[ANN] Spondoolies-Tech - carrier grade, data center ready mining rigs

[Soros Shorts]

if you have only one unit, put it there .. Host everything you have above that one unit with jtoomim.

One unit has 2 power supplies, drawing 7A each from the wall. That is 14A maxed out .. but for sustained load, this would represent 80% of the circuit size (i.e. the circuit should be at least 17.5 Amps).

So .. as the breakers go .. a 20 Amps circuit will host you one unit. A 30 Amps circuit will host you 1 1/2 units

If you get all three phases independently at the rack level, triple that. I guess the sales guy was telling you what you paid for, not what you need.

Actually, it's trickier than that. You don't get 208V as an independent phase-to-neutral voltage. You have to wire up to two phases to get that (hot-to-hot). A three phase 208V PDU has 1/3 of the sockets wired up between phase A and phase B, 1/3 wired between B and C, and 1/3 wired between C and A. This means if you put 7 amps on one plug (AB), that's 7 amps on phase A and 7 amps on phase B. If you put 3 plugs on AB, that would be 21 amps on A and 21 amps on B. Then you couldn't plug anything into phase C. If instead you put one on AB, one on BC, and one on CA, then you get 14 amps on each phase. You could then stick another plug on AB, and you'd get 21 amps on A and B, but only 14 amps on C. That's four plugs total, or two SP30s. Any additional plugs you put on BC or CA would bring B or A up to 28 amps, which exceeds the 80% limit for safe utilization. If you found that you could get away with that, or if you had the ability to boost voltage to 230 or 240V (or underclock your SP30s), you might be able to add one plug on BC and another on CA, putting all three phases at around 28A, with six plugs and three SP30s.

This is incorrect, see this article:

http://www.raritan.com/blog/detail/how-to-calculate-current-on-a-3-phase-208v-rack-pdu-power-strip

Lol. The article makes it sound like the topic is rocket science just to make you download and run their tool. It is actually relatively easy to compute by hand. If you have 7 Amps in AB and 7 Amps in BC the combined Phase B current is 12.1 Amps.

Code:

AB current: 7 Amps
BC current: 7 Amps
Phase B current = sqrt((7 + 7*sin(120))**2 + (7*cos(120))**2)
                = sqrt(110.25 + 36.24)
                = 12.1 Amps

[1714493651]

1714493651

[1714493651]

1714493651

[1714493651]

1714493651

[Biodom]

Spondoolies: a question...
Is there a hardware ID or other identifier (other than a order number) that would identify a machine.
Is this information being transmitted to the owner upon order fulfillment?
The reason-in a hosting environment, with many machines coming and going, mistakes could be made.
How can we ensure that a particular miner is mine and not someone else's?
I know that this is not your problem, but S3 all look the same and if there was such an ID, I don't know where to look (in case of S3)

The machines has serial ID.

that is visible in FW/SW?

And on the miner physically.

I am sure that you have this covered, but assuming that buyer has no physical contact with the machine(s), some other ID process has to be in place.
For example, buyer gets a shipping email with machine's serial number, then logs in into "his" machine at the hosting site, checks the Serial number in the FW/SW and that should be the same as in his/her email-it is all good. You would think that people cannot make such mistakes (misplace, mislabel, etc), but as a matter of fact they do, especially since there is nothing to distinguish one machine from another (color, etc.). I would actually suggest for hosting sites with hundreds of machines some form of barcoding with double label-one goes on the machine, another into customers file and they have to match. There are barcoding printers that can do a small double label.

I will have this issue too when my machine will be delivered to the hosting provider; It is important to be sure that we are on the right machine. Confusion may happened; and some machines have issues. Tracking them is important. And we can expect too, and I hope so, this is common with electronic manufacturing, the production will be more mature on September, and we can expect to see a slight and positive differences on the delivered machines (SMD resistor values adjusted and so on, better decoupling..).

Is the number sent to the owner when the delivery begins ?
Is there an eeprom on the machine that can be checked once logged ?

I would like to have this answered too since it was not explained yet, except a small blurb about serial number physically on machine, which is meaningless for the remote owner. ID of the machine once logged in that matches the serial number send by email to owner (not hosting provider)-is it there and in what fashion?

[wh00per]

I would like to have this answered too since it was not explained yet, except a small blurb about serial number physically on machine, which is meaningless for the remote owner. ID of the machine once logged in that matches the serial number send by email to owner (not hosting provider)-is it there and in what fashion?

SEE https://github.com/Spondoolies-Tech/minepeon/blob/Spond0.1/etc/cron.d/hourly/pandp_register.sh


board_id(){
   EEPROM_DEVICE=/sys/bus/i2c/devices/0-0050/eeprom
   board_id=`dd bs=12 skip=7 count=1 if=$EEPROM_DEVICE 2>/dev/null`
}

The ID is part of the registration script. Keep the unit registered, and you can track it after.

[Biodom]

I would like to have this answered too since it was not explained yet, except a small blurb about serial number physically on machine, which is meaningless for the remote owner. ID of the machine once logged in that matches the serial number send by email to owner (not hosting provider)-is it there and in what fashion?

SEE https://github.com/Spondoolies-Tech/minepeon/blob/Spond0.1/etc/cron.d/hourly/pandp_register.sh


board_id(){
   EEPROM_DEVICE=/sys/bus/i2c/devices/0-0050/eeprom
   board_id=`dd bs=12 skip=7 count=1 if=$EEPROM_DEVICE 2>/dev/null`
}

The ID is part of the registration script. Keep the unit registered, and you can track it after.

That is making this more, not less confusing for me. Basically, you are saying that during installation (registration) you can call it whatever-this makes it worse than I thought, actually.
Here is analogy from another field. Say, i am a patient who wants genetic analysis done for me. It is important to get MY analysis, not some other fellow.
Yet, i give my DNA sample to the lab and have no control of it afterwards. I better be sure that lab does not mix patient's samples and is accurate in analysis.
back to mining-it is of first importance for remote owner to have a firmware (not programmable) machine ID. Is it there or not?
EDit: saw your post, so this code

Code:

EEPROM_DEVICE=/sys/bus/i2c/devices/0-0050/eeprom

is where device ID is and it is in FW, not SW, correct?

[wh00per]

No, what I'm saying is that one can get the ID from EEPROM_DEVICE=/sys/bus/i2c/devices/0-0050/eeprom, no matter what the machine is registered with.

When enabled, the machine reports back to SP-T hourly. That way, they can monitor what machines are used, what firmware they have, and notify the users for firmware updates. Keeping the script alive, SP-T will have a history on your machine very hard to fake. Once the code is changed, (as per my understanding) you lose the warranty. Why would a normal customer change it and lose all the benefits coming with SP-T support? How would I change my registration ID to match yours? Definitely if you keep your registration script working, SP-T will have your ID logged, together with all other registration data. They can definitely compare changes between the actual status and the recorded data while the unit was in the burn-test, first powered up .. and so on.

The EEPROM (Board) ID is not SW, nor FW. It's a hardware ID and does not depend on the firmware or the software installed on your machine.

[Guy Corem]

Currently, we don't email the serial ID before shipping.
It's very hard, if not impossible to implement.
We prepare the shipping documents in advance.
Flextronics select one tested machine randomly and attached the shipping documents.

Biodom, if it's important to you to know the serial, you can ask the DC owner to email it upon receiving the machine.

We'll try to find ways to improve the procedure, but it will take time.

Guy

[kendog77]

Currently, we don't email the serial ID before shipping.
It's very hard, if not impossible to implement.
We prepare the shipping documents in advance.
Flextronics select one tested machine randomly and attached the shipping documents.

Biodom, if it's important to you to know the serial, you can ask the DC owner to email it upon receiving the machine.

We'll try to find ways to improve the procedure, but it will take time.

Guy

This is really only an issue if you don't trust your hosting provider.

If a hosting provider received a defective SP30 that they personally own, it would be very easy for them to swap the defective unit with a good customer unit, and the customer would have no way to know this occurred.

[wh00per]

Again, as I understand it, SP-T will  service all machines they built as per the warranty terms. For them if one is broken, it must be fixed, and it does not matter who owns it ..

I agree with Guy hare .. for the time being, just have the DC communicate it, and Zvi can update the MinePeon stats page to include it on the screen (below the MAC address) in the next software release. At least, when first powered up, you know what machine #ID you own. It still does not take out of the picture the "machine swap" before the first power-up in the end-user data-center, but at least you can track everything afterwards.

[Guy Corem]

I guess big orders have priority over "meaningless" small orders

No. Wissam didn't wait and contacted us after receiving the newsletter with the spec change.
He was offered the same options everyone else will be offered.
Some others also contacts us and we usually find a common ground.
Next week we'll contact all the rest of the customers.
We're a very small team, please be patient and let us contact you.
We do hire more support staff.

Guy

[Guy Corem]

Again, as I understand it, SP-T will  service all machines they built as per the warranty terms. For them if one is broken, it must be fixed, and it does not matter who owns it ..

I agree with Guy hare .. for the time being, just have the DC communicate it, and Zvi can update the MinePeon stats page to include it on the screen (below the MAC address) in the next software release. At least, when first powered up, you know what machine #ID you own. It still does not take out of the picture the "machine swap" before the first power-up in the end-user data-center, but at least you can track everything afterwards.

I admit that the method isn't bullet proof and needs to be improved. We'll look into it.
Of course we'll service any machine under warranty terms.
I've asked Zvisha to add the EEPROM serial #ID to UI

We're working on a list of DCs we'll recommend after proper due diligence by us.

Guy

Edit:
Note that the default hostname contains the EEPROM serial #. It begins with FL14:
http://storage.googleapis.com/spond_public/images/SP-30%20Hashrate.png

[Guy Corem]

Seems that spondoolies's side effect make appears some new business model !

For the good of the Bitcoin ecosystem: http://www.coindesk.com/gold-rush-arms-race-bitcoin-mining-north/

[Biodom]

Currently, we don't email the serial ID before shipping.
It's very hard, if not impossible to implement.
We prepare the shipping documents in advance.
Flextronics select one tested machine randomly and attached the shipping documents.

Biodom, if it's important to you to know the serial, you can ask the DC owner to email it upon receiving the machine.

We'll try to find ways to improve the procedure, but it will take time.

Guy

This is really only an issue if you don't trust your hosting provider.

If a hosting provider received a defective SP30 that they personally own, it would be very easy for them to swap the defective unit with a good customer unit, and the customer would have no way to know this occurred.

It is a shocker if true, and goes somewhat against the claim that it is fully data center ready, in my opinion.
Bitcoin teaches us trust-less transactions. I don't want to find by trial and error who I should trust, although I always trust first, only to be discouraged sometimes afterwards.

In my opinion, it is easy to remedy. Just make sure that the initial host name (which contains eeprom number as Guy edited) can only be changed by the owner (using perl, for example) and send each owner email with perl login/unique passwd (no big deal using random passwd generator), otherwise it WILL stay FOREVER on default hostname that includes eeprom. If owner will try to change the hostname and it would not work, then it would mean that it is accidentally not his/her machine. In addition, DC owner would be asked to send info about what is customers equipment serial number BEFORE powering it up and connecting it to the net. I think that this is important from the point of ownership if you want to sell your machine. Obviously, you cannot sell what is not yours.

[Biodom]

Seems that spondoolies's side effect make appears some new business model !

For the good of the Bitcoin ecosystem: http://www.coindesk.com/gold-rush-arms-race-bitcoin-mining-north/

So, SPT wants to be MBP franchisee? I hope, not, but the article puts two together.

Corem continued by saying that the market seems to be moving in favor of centralization, but argued that other players in the space, including Spondoolies, are actively moving to keep the network as decentralized as possible. This can be seen in efforts like MegaBigPower’s franchisee program, which begun explicitly as a means of expanding the network’s capacity in a decentralized manner.

[Guy Corem]

No method is really bullet-proof. The eeprom serial is saved in the management board.
In theory, the DC owner can replace ASICs boards and create better and worst machines.

I think that gmaxwell first suggested to burn the owner public key in the ASIC OTP.
(Last time I saw it here: https://en.bitcoin.it/wiki/Mining_manufacturer_tech_guidelines)

But I won't believe it will happen. Too complicated to implement.

Guy

[Guy Corem]

Seems that spondoolies's side effect make appears some new business model !

For the good of the Bitcoin ecosystem: http://www.coindesk.com/gold-rush-arms-race-bitcoin-mining-north/

So, SPT wants to be MBP franchisee? I hope, not, but the article puts two together.

Corem continued by saying that the market seems to be moving in favor of centralization, but argued that other players in the space, including Spondoolies, are actively moving to keep the network as decentralized as possible. This can be seen in efforts like MegaBigPower’s franchisee program, which begun explicitly as a means of expanding the network’s capacity in a decentralized manner.

Nope.

We'll sell ASICs and miners to anyone interested to buy them though.
It was mentioned few pages ago that MBP has their own ASIC effort.

Guy

[wh00per]

Definitely, if MBP wants SP-T miners to populate their data-centers in franchise I don't see why not.

Centralization and decentralization is a fun game . every 5-6 years there is a trend of one flavor or the other .. in an alternating manner.  And that can be seen across all industries. What we get out of it, eventually, is a central structure at the top with very mobile and empowered legs allowing the benefits and drawbacks of both. The extremes are being seen as the main changers over the short periods of time, until the environment adapts. There, one can profit or get hurt .. however, it pays to pick only the fruits of both sides.

[jtoomim]

Actually, it's trickier than that. You don't get 208V as an independent phase-to-neutral voltage. ... If instead you put one (7 amps) on AB, one on BC, and one on CA, then you get 14 amps on each phase. ...

This is incorrect, see this article:

http://www.raritan.com/blog/detail/how-to-calculate-current-on-a-3-phase-208v-rack-pdu-power-strip

Oh, thanks for the correction. I don't actually use this type of PDU, thus the mistake.

[jtoomim]

It is a shocker if true, and goes somewhat against the claim that it is fully data center ready, in my opinion.

In my opinion, it is easy to remedy. Just make sure that the initial host name (which contains eeprom number as Guy edited) can only be changed by the owner (using perl, for example) ...

All that needs to be changed is that the customer needs to be informed what their serial number is. You can always check the serial number, regardless of whether the hostname is currently the serial number or not.

I believe there is a sticker on the machine itself with that serial number/hostname in the back of the SP30 near the PSU. I'll check when I get to our datacenter later today.

If any of our customers are concerned about this issue, I am willing to video chat with them while unboxing their machines to give them a chance to record their serial number before we have a chance to swap it out. If you're less paranoid and trust us somewhat (which you probably do, if you're willing to send us miners in the first place), I can just write down and email you your FL serial numbers as I unbox them.

[johnray]

Currently, we don't email the serial ID before shipping.
It's very hard, if not impossible to implement.
We prepare the shipping documents in advance.
Flextronics select one tested machine randomly and attached the shipping documents.

Biodom, if it's important to you to know the serial, you can ask the DC owner to email it upon receiving the machine.

We'll try to find ways to improve the procedure, but it will take time.

Guy

This is really only an issue if you don't trust your hosting provider.

If a hosting provider received a defective SP30 that they personally own, it would be very easy for them to swap the defective unit with a good customer unit, and the customer would have no way to know this occurred.

It is a shocker if true, and goes somewhat against the claim that it is fully data center ready, in my opinion.
Bitcoin teaches us trust-less transactions. I don't want to find by trial and error who I should trust, although I always trust first, only to be discouraged sometimes afterwards.

In my opinion, it is easy to remedy. Just make sure that the initial host name (which contains eeprom number as Guy edited) can only be changed by the owner (using perl, for example) and send each owner email with perl login/unique passwd (no big deal using random passwd generator), otherwise it WILL stay FOREVER on default hostname that includes eeprom. If owner will try to change the hostname and it would not work, then it would mean that it is accidentally not his/her machine. In addition, DC owner would be asked to send info about what is customers equipment serial number BEFORE powering it up and connecting it to the net. I think that this is important from the point of ownership if you want to sell your machine. Obviously, you cannot sell what is not yours.

You are right, we have trust, but checking again on our side, is always better. If we are able to deduce the production week using the eeprom content is good to start with; Anyway having the serial number case/eeprom sent once the shipment begins is better.

[Biodom]

Currently, we don't email the serial ID before shipping.
It's very hard, if not impossible to implement.

At the risk of sounding 'paranoid', i don't see why it is difficult to sent such email (containing serial #) not before, but WITH order shipment notification.
A person who packs it could also record the serial # in the customers db, then customer service would email as part of shipment notification.
I get the point that DC can do it for you, but it is suboptimal, as it is not an independent confirmation with all due respect.
Having videochats with some customers would be really taxing in time, believe me, nobody would really want to do it if a simpler solution presents itself.