BetaNet Node Final Hardware Requirements are here!

Using the answers given by the 101 responders to the BetaNet Questionnaire which closed on Tuesday, March 27th, 2020 and after extensive testing in the NodeLab, the xx network Node Team has fine tuned the BetaNet node and gateway specifications as follows:

Node:

CPU: 8 core modern CPU with Hyperthreading such as AMD Ryzen 7 2700x

GPU: Nvidia Turing or Tesla GPU, Nvidia GeForce RTX 2070 or greater

RAM: 16GB DDR4 or more (an upgrade path to 32GB is recommended)

Storage: 1TB Professional High Speed NVME (PCI) SSD (Example: Samsung 970 PRO SSD 1TB - M.2 NVMe)

Bandwidth: 100mbit up / 100mbit down with Static IP

Other configurations may also work and will be approved on a case by case basis. The software does support full CPU nodes, but higher core counts are required. As the software becomes more mature and more power is extracted from the GPU, it is likely the hardware requirements for such nodes will increase.

We are actively testing setups with multiple high end (5+) Nvidia Pascal GPUs and expect to report in the next few weeks if such setups can meet requirements. That being said, it is likely that support will be dropped from such configurations as the platform matures.

Gateway:

Gateways are low powered machines which act as the public face for a node. Every node must have a gateway. Most gateways are expected to run in the cloud, but it is possible to run them as physical devices. They offer protection to the node if they are on a seperate IP address and network connection.

CPU: modern dual core (Example: AMD Ryzen 3 2200g)

GPU: None

RAM: 2GB or more

Storage: 250GB (database instance)

Bandwidth: 100mbit up / 100mbit down (Must be separate from node bandwidth) with Static IP

On the BetaNet forum, you can find the exact specifications we have been testing to better inform your decisions. Although, it is important to note that our test nodes do not have reliable SSDs.

These specifications may change over the life of the BetaNet, but we will strive to keep them as consistent as possible. Please note these specifications are highly likely to change going into the MainNet.

Node Registration

On Friday, the xx network will begin registering the first 150 nodes in the BetaNet! As discussed in our April 16th blog, The Road to BetaNet, we are partnering with Tokensoft on “registration” to meet the Cayman, and U.S. regulatory requirements to protect nodes and the network from legal issues and to allow as many nodes as possible to receive xx coin compensation.

Starting on May 1st at 24:00 GMT we will begin registration based on your color group (previously selected in May 2019). The Blue Group will be invited to register first at 24:00 GMT May 1st. The White group will be invited to fill any open slots on May 6th at 24:00 GMT and finally registration will open to the Teal group on May 11th at 24:00 GMT.

On July 1st, the xx network BetaNet will launch with the first 150 nodes, and by September will mature to invite all 600 nodes, and beyond!

3 Likes

Fantastic. Going shopping now! Can’t wait to register on Friday. Thanks for all your hard work, team.
Oh, as an aside, am I to presume in both instances we are using some Linux derivative as the OS?

Yes, we will be using linux

Amazing! Finally a timeline! Can’t wait to start!

Thanks for the update! One question: it’s not clear to me who will be operating the gateway machines. Will each betanode runner have to maintain both the node and its associated gateway?

Yes, node operators maintain both a node and a gateway.

The gateway masks the node from the internet at large and is designed to be set up in the cloud with a load balancer to take the brunt of any DDOS attack.

That being said, in BetaNet we expect some to run them on the same machines as their nodes proper given the reduced danger and importance of a BetaNet. Doing so would require a 200mbit connection.

1 Like

Is it possible to use a CPU 8 core with Hyperthreading AMD Ryzen 7 1700? All other parameters are suitable.

I think the specs should be a little more precise. There are tremendous differences in performance in 8 core CPUs, you can get Xeon servers with 32 cores that perform worse than a 4 core Ryzen CPU. It’d be probably useful to know the requested CPU Mark value (for GPU and CPU-only nodes) so we can compare it here: www .cpubenchmark .net /cpu_list.php

Similarly for the storage, what are exactly the performance requirements? How many R/W IOPS? What transfer rate? Is it a sequential or random access pattern? There could be RAID-0 SATA SSD configurations that perform better than a single NVME disk or that would be enough for the purpose.

Finally, will it be feasable to change the IP address of the node/gateway during the betanet, eg. if the machine is moved between cloud and in house or viceversa?

3 Likes

I think the specs should be a little more precise.

Your point is well taken. I will talk with my team about providing numbers in terms of benchmarks.

Similarly for the storage, what are exactly the performance requirements

This is an interesting question. The component which really uses this isn’t complete yet and will probably not be ready at launch. I will try to get back with an answer about this as well.

Finally, will it be feasible to change the IP address of the node/gateway during the BetaNet

The platform fundamentally allows for this, although the update process has not been fully implemented yet. We may finish it before launch or shortly after.

Maybe, let me get back to you

1 Like

@alexdupre

What I’ll be doing on my node is an easy way to overshoot storage issues. Just get 2 500GB SSDs and set them to RAID-0. That doubles the throughput by doubling the number of I/O lanes, and the cost is within a few bucks of a single larger drive. So, low-cost 2x throughput = easy sauce.

1 Like

I’ll be running on AWS/Google Cloud. Which instance seems most suitable?
Is it possible to run it on an instance without GPU?

I’m ready but I think the equipment is too expensive to achieve a decentralization of the network, not everyone has $1500 equipment to run a node. IT’S NOT A GOOD IDEA AND IT WILL NEVER BE DECENTRALIZED

The incentives though (7000 xx coin per month) mean that would be paid off in a month with the sale price of 0.35.

7000 xx-coin per month?? Yeah

I’ve probably read the agreement completely wrong… let’s wait for official word.

Yes… i think… ask now telegram group

It would be very very much appreciated if some of you, more expert guys, could share some suggested specs with the rest of us, beyond the general ones indicated by the xx team. For instance, I miss some guidelines on suggested motherboards, SSDs, power supplies, (silent) fans… Many thanks!

That is the correct amount

I have several questions regarding these requirements. It seems you’re aiming a little lower than the ultimate best computer you can buy.

8 core CPU
Ryzen 5 3600 outperfoms 2700 in benchmarks, is cheaper but has minus 2 cores. Why do you insist on the 8 cores?

RTX 2070 or higher?
That’s only one tier away from the best ever for consumer grade GPUs, so what makes this card the chosen one?
Also there are variants of this card with slight changes to core and memory. Can you elaborate on that?
What’s the load you expect to put the card on. Should I go with one with bigger heatsinks and 4 fans since it’s going to be always close to 100% or a small factor/cheaper brand will make the cut?
Do you expect to have to upgrade the gpu in the next year, if so can I just add a second 2070? If so I’ll have to buy a motherboard with that in mind.

1TB NVME SSD professional grade.
Most users have a small super fast NVME for boot drive and files go on a not so fast SSD even or a fast HDD. Isn’t that enough for you, run the OS and your client from the 128GB NVME and the keep the database separately?
Also, why NVME/ultra fast? I can use RAID0 like someone already said here before and get even faster read/write speed.

Static IP
I don’t know on easy and affordable it is around the globe but if that’s anything like here you can just quit trying to host a node.
I called my ISP 2 years ago to do that on my office and the reply was “that’s too difficult and expensive for you to consider”. I would have to upgrade my connection (which is 500/100, practically the best you can have around here) to enterprise (which is slower and a lot more expensive) with direct connection their servers.
My IP rarely changes however, I upgraded my plan for 500/100 in December and have the same IP ever since, but I can’t guarantee they won’t change it.
I use a free dynamic DNS instead - such has noip . com - that works like a charm.

Gateway
So far I had no idea a separate computer and location had to be necessary. I almost feel tricked into this.
My understanding, and since you require a separate IP/location is that you intend people to use cloud services like a VPS to do so.
Does it have to be located in the same country has the node? Many countries won’t have that opportunity or will be too expensive to do so.
You require 250GB (which is a lot for VPS), if that’s the database size, why the node requires 1TB?
If there’s downtime on the gateway, can the node automatically connect to a backup gateway?
Again, can the static IP be solved with dynamic DNS?
But most important, why does your system doesn’t contemplate this and implements the masking on the node itself?

This is the beta, no one knows what’s the future holding and although I wish for the best and believe in this project, I don’t want to buy an overpriced computer that I’m not gonna have other use, but also I don’t want to cheap out on the specs I’m inquiring and have to upgrade the node just a few months down the road, and I’m getting a 700€ difference between my “low cost” approach and your requirements. Just the small nvme+hdd vs 970pro is a 250€ difference.

In my case all my mining equipment is not suitable to donate parts. ASICS have nothing to take from, mining rigs all have weak cpus and I only AMD gpus so I’m getting everything new and I always want the best bang for the buck.

I’m sorry for the long post, but there are too many things regarding these requirements that I need to understand.

2 Likes