Celestia Light Node Performance Analysis


I’m one of the lucky people who were selected by the Celestia team to run light node about a month ago. I will review the current status of my node and its performance over the last month. Before I start this review, I would like to state that although the name of the node I run is “light”, the tasks it bears are not light at all. Celestia’s light node need low system requirements and this feature created an opportunity for many people around the world to be able to set up nodes and make Celestia more decentralized. In this article, I will follow these steps:

  1. I’ll review the minimum system requirements to run light node and the sources that my server has.
  2. I’ll share my opinions about the effect of geographical distribution on decentralization and the location of my server.
  3. I’ll review the performance of my server in the lights of the data that I collected in the last 30 days (CPU, disk usage, reading and writing speed of disk, network speed, instant internet speed, etc.)
  4. I’ll analyze the real-time performance of my server with Netdata.
  5. I’ll review the data of my node on Tiascan.
  6. I’ll make an overall assessment.



The image below is about the Celestia Node system requirements. As can be seen, light node appealing with low system requirements to people all around the world who do not have high-speed internet and powerful hardware.

I rented Hetzner’s server called CPX21 and as you can see, its features are quite enough for the job. But you’ll know when you read the second title that even though it has powerful system features, I’m not satisfied with running node on its data center.

  • Memory: 4 GB RAM
  • CPU: 3AMD
  • Disk: 80 GB SSD
  • Bandwidth: 800 Mbit/s for download; —800 Mbit/s for Upload


My node is located in Hetzner’s data center in the Finnish capital right now. I think that if we want to create a truly decentralized product, we should not use such data centers, we should run our nodes in our homes or workplaces. But one of the biggest obstacles to this idea is that this type of work often demands high system requirements. Since Celestia’s light node does not require such a high system requirement, we can easily run a node at home even with the “Orange Pi 5”. My original plan was to run this node in the house where I live in Turkey and to help create a geographically decentralized network too, but due to some problems at customs, my Orange Pi 5 order did not arrive in time and I could not set up a node at home.

The image below is from Tiascan. We can see the location of the nodes running on the network in real-time. I unfortunately think that the results are not promising. Although the numbers on the map do not match the numbers on the right, they at least give an idea. Unfortunately, the majority of servers are hosted in data centers located in Germany and Finland. However, this is not surprising. Other networks also have this problem. I know a lot of people living in Turkey who are involved in Celestia’s award-winning Testnet process and almost none of them run their nodes in Turkey. There are some reasons for this. It is not easy for us to reach products such as “Orange Pi 5”, the companies providing server services in Turkey are not known enough, we do not have experience, the internet infrastructure of the country is not good enough to handle these matters etc…. I hope that these conditions can change in the future (I wish this kind of progress to be made all over the world) and networks may be truly decentralized structures.


Renting a server from a good source also has some advantages. These advantages are not irreplaceable for me, so I use them when they are offered for free. One of the things that Hetzner provides us for free is the last 30 days of hardware performances and it’s within a single click’s reach.

As we can see in the CPU graph, CPU usage has never risen above 50% in the 1-month period, and the average resource usage has been 20%. One of the reasons these numbers are low is because I have more CPU than the minimum requirements and Celestia demands very low hardware. When we see the disk data, we see that the first 15 days are not very active. It was only on the first day that the node started to work and naturally, disk usage was high. After April 19, transactions on the disk began to increase. I think something changed with a new update that came in those days. But it’s not a problem since even it’s updated, it still consumes very few resources. There is also no abnormal change in network traffic. In summary, everything has worked properly in the last 1 month. Finally, with the following command below, we can also look at the current disk usage. It doesn’t seem to be a problem there either.

df -h 


There are many different programs for Linux-based operating systems that you can track instant performance. I decided to use Netdata because of its open source and my previous experience with it. Since it uses a very small number of resources, you will not experience performance loss on your server and you will have the opportunity to visually examine a lot of data such as CPU, RAM usage, disk I/O, network traffic, etc. Because the data is updated 1 time/s, you’ll have a chance to take action quickly. And thanks to the alarm feature in the system, you can also receive a notification to your e-mail address in case of an emergency situation. Since there are many sources that shows you how to install Netdata, I will not include the installation phase here.

  • A – Node Performance Overview: When I look at the resource consumption in general, I don’t see any problems.
  • B – Network Usage: There does not seem to be a problem with the connection.
  • C – RAM Usage: The amount of RAM on my server is seems sufficient. There are still enough resources to use.


With Tiascan, you can see and review the data of all validators and nodes that are working in the Celestia Network in real time. If you want to see my node, you can click here. According to the data on this page, my node is working since 30 March 2023 and last restart is from ten days before. My uptime score is %96, although its relatively good score it’s not enough. Many node has more or less %100 uptime point. I’m not sure if this is a problem because i did all of the commands that has been sent by the team completely. I’m checking everything in a regular basis and i haven’t detect any issue to this day.


I became a part of Celestia’s prize award Testnet a month ago and I learned so much in this  1-month journey. Testnet process passed more smoothly than I expected. I almost haven’t faced any problem at all and when i did, it solved quickly by the team. The server I participated Testnet didn’t cause any problem in terms of performance and the main sources was sufficient.

Yazar Hakkında

Ata Tekin

Yorum Ekle

Yazar: Ata Tekin

Bu İnternet Sitesi Hakkında

Değerli okuyucu, paylaştığım her şeyde kendimden bir parçayı sana veriyorum. Bu parçaların hayatında fark yaratması dileğiyle...