Elrond VPS guide
Thinking about where you will put your nodes on 2020?
It’s plenty of options to run elrond nodes, the goal is to fit well the operator profile and resources with the existing options to do it.
Let’s explore them and later you will see comparison tables between different providers.
Home running vs Datacenters
One of the strengths of Elrond Network is that it has very low barriers to have new node runners, it’s really the backbone of Blockchain culture, a democratic and decentralized system.
Home nodes is good option if you have some PC that has the minimum resources to run a node:
- Dual Core
- 4GB ram
- 200 GB disk space
- 100 Mbps
The most difficult part is internet connection, if you have fiber connection in your home with at least 100 Mbps you will be able to run well your node. But if you lose connection from time to time this can take down your node score.
A datacenter ensure us reliability in terms of no shutdowns on electric supply or internet connection.
If you have the machine/s in your home and prefer don’t have monthly fix costs, and can ensure with UPS energy for internet and server, go with home option.
One option very reliable and energy efficient are the Intel Nuc barebones. You can make cheap configuration for about 300 eur to run a couple of nodes in house.
VPS vs Dedicated Server
In datacenters you have 2 main options to rent resources:
- VPS: they are Virtual Machines that run above hardware to “simulate” a server.
- Dedicated Server: you rent directly one complet machine (the iron).
Last years we have see an important growth in VPS offer, the main reason is to improve the use of resources. Imagine that you are a datacenter owner, if you purchase one server for 3000 € and you rent it for example for 200€/month. The machine itself can be idle 80% of the time, for example if your customer puts a CRM on it and his employees only use the software from monday to friday on work hours. The same datacenter can put 4 VM on this server to ensure that the hardware is working more hours.
Usually what happens is that on a farm for VPS they have for example 100 GB of ram in total and of course they will sell above this, because they know that is very probable that not all the VM use max RAM in the same minute. Then you have datacenters that sell double ram that is available, others sell triple… and so on. Here depends of the strategy of the datacenter: cheap prices vs ensure reliable performance.
The good point with VPS is that is very hard to have your machine down, because in the layer of hardware some ram slot or disk may die, your VM balance to another one and you don’t notice hardware issues.
Sometimes you can experience hills and valleys on performance, for example if your VPS share hardware with a Hospital that on Friday execute an AI algorithm that put 100 VM up at full speed, maybe your nodes get affected and the 8 cores that you rent are working during these hours as they are 2 cores.
If you want more reliability in terms of performance you need Full Servers, because all the hardware is yours, and if you rent a 10 cores Xeon you have them always. On the other hand if some hardware component dies you can get your machine out of service some hours. But it really happens at very very low rate today.
Upstream bottleneck
There are plenty of applications to put on servers in the datacenters, but elrond nodes are very critical on internet traffic use.
Take carefully this in mind, in a datacenter there’s a capacity of internet connection. For example a datacenter with 100 Gbps, the customers can have nominal use of 30 Gbps as an example. But if in this datacenter say there’s plenty of webservers for Germany during some hour on the day the traffic on this datacenter goes up to 80 Gbps.
Now imagine that we put 300 nodes of elrond here (300 x 100mbps = 30Gbps), in this datacenter at some hours every day we don’t have enough internet capacity(80+30=110Gbps), and our nodes will become slow connection to the elrond blockchain.
If you are the datacenter owner you can take out elrond nodes because they affect his timelife customers.
The contabo case
I need to write some lines about this because in elrond project it become something notable. I hear that on Battle of Nodes we had near 1/3 of the nodes on contabo.
On VPS solutions contabo has prices that are about 60% cheaper than the competitors price.
As nodes on elrond are processing blocks and connecting to each other continously we can saturate one datacenter if we put plenty of us together.
If you are in love with contabo and don’t plan to go out, what I recommend to you is don’t force specs. Rent 2 cores more that you theorically need, for example rent an VPS with 8 cores to run 3 nodes (my setup on BoN).
And if you can, split your nodes. Say that you want 9 nodes on mainnet, you can put 3 on contabo and 6 on other datacenter. You still will have good average cost per node and the network will perform better.
Finally, contabo will not be there always, if elrond growths and we put 1000 nodes (10Gbps) there, contabo will begin to shutdown our accounts.
Amazon, Microsoft and Google Cloud Services
There’s plenty of good usecases of this services, like Accounting, Human Resources, Data Analisys, Research and a lot more. When you have plenty of hours with low or zero process and the some peaks or hours of full load. On this scenarios the fees scheme of this services went well.
Elrond Team for example has used AWS (Amazon virtual machines), on testnet to make try and errors. Here it works well because you are able to deploy 100 nodes and keep them running 48h, you don’t need to pay whole month of this resources.
But on mainnet nodes our continous use of CPU and bandwith will make AWS, Azure and Google Cloud not good solutions, I discard them for elrond nodes no doubt.
VPS comparison
On VPS the problem usally is the disk space, as VPS has additional software layer because it uses VMs, it performs better on SSD disk technology than traditional HDD disks. The existing offer of VPS is more frequent with SSD than HDD and usually it’s not enough for elrond nodes (200GB of space for each node).
Second difficult resource is internet connection, be sure that you have at least 100Mbps speed per node and unlimited traffic or a minimum of 10TB/node monthly.
The RAM is never a problem in the setups you will find and the CPU cores offered are good too.
Here you can find some options that I find, only those ones that have enough disk space, and unlimited traffic or 10TB/node. The nodes capacity that you will see is my recommendation, not the limit, because I always gave some margin to the setups to ensure reliability on peaks in elrond network.
Really the most difficult resource to find on VPS providers is disk, the regular setup comes with 30–50GB of disk, and is really difficult to see VPS with huge space to fit 8–10 nodes.
Full Server comparison
As there’s a minimum cost in terms of rack space in the datacenter, internet connection and so on. The cost of the server is not lineal, I mean you will see lot more cost/node if you try to rent a machine for 2 nodes than a powerfull machine for 10 nodes.
In the other hand, the tipical setup will have 1Gbps speed, make this go up to 2Gbps is really expensive (like 300eur/month). For this reason I will focus on machines with capacity between 4 nodes and 10 nodes.
About disk space, for big numbers of nodes like 8–10 in a machine, I recommend RAID with HDD traditional disks, as this setups are really cheap and will bring you 2TB that is the space for 10 nodes.
As we see on VPS, the RAM and cores are not a problem in general. I will focus always on cores, not threads capacity.
If you will deploy more than 5–6 nodes I recomend you to split your setup, to avoid risk. In this case is better to have 2 servers of 3 nodes each, in different datacenters instead of having all your nodes in one unique server. The cost per node will be very similar.
Diversity is the key
Each of us need to find our best solutions, but from the network perspective is really good to have good mix of solutions. Distribute the nodes from home computers, to VPS or Full servers and on a wide number of countries.
Remember that at any moment you can move your nodes from one solutions to another ones if you see poor performance or reliability.
Do you know other alternatives? share them with us on the Validators Telegram channel of elrond: https://t.me/ElrondValidators
