FutureBit Apollo III - Next Gen Home Mining

[FP91G]

Tell me, why would a small miner want to set up their own node?
I can do it on my own PC, but my connection quality will be worse than in a data center.

to privately bounce your own software off of your local copy of the full blockchain.

I understand this, but I'll only do it if I don't trust other solo pools and have enough hashrate to have a good chance of finding a block solo. I won't be able to achieve the same speed at home compared to a data center. I view mining as a commercial activity, and I want my efforts to maximize my profits.

[crypto_curious]

Testing shows that RAID 0 with SSD drives does not perceptively increase read/write speeds.  The creator of the APOLLO III Solo Node mentioned that he was interested in implementing the "mirror drive" feature sometime earlier in the development process.  I am still intrigued, and looking forward to seeing it.  It would increase local data security.

Hm... RAID 0 can significantly improve performance, although with NVMe drives the gains are often minimal because a single drive may already saturate the available bandwidth. So that's a fair point.

As for RAID 1, I don't see much value in it for a Bitcoin node. The blockchain is already replicated across thousands of nodes worldwide, so there's little benefit in mirroring it locally. For storing only the blockchain, RAID doesn't provide much practical advantage.

"It would increase local data security."

Only if the drive also stores your own important data. For a dedicated blockchain drive, RAID 1 simply doubles the SSD space and write wear for little real benefit. I'd rather keep a second NVMe as a spare and replace the primary drive when it eventually fails, instead of wearing out both drives at the same time. If they're identical models with similar usage, they may even fail around the same period, reducing perceived protection RAID 1 is supposed to provide.

[scryptr]

RE-DOWNLOADING THE BTC BLOCKCHAIN IS A PAIN
-----------------------------------------------------------

It is my data, I want to be able to spin up quickly.  And, RAID 0 can crash on a power outage.

        L8r...       scryptr

[philipma1957]

when it finally comes I plan to put in a 2tb nvme.2 ssd.

and I will clone it and store the clone. so if the first crashes i will have a clone with 2009-2026 on it.

[scryptr]

WAITING ON THE WORD
---------------------------

Rather than stretching the argument further, I'd rather hear from jstefanop as to his intentions on adding the mirror drive feature to the Solo Node OS.  Luckily, I purchased the pair of 2TB NVME drives prior to the price increases inflicted on the computing market by the current memory shortage.

The "clone and store" option also seems viable.  A corrupted drive could simply be replaced, reformatted, and be used to hold a fresh clone.

        Thanks for the replies...       --scryptr

[crypto_curious]

RE-DOWNLOADING THE BTC BLOCKCHAIN IS A PAIN
-----------------------------------------------------------

It is my data, I want to be able to spin up quickly.  And, RAID 0 can crash on a power outage.

        L8r...       scryptr

Whatever rocks your boat  
RAID0 is not any worse than plain single drive after a power outage - it will recover just fine. Journaling filesystems solved it decades ago.
 am not sure what jstefanop is using for his OS on Apollos, probably Ext4. Software mdraid RAID0 or RAID1 on Ext4 should be totally fine.
Personally, I use Btrfs RAID1 software mirror for my personal data. It has one advantage over mdraid - data checksumming.

when it finally comes I plan to put in a 2tb nvme.2 ssd.

and I will clone it and store the clone. so if the first crashes i will have a clone with 2009-2026 on it.

That's the way to go! Backup, not online mirror aging the second drive unnecessarily.

[philipma1957]

Quando Quando Quando


https://www.youtube.com/watch?v=jQWp1azgPXs

[jstefanop]

Testing shows that RAID 0 with SSD drives does not perceptively increase read/write speeds.  The creator of the APOLLO III Solo Node mentioned that he was interested in implementing the "mirror drive" feature sometime earlier in the development process.  I am still intrigued, and looking forward to seeing it.  It would increase local data security.

Hm... RAID 0 can significantly improve performance, although with NVMe drives the gains are often minimal because a single drive may already saturate the available bandwidth. So that's a fair point.

As for RAID 1, I don't see much value in it for a Bitcoin node. The blockchain is already replicated across thousands of nodes worldwide, so there's little benefit in mirroring it locally. For storing only the blockchain, RAID doesn't provide much practical advantage.

"It would increase local data security."

Only if the drive also stores your own important data. For a dedicated blockchain drive, RAID 1 simply doubles the SSD space and write wear for little real benefit. I'd rather keep a second NVMe as a spare and replace the primary drive when it eventually fails, instead of wearing out both drives at the same time. If they're identical models with similar usage, they may even fail around the same period, reducing perceived protection RAID 1 is supposed to provide.

It's a redundancy feature for people solo mining. We have clients using the solo node for 100s of petahash of hashrate...if one drive dies or gets corrupted raid setup is way more robust.

Will have the RAID feature complete for both solo node and Apollo 3 once we get all the pre-orders out in the next few weeks.

RAID 0 also has the benefit of nearly double random read bandwidth, with makes building blocks even faster. Almost all major pools use Muti-raid setups across 10s of thousands of dollars of infrastructure. The Solo node pretty much will replicate 90% of that for a fraction of the cost.

[crypto_curious]

Testing shows that RAID 0 with SSD drives does not perceptively increase read/write speeds.  The creator of the APOLLO III Solo Node mentioned that he was interested in implementing the "mirror drive" feature sometime earlier in the development process.  I am still intrigued, and looking forward to seeing it.  It would increase local data security.

Hm... RAID 0 can significantly improve performance, although with NVMe drives the gains are often minimal because a single drive may already saturate the available bandwidth. So that's a fair point.

As for RAID 1, I don't see much value in it for a Bitcoin node. The blockchain is already replicated across thousands of nodes worldwide, so there's little benefit in mirroring it locally. For storing only the blockchain, RAID doesn't provide much practical advantage.

"It would increase local data security."

Only if the drive also stores your own important data. For a dedicated blockchain drive, RAID 1 simply doubles the SSD space and write wear for little real benefit. I'd rather keep a second NVMe as a spare and replace the primary drive when it eventually fails, instead of wearing out both drives at the same time. If they're identical models with similar usage, they may even fail around the same period, reducing perceived protection RAID 1 is supposed to provide.

It's a redundancy feature for people solo mining. We have clients using the solo node for 100s of petahash of hashrate...if one drive dies or gets corrupted raid setup is way more robust.

Will have the RAID feature complete for both solo node and Apollo 3 once we get all the pre-orders out in the next few weeks.

RAID 0 also has the benefit of nearly double random read bandwidth, with makes building blocks even faster. Almost all major pools use Muti-raid setups across 10s of thousands of dollars of infrastructure. The Solo node pretty much will replicate 90% of that for a fraction of the cost.

I am sceptical to the idea of NVMe RAID on the Apollo because the Apollo is a home hobbyist device, not a platform for professional mining operations. That's why I've argued against it in my previous posts. Of course, I appreciate new features, but personally I don't think this is a feature suited to professional use. For hobbyists, it can be offered as an option, but the benefits are marginal.

Now, let's look at what a 100 PH/s operation actually looks like.

I checked the ASIC Miner Value website. Using one of the latest miners, the Bitmain Antminer S23 Hyd 3U (1.16 PH/s), you'd need around 86 units. That's well over half a million dollars' worth of mining hardware, consuming roughly 1 MW of power. We're talking about a serious operation with dedicated power infrastructure, water cooling, redundant internet, monitoring, physical security, and staff on site.

The minimum standard for serving the blockchain to that operation is not a US$1,000 black plastic brick sitting on the owner's desk over flaky Wi-Fi. It's a proper Dell, HPE, or similar enterprise server with plenty of x86-64 cores, ECC RAM, redundant storage, redundant power supplies, mounted in a rack. A server like that should never go offline except for scheduled maintenance. I have a home server running 24/7 on consumer-grade hardware, and it hasn't gone down due to a fault since I built it in 2018.

If someone is running a 100 PH/s mining operation and using an Apollo as their primary blockchain server, they're doing it wrong.

[jstefanop]

Testing shows that RAID 0 with SSD drives does not perceptively increase read/write speeds.  The creator of the APOLLO III Solo Node mentioned that he was interested in implementing the "mirror drive" feature sometime earlier in the development process.  I am still intrigued, and looking forward to seeing it.  It would increase local data security.

Hm... RAID 0 can significantly improve performance, although with NVMe drives the gains are often minimal because a single drive may already saturate the available bandwidth. So that's a fair point.

As for RAID 1, I don't see much value in it for a Bitcoin node. The blockchain is already replicated across thousands of nodes worldwide, so there's little benefit in mirroring it locally. For storing only the blockchain, RAID doesn't provide much practical advantage.

"It would increase local data security."

Only if the drive also stores your own important data. For a dedicated blockchain drive, RAID 1 simply doubles the SSD space and write wear for little real benefit. I'd rather keep a second NVMe as a spare and replace the primary drive when it eventually fails, instead of wearing out both drives at the same time. If they're identical models with similar usage, they may even fail around the same period, reducing perceived protection RAID 1 is supposed to provide.

It's a redundancy feature for people solo mining. We have clients using the solo node for 100s of petahash of hashrate...if one drive dies or gets corrupted raid setup is way more robust.

Will have the RAID feature complete for both solo node and Apollo 3 once we get all the pre-orders out in the next few weeks.

RAID 0 also has the benefit of nearly double random read bandwidth, with makes building blocks even faster. Almost all major pools use Muti-raid setups across 10s of thousands of dollars of infrastructure. The Solo node pretty much will replicate 90% of that for a fraction of the cost.

I am sceptical to the idea of NVMe RAID on the Apollo because the Apollo is a home hobbyist device, not a platform for professional mining operations. That's why I've argued against it in my previous posts. Of course, I appreciate new features, but personally I don't think this is a feature suited to professional use. For hobbyists, it can be offered as an option, but the benefits are marginal.

Now, let's look at what a 100 PH/s operation actually looks like.

I checked the ASIC Miner Value website. Using one of the latest miners, the Bitmain Antminer S23 Hyd 3U (1.16 PH/s), you'd need around 86 units. That's well over half a million dollars' worth of mining hardware, consuming roughly 1 MW of power. We're talking about a serious operation with dedicated power infrastructure, water cooling, redundant internet, monitoring, physical security, and staff on site.

The minimum standard for serving the blockchain to that operation is not a US$1,000 black plastic brick sitting on the owner's desk over flaky Wi-Fi. It's a proper Dell, HPE, or similar enterprise server with plenty of x86-64 cores, ECC RAM, redundant storage, redundant power supplies, mounted in a rack. A server like that should never go offline except for scheduled maintenance. I have a home server running 24/7 on consumer-grade hardware, and it hasn't gone down due to a fault since I built it in 2018.

If someone is running a 100 PH/s mining operation and using an Apollo as their primary blockchain server, they're doing it wrong.

Well thats your opinion, but unfortunately bad hardware and software in the space has made this perception a reality. We have optimized the solo node even beyond what a stock "server" can do in terms of raw performance per dollar. ARM embedded systems are designed for no downtime as much as enterprise servers, and that enterprise server will probably consume 1000 watts...8 cores, 16GB LDDR5 RAM, 2.5 GB ethernet and RAID NVMe gets you 99% of the way there on 20 watts. He's been up for 6 months straight found two blocks with zero downtime.

Perception that bitcoin mining requires  crazy infrastructure is wrong...now if you dont want to risk that 1% for a few grand thats entirely up to you, have also helped install solo operations on 10k worth of node hardware but still think thats overkill.

[crypto_curious]

Well thats your opinion, but unfortunately bad hardware and software in the space has made this perception a reality. We have optimized the solo node even beyond what a stock "server" can do in terms of raw performance per dollar. ARM embedded systems are designed for no downtime as much as enterprise servers, and that enterprise server will probably consume 1000 watts...8 cores, 16GB LDDR5 RAM, 2.5 GB ethernet and RAID NVMe gets you 99% of the way there on 20 watts. He's been up for 6 months straight found two blocks with zero downtime.

Perception that bitcoin mining requires  crazy infrastructure is wrong...now if you dont want to risk that 1% for a few grand thats entirely up to you, have also helped install solo operations on 10k worth of node hardware but still think thats overkill.

I appreciate the work you've put into making the Apollo such a capable platform.

That said, it's still a small ARM-based system without ECC memory or redundant power supplies. If it fails, the entire mining operation stops. You can reduce that risk by running multiple Apollo units... or by going to back to the strategy I outlined in my previous post.

At the end of the day, there's no single right solution. Different users have different priorities, so having multiple options is a good thing.

[heslo]

Tracking number received! Not long now 

[crypto_curious]

Nothing for me yet 

[heslo]

Any chance I can carry over my best share from the Apollo II to the III when I get it by backing up or copying across stats?

[Opazzno]

What model of ASIC chip is in the Apollo 3?

[philipma1957]

What model of ASIC chip is in the Apollo 3?

it is a future bit designed chipset.

not bitmain
not avalon
not whatminer made chips