear Experimental Reactor also generate massive data streams, demanding speeds beyond the current 100Gbps and 400Gbps standards.

    Insights and Follow-Up Tests
    Testing an operational network connection over long distances using real data from the Large Hadron Collider provides unique insights into data transport and storage at scale. These types of tests, regularly conducted by SURF in collaboration with various network partners, are essential for enhancing infrastructure to support data-intensive research.

    SURF continues to innovate by testing, optimising, and expanding its network, global network hubs, storage capacity, and computational power. The goal is to ensure that data-intensive research can be conducted effectively now and in the future.

  • SURF, Nikhef, and CERN are preparing for the High Luminosity LHC

    SURF facilitates the high-bandwidth connections between Amsterdam and Geneva. With the advent of the High Luminosity Large Hadron Collider at CERN, traffic is expected to increase fivefold in the coming years. This upgrade will enable more precise measurements, allowing for more detailed research into the origins of the universe. As a result, the amount of research data is projected to grow by five to seven and a half times compared to current levels. All this data must be distributed via various networks, including SURF’s network, to universities and research groups in the Netherlands and worldwide.

    To successfully handle this massive data volume, SURF is working with partners on the optical network. For the upcoming test, they are collaborating with supplier Nokia. Nokia has developed a new generation of network cards that allow for higher transmission rates between Amsterdam and Geneva. The test is scheduled for early next year, but preparations are already in full swing: