Monetizing gaming: running several servers in the decentralized network, hardware owner perspective

We are pleased to announce that Playkey is becoming increasingly popular in the Russian IT-community. Just recently we have published a translation of the article about stress testing cloud gaming services in poor connection conditions. And as we continue sharing translations, this one is about working with our decentralized network.

We wholeheartedly thank newspirit2019, a user of a Russian resource Habr, for the permission to translate and publish the experience he had with our decentralized network. This user has connected to Playkey network to rent his hardware out to gamers.

This is a translation of the original post written by Habr user newspirit2019, it is published with the author’s permission.
The original article can be found here: https://habr.com/ru/post/497522/

Just a few days ago I discovered a Habr article titled “Decentralized network as a GFN alternative” and decided I should share my experience with such network. I was one of the first to rent out my hardware for gaming. I am not a gamer, I just happen to be an owner of several rather powerful PCs, which are now connected to this network.

Let’s make it clear from the start, my servers are used to stream games to people who connect to a cloud gaming network. The article I mentioned previously lists three solutions: Playkey, DROVA and SONM. I opted for Playkey and I am describing my experience with them.

Technical side

I’ll give a brief description of the inner mechanics of such service. Cloud gaming service looks for owners of powerful PCs who would be willing to rent out their hardware for money. When a player connects to the service, it detects the closest to the player server. Then a connection is made and the chosen server is used for gaming. This way users get to enjoy their games with minimal input lag and the service generates profit and shares it with the server host.

How did I get involved in that?

I have about 25 years of experience working in the IT industry. Most of that time I have been running a small private enterprise which focuses on developing navigation systems. I do enjoy video games, although I am quite far from being a gamer. The company owns about 20 high-end machines and a large portion of these resources remains unused.

So, some time ago I started looking for a way to make these machines generate profit for the company. I discovered quite a number of both local and foreign services. Most of these offers implied mining, which was not even an option for me, as the industry was filled with all sorts of possible scams.

However, the idea of dedicating these machines to gaming was very different. I applied for the beta testing program and was accepted right away. And yet, I was not invited to actually participate until about 18 months later.

I was tempted by the fact that the service required just the hardware while installing, setting up and updating the software would be taken care of by them. And since my spare time was very limited, that was a huge advantage. Besides, it was possible to host several virtual machines on a single physical machine, I ended up doing just that.

Once the hardware was all set up and ready I decided to give the system a try and connected to the decentralized network as a player (I was just a few kilometres away, so I was connected to my own server). The difference between this and a game running in a cloud was crystal clear. It felt nearly identical to a game run on my own PC.

Hardware and Networks

I tested the network on various hardware configurations. The PCs in this test were workstations powered by Intel CPUs ranging from i3 up to i9, with RAM modules varying in both capacity and frequency. Storage was taken care of by both HDD and SSD units plugged in via SATA and NVME. As for the GPUs I went for Nvidia, using several GTX 10x0 and RTX 20x0.

As a result, I ended up having four server machines as the final solution for the beta. All of them powered by i9–9900 CPUs and equipped with 32 or 64 gigs of RAM. Each rig was configured to host three pretty powerful virtual machines. That is 12 virtual machines in total. Afterwards, I located hardware on a deep (1 meter wide) shelf. All of the cases were well ventilated, equipped with proper cooling solutions and dust filters.

As for the routers, their bandwidth varied from 100 Mbps to 10 Gbps.

Consequently, I learned that 100 Mbps bandwidth is not enough for the decentralized network, so an average household router would do no good here. Then again, these routers can be an issue even when it comes to regular content browsing. However, gigabit routers with two or even four cores CPUs were a perfect match for the task.

Server Load

I joined the beta test program before the Pandemic crisis started. Back then the machines would be in use for about 25–40% of the time. These numbers skyrocketed when the health situation worsened, though. These days with the majority of people stuck at home most of the machines see as much as 80% of active use time. As a result, all the testing, monitoring and maintenance have to be done in the morning hours to keep the machines up when they are needed the most.

The spike of interest meant that the amount of maintenance work has also spiked. Both my colleagues and I have to pay more attention to the systems but we are doing just fine nonetheless. There are occasional issues that require some fixes, but the overall experience is smooth.

I can easily access the information on use of the virtual machines from the administration panel. It gives data on the status of each virtual machine, resources used, games played and total time in each game. There is plenty of interesting data to explore.

Hardware Maintenance

Just as I have mentioned earlier, the experience is not flawless. The main issue is the absence of automatic system health monitor that would notify server owners of any issues. I certainly hope that this feature will be added soon. In the meantime, I have to manually monitor my control page and check hardware temperatures, network status and so on. Luckily, I have enough experience in IT, but this could be an issue for a less advanced server owner.

Fortunately, most of the issues were solved at early stages of testing. A guide or a FAQ list would definitely be helpful, but that is something that can be addressed later.

The Juiciest Part — Profits and Expenses

It goes without saying that this is nothing like SETI@home, and the main goal here is to make money. And the ideal solution for this would be a powerful host server running several VMs, as this becomes far more profitable to run with all the costs taken into account. There is also some time and knowledge required to get everything set up, but with enough time and motivation anything can be learned.

Power consumption in these scenarios is much lower compared to mining. I explored different opportunities to make some money, including crypto currencies, so the comparison is not unjust. Below you can find average power consumption per month:

• 1 server (i5 + 1070) — single virtual machine ~80 kW/month
• 1 server (i9 + 3*1070) — 3 virtual machines ~130 kW/month
• 1 server (i9 + 2*1070ti + 1080ti) — 3 virtual machines ~180 kW/month

In early testing stages the reward was just a formality, something in the neighborhood of $4–10 per VM.

Later the reward was increased up to $50 per month per virtual machine, assuming that the machine stays in the network all the time. This is a fixed sum, although the service promises to implement payments on a per minute basis. So, we would be looking at a total of $56 per month per VM. That is a reasonable sum, even with all the taxes, banking fees and ISP and electricity bills in mind.

Based on my estimations such hardware, if bought exclusively for decentralized gaming, has a three year payback period. Whereas average lifetime of PC components (including both wearing out with use and obsolescence) is about four years. That is why it is probably a good idea to join the project only if you already own the hardware. Another positive aspect is that the demand in the service keeps growing. And with the soon planned per minute pay it will likely become more profitable with a shorter hardware payback period.

My Thoughts about the Service and its Future

Decentralized network seem to be a good opportunity for gamers with powerful PCs to cover their gaming hardware expenses. They do not need cloud gaming, but since they already own a powerful rig, why not use it to cover its own cost. Besides, this is also a great opportunity for businesses with powerful hardware that is not used to its full capacity. These potential resources can easily be monetized with such networks, which is especially important in the ongoing crisis situation.

Decentralized gaming is a smart solution accessible to a wide range of users. Gamers get smooth cloud gaming experience with great picture quality and minimal latency as the servers are just a few kilometres away. Hardware owners profit off their hardware being rented out. The larger such networks get, the more efficient they become.

In the near future I see cloud and decentralized gaming coexist and complete each other. In the current situation, as cloud gaming services experience increased loads, decentralized networks seem to be the right solution. And even once the health crisis is over, the gaming industry will keep growing, so decentralized networks will grow and develop as well.