SgtSpike (OP)
Legendary
Offline
Activity: 1400
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 03:25:34 AM |
|
I am considering partnering with someone who has funds, renting some office space, retrofitting it with whatever electrical and cooling is necessary for a small datacenter, and renting out server "slots", for people who can no longer host their rigs in their apartments/houses/basements.
The service would only use a basic internet service with connections only allowed for VNC/SSH and the bitcoin mining applications themselves. No webhosting, etc allowed. Mining rigs ONLY.
The service would track electricity usage, and would charge a base rate per rig + actual electric usage each month. If mining became somewhat unprofitable for a week or two, you could simply turn it off, and only be charged the base rate for that period.
You wouldn't have remote control of the power directly, so in the even of a hard lockup or shut down, we would provide a "request power on" button in a web interface. Your rig would be powered back on within 24 hours. For this reason, it is imperative that you already have a stable configuration prior to shipping to our facility.
All management of your mining rig(s) would be done by you. We wouldn't touch it. You would be responsible for remotely accessing your server, once we unpackage it and start it up, to get the mining running. You would also be responsible for your own software security. Given that many machines would be on the same network, you should set up a software firewall appropriately to isolate your rig(s) from the other rigs.
A security firm would be hired for installation and monitoring of a security system intended to protect the building from theft. Insurance would also be purchased to cover the contents of the building in the event of damages.
Rates would be something along the lines of: $20-$30 per rig per month $0.10-$0.15 per kwh per month
Would you use such a service at such rates? If so, how many rigs would you rent space for? Vote in the poll!
Also, do you see any potential problems that I have not addressed? Any potential security or performance issues?
|
|
|
|
Syke
Legendary
Offline
Activity: 3878
Merit: 1193
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 04:08:57 AM |
|
I am considering partnering with someone who has funds, renting some office space, retrofitting it with whatever electrical and cooling is necessary for a small datacenter, and renting out server "slots", for people who can no longer host their rigs in their apartments/houses/basements.
Is a "slot" a 4U spot? The service would track electricity usage, and would charge a base rate per rig + actual electric usage each month. If mining became somewhat unprofitable for a week or two, you could simply turn it off, and only be charged the base rate for that period.
What amperage circuits would be provided? Does each slot get a 20 amp circuit? What is the temperature control of the facility? You wouldn't have remote control of the power directly, so in the even of a hard lockup or shut down, we would provide a "request power on" button in a web interface. Your rig would be powered back on within 24 hours. For this reason, it is imperative that you already have a stable configuration prior to shipping to our facility.
But I could install my own remote-power device, correct? All management of your mining rig(s) would be done by you. We wouldn't touch it. You would be responsible for remotely accessing your server, once we unpackage it and start it up, to get the mining running. You would also be responsible for your own software security. Given that many machines would be on the same network, you should set up a software firewall appropriately to isolate your rig(s) from the other rigs.
Will you install video cards? Rigs shipped with heavy video cards installed are likely to be damaged by the shipping companies. You really should enforce separate rig networks, otherwise customers could interfere with each others, like with incorrect networking settings, either through mistakes or on purpose. Also, do you see any potential problems that I have not addressed? Any potential security or performance issues?
What is the length of contract for a slot? Do you guarantee uptime (network, electricity, etc) of the facility?
|
Buy & Hold
|
|
|
SgtSpike (OP)
Legendary
Offline
Activity: 1400
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 03:29:45 PM |
|
I am considering partnering with someone who has funds, renting some office space, retrofitting it with whatever electrical and cooling is necessary for a small datacenter, and renting out server "slots", for people who can no longer host their rigs in their apartments/houses/basements.
Is a "slot" a 4U spot? The service would track electricity usage, and would charge a base rate per rig + actual electric usage each month. If mining became somewhat unprofitable for a week or two, you could simply turn it off, and only be charged the base rate for that period.
What amperage circuits would be provided? Does each slot get a 20 amp circuit? What is the temperature control of the facility? You wouldn't have remote control of the power directly, so in the even of a hard lockup or shut down, we would provide a "request power on" button in a web interface. Your rig would be powered back on within 24 hours. For this reason, it is imperative that you already have a stable configuration prior to shipping to our facility.
But I could install my own remote-power device, correct? All management of your mining rig(s) would be done by you. We wouldn't touch it. You would be responsible for remotely accessing your server, once we unpackage it and start it up, to get the mining running. You would also be responsible for your own software security. Given that many machines would be on the same network, you should set up a software firewall appropriately to isolate your rig(s) from the other rigs.
Will you install video cards? Rigs shipped with heavy video cards installed are likely to be damaged by the shipping companies. You really should enforce separate rig networks, otherwise customers could interfere with each others, like with incorrect networking settings, either through mistakes or on purpose. Also, do you see any potential problems that I have not addressed? Any potential security or performance issues?
What is the length of contract for a slot? Do you guarantee uptime (network, electricity, etc) of the facility? Good questions. A slot could be 4U, or could be any size tower up to a full size tower. Provisions will be made for both, with expectations of most people using towers, since they are cheaper. Circuits would be 20 amp. Each machine wouldn't necessarily be on its own circuit, but we would take a look at the hardware inside each machine, determine an estimate for amperage draw at full load, and balance the machines across the electric circuits appropriately. Please clarify regarding your question about temperature control - the temperature would be controlled. Yes, you could provide your own power-control device for us to install if you would like. I meant to mention that as well - everyone would have to uninstall their video cards prior to shipping the system to us. They should be clearly labeled as to what slot each one is to be installed in. Video card installation will be done by us at no additional charge. I would like to hear more information about enforcing separate networks for each rig. What would be necessary to accomplish this? It is something I have not done before. I am assuming a dedicated IP address for each machine would be necessary at a minimum. Probably a special type of router/firewall device... I do agree this would be the optimal route to go though. Perhaps I could include a small monthly charge for each dedicated IP address a person wanted, and they could choose whether they wanted to go that route or not. Billing would be monthly with no contract. Invoicing would be done on the same day of every month. The base fee would not be pro-rated if a person decided to pull their rig out early. In order to retrieve their rig, a person would have to pay shipping costs as well as their latest month's bill. Not sure about whether uptime would be guaranteed or not. On the one hand, I think that it should be. On the other hand, a day of downtime with 40 rigs could put me out of business. I need to find the optimum solution with regards to that potential issue, and am open to suggestions.
|
|
|
|
alans
Newbie
Offline
Activity: 50
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 05:06:17 PM |
|
You could set it up so that each user gets an individual vlan and a vpn account. That users machines are put on their own vlan segmented from the rest of the data center. To access your machines you would have to access the centers VPN service first. The access to individual vlans could be controlled at the user level of the vpn service.
From the end users point of view they would connect their home pc to a vpn connection and then be able to vnc/ssh/telnet/ping whatever was open on their systems in the data center.
From the centers perspective each user is assigned a vlan and their systems are placed on it. You'd need a router and someone savvy to set it up. The center would also need to host a vpn service that provided access control mechanisms to restrict each account's rights to access only certain vlan's behind the firewall.
That's how it has been done in the traditional data center hosting services I've used in the past. They typically use proprietary commercial versions but I believe there may be open source solutions for both the vlan management and vpn solutions.
|
|
|
|
SgtSpike (OP)
Legendary
Offline
Activity: 1400
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 05:12:58 PM |
|
Thanks alans, I'll have to do some research on that front. It might be cost-prohibitive to set up something like that for a small/startup operation, but worth checking in to nonetheless.
|
|
|
|
alans
Newbie
Offline
Activity: 50
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 05:21:37 PM |
|
Don't let the startup cost scare you too much. For a router you could setup a pfsense box with 1 NIC for the external connection and 1 NIC for each user. pf sense would provide firewall and connectivity (dhcp...) services for the clients and routing for the vpn clients. Then another system to run OpenVPN or some other protocol. The routing setup might get a little tricky once you max out the number of NIC's you can put in the pfsense box but by then you should have the income to justify a real router...
|
|
|
|
SgtSpike (OP)
Legendary
Offline
Activity: 1400
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 06:00:12 PM |
|
Don't let the startup cost scare you too much. For a router you could setup a pfsense box with 1 NIC for the external connection and 1 NIC for each user. pf sense would provide firewall and connectivity (dhcp...) services for the clients and routing for the vpn clients. Then another system to run OpenVPN or some other protocol. The routing setup might get a little tricky once you max out the number of NIC's you can put in the pfsense box but by then you should have the income to justify a real router...
It would take 20-25 rigs to break even on monthly costs at the price levels described above, without accounting for initial expenses. Even maxing out the slots in a pfsense box wouldn't bring me close to that number, unless several people rented several slots each. A "real router" might be cheaper and quicker to go with in the first place. What sort of router would be required to run these services? A Cisco solution, perhaps? Or something like this? http://www.newegg.com/Product/Product.aspx?Item=N82E16833706014
|
|
|
|
alans
Newbie
Offline
Activity: 50
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 06:36:55 PM |
|
You would only really need one port or one nic in a pfsense box for each customer. If they ran >1 machine then you'd just need a hub/switch to connect each of their machines back to that master port/vlan.
|
|
|
|
infested999
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 07:04:40 PM |
|
You would only really need one port or one nic in a pfsense box for each customer. If they ran >1 machine then you'd just need a hub/switch to connect each of their machines back to that master port/vlan.
Couldn't you just use 1 NIC and a large switch for everyone. Can pfsense create 100+ Virtual LANs?
|
|
|
|
SteveFL
Newbie
Offline
Activity: 33
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 07:19:31 PM |
|
You're going to be hard pressed to offer space for the prices your mentioned. Don't forget fire suppression, multiple ISPs, UPS/Generator and redundant power feeds.
Without the ISPs you're a DDoS away from losing customers. Without a UPS/Power conditioner one lightning strike could do you in. Multiple power feeds are also a concern along with standby power. How long is acceptable downtime?
There's a reason most CoLo facilities charge $150-$200 base + a charge per extra 1U per month.
|
|
|
|
infested999
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 07:41:14 PM |
|
You're going to be hard pressed to offer space for the prices your mentioned. Don't forget fire suppression, multiple ISPs, UPS/Generator and redundant power feeds.
Without the ISPs you're a DDoS away from losing customers. Without a UPS/Power conditioner one lightning strike could do you in. Multiple power feeds are also a concern along with standby power. How long is acceptable downtime?
There's a reason most CoLo facilities charge $150-$200 base + a charge per extra 1U per month.
multiple ISPs is usually not possible in most places in America because Comcast has a monopoly. I don't think we have to worry about lightning strikes with the modern PSU's in the machines now-a-days. BTW SteveFL are you in Florida?
|
|
|
|
SgtSpike (OP)
Legendary
Offline
Activity: 1400
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 09:36:20 PM Last edit: July 26, 2011, 09:57:32 PM by SgtSpike |
|
You would only really need one port or one nic in a pfsense box for each customer. If they ran >1 machine then you'd just need a hub/switch to connect each of their machines back to that master port/vlan.
Still, if I get 20 customers with 2 rigs, I'd need 20 ports. That's going to either be extremely costly on a pfsense box (with NICs that have four ports each), or would require some special router. I would rather go the router route to begin with, but I have no idea what I should be looking for as far as features go. You're going to be hard pressed to offer space for the prices your mentioned. Don't forget fire suppression, multiple ISPs, UPS/Generator and redundant power feeds.
Without the ISPs you're a DDoS away from losing customers. Without a UPS/Power conditioner one lightning strike could do you in. Multiple power feeds are also a concern along with standby power. How long is acceptable downtime?
There's a reason most CoLo facilities charge $150-$200 base + a charge per extra 1U per month.
The point of this would be to offer low-cost CoLo for miners. It wouldn't have all the features that your typical datacenter might have, and probably won't have the 99.% uptime that your typical datacenter might have. - Fire suppression is important, but I would only go so far as the fire marshall/insurance required on that front. - I wouldn't have a UPS/Line conditioner or generator. People could provide their own if they like. The most I will provide is a fuse-protected power strip. - There wouldn't be multiple power feeds. We VERY rarely get any sort of power outages here, so I'm not worried about that. In the city, we probably get 1 power outage every two years. - DDOS is a major concern, and something I am still researching (and open to suggestions on). I have several potential providers, but am unsure that any of them could provide DDOS prevention in any capacity.
|
|
|
|
NothinG
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 10:21:48 PM |
|
- DDOS is a major concern, and something I am still researching (and open to suggestions on). I have several potential providers, but am unsure that any of them could provide DDOS prevention in any capacity.
Throw pfSense in front of the servers.
|
|
|
|
SgtSpike (OP)
Legendary
Offline
Activity: 1400
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 11:16:48 PM |
|
- DDOS is a major concern, and something I am still researching (and open to suggestions on). I have several potential providers, but am unsure that any of them could provide DDOS prevention in any capacity.
Throw pfSense in front of the servers. Thanks. It's that easy, huh? I'll definitely have to do some reading/experimentation on pfsense. Thinking about cooling now.... I figure for 40 rigs, it'll probably be, on average, around 500w of electric usage per rig. So 20kw. That's fine, but then I'll need an estimated 70,000 BTU a/c (3-4 ton) to cool it. Yikes! ![Shocked](https://bitcointalk.org/Smileys/default/shocked.gif) That'll definitely bring the initial investment cost up... Right now, it's looking like initial investment will be $6000 to $9000. Quite a large amount for relying on bitcoins to recover it. I wonder if there would be enough interest from investors to put up the capital for it. Heck, that's only 642 BTC. ![Tongue](https://bitcointalk.org/Smileys/default/tongue.gif) I am definitely happy to see the poll results thus far. If there are truly at least two people willing to pay for 9+ rigs each, that would be a great jumpstart to such a business. The two who voted that way - if you wouldn't mind, PM me so we can talk in further detail.
|
|
|
|
NothinG
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 11:36:43 PM |
|
- DDOS is a major concern, and something I am still researching (and open to suggestions on). I have several potential providers, but am unsure that any of them could provide DDOS prevention in any capacity.
Throw pfSense in front of the servers. Thanks. It's that easy, huh? I'll definitely have to do some reading/experimentation on pfsense. I setup a pfSense box at my house running 10/100/1000 nic [motherboard/lan (private)] and an 10/100 nic [pci/wan (public)], 1GB DDR3, 4x i5, 80GB HDD (for squid). Had one of my buddies DDoS me with 2 servers (1Gb/s each) from a DC totaling 2Gb/s, my home network maxed out at 300Kb/s but I could still browse websites and talk on messenger. pfSense had blacklisted his IP leaving me with a slight lag but still able to view sites like youtube without a hitch because of how I had caching/squid setup.
|
|
|
|
Dargo
Legendary
Offline
Activity: 1820
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 26, 2011, 11:39:03 PM |
|
This is probably a crazy idea, but I'm going to toss it out there anyway. What if you simply don't offer customers VNC/SSH at all? You could stipulate that customers all use the same OS - BAMT let's say. It is the customers responsibility to get the rig optimized and running rock-solid stable before they ship it to you. Then, if they want to make an adjustment, they would simply email you a new configuration file, and you replace this file with the old one and reboot the rig. You could charge a small fee for this. They would do this at their own risk, since if it didn't work, they would have to either revert to the old file, or you would charge them an hourly rate to trouble shoot. You might even have them ship the rigs with an extra copy of the OS configured how they want, so if the rig goes haywire, you can stick in the backup copy and reboot (again for a small fee). If you are worried about customers installing something weird on the OS (even malware) that could bolix up the network, you could stipulate what version of the OS they are going to run, which they could download from your web page, configure how they like, then just send you the config file which you would put into your own copies of the OS, and boot up. The BAMT config file is pretty simple, so it would be easy to check to make sure they didn't add any abnormal scripts. The author of BAMT might even be willing, for a reasonable fee, to create a more stripped down and newbie friendly version of the OS to suit your needs.
Personally, I'm not sure how much need I would have for access to my software once I get everything set up the way I like. And the only personal info I might have on the rig would be my public bitcoin key and/or my worker names and passwords. Most of the adjustments I might want to make could be handled by replacing the config file. It might also be a hell of a lot easier for you to maintain a network like this if very rig is running the same OS. One major downside here is that it is harder to overvolt cards in linux (but pretty much everything else is relatively easy at this point). I'm not sure how much of a con this is, though, since it probably isn't wise for people to be overvolting cards that are going to be run at a remote location (i.e. cards that they can't keep close eye on).
Anyway, just a thought I had that I think would work for me as a customer.
|
|
|
|
SgtSpike (OP)
Legendary
Offline
Activity: 1400
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 27, 2011, 12:37:04 AM |
|
This is probably a crazy idea, but I'm going to toss it out there anyway. What if you simply don't offer customers VNC/SSH at all? You could stipulate that customers all use the same OS - BAMT let's say. It is the customers responsibility to get the rig optimized and running rock-solid stable before they ship it to you. Then, if they want to make an adjustment, they would simply email you a new configuration file, and you replace this file with the old one and reboot the rig. You could charge a small fee for this. They would do this at their own risk, since if it didn't work, they would have to either revert to the old file, or you would charge them an hourly rate to trouble shoot. You might even have them ship the rigs with an extra copy of the OS configured how they want, so if the rig goes haywire, you can stick in the backup copy and reboot (again for a small fee). If you are worried about customers installing something weird on the OS (even malware) that could bolix up the network, you could stipulate what version of the OS they are going to run, which they could download from your web page, configure how they like, then just send you the config file which you would put into your own copies of the OS, and boot up. The BAMT config file is pretty simple, so it would be easy to check to make sure they didn't add any abnormal scripts. The author of BAMT might even be willing, for a reasonable fee, to create a more stripped down and newbie friendly version of the OS to suit your needs.
Personally, I'm not sure how much need I would have for access to my software once I get everything set up the way I like. And the only personal info I might have on the rig would be my public bitcoin key and/or my worker names and passwords. Most of the adjustments I might want to make could be handled by replacing the config file. It might also be a hell of a lot easier for you to maintain a network like this if very rig is running the same OS. One major downside here is that it is harder to overvolt cards in linux (but pretty much everything else is relatively easy at this point). I'm not sure how much of a con this is, though, since it probably isn't wise for people to be overvolting cards that are going to be run at a remote location (i.e. cards that they can't keep close eye on).
Anyway, just a thought I had that I think would work for me as a customer.
I like crazy. ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) Interesting idea, and I think it has merit. I suppose the biggest difficulty would be for those who do not know how to configure a rig in the given OS environment that is chosen. I don't think that renters would be very fond of the idea though, but I am interested in fleshing it out and getting more opinions.
|
|
|
|
Dargo
Legendary
Offline
Activity: 1820
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 27, 2011, 03:15:04 AM |
|
ya, I guess this would really be plan B if it proves too difficult to offer remote connection for everyone. You don't need to know much about linux to confgure BAMT. You need to know how to find your adapter #s, but the rest of the config is just a matter of specifying pools, clocks, fan speeds, etc. Your customers are going to have to know how to run headless, connect remotely, and configure for automated operation anyway, and this will be much easier if it is all ready to go via a config file. You can monitor things like temp, hash rate, share submission rate, and set email alerts if these get below/above specified limits. You probably don't want to get too complicated, but it wouldn't be that hard to have various configurable recovery routines for the more common difficulties. Also, there is no reason (I think) why the config file needs to be on the usb - this could be fetched from somewhere on the network. And customers could login to their accounts to modify the file themselves (if you want to get fancy you could have a web interface for this so they are just filling in boxes). And if the rig can be alerted when the config file is changed, it could just fetch the new file and restart. In addition to the config file, you could offer a few basic web options, such as 1) shut down rig, 2) restart miners, 3) reboot rig, 4) fetch new config file and restart. So I think you could offer customers quite a bit of control without any direct connection to their rigs. Perhaps you could offer this as the "basic" package, and charge more for those who simply must have remote connection (or charge for remote connection on a per-use basis). Another advantage of this is that it makes software troubleshooting very simple. If a customer's rig isn't acting right, the first step is to revert to an earlier configuration that is know to be stable (they do this). If that fails, you yank the usb, put in a new one, and reboot. If that fails, then it is probably a hardware problem, and the customer can decide to have the rig shipped back to them, or you offer technical support at an hourly rate.
|
|
|
|
SgtSpike (OP)
Legendary
Offline
Activity: 1400
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 27, 2011, 04:00:23 AM |
|
ya, I guess this would really be plan B if it proves too difficult to offer remote connection for everyone. You don't need to know much about linux to confgure BAMT. You need to know how to find your adapter #s, but the rest of the config is just a matter of specifying pools, clocks, fan speeds, etc. Your customers are going to have to know how to run headless, connect remotely, and configure for automated operation anyway, and this will be much easier if it is all ready to go via a config file. You can monitor things like temp, hash rate, share submission rate, and set email alerts if these get below/above specified limits. You probably don't want to get too complicated, but it wouldn't be that hard to have various configurable recovery routines for the more common difficulties. Also, there is no reason (I think) why the config file needs to be on the usb - this could be fetched from somewhere on the network. And customers could login to their accounts to modify the file themselves (if you want to get fancy you could have a web interface for this so they are just filling in boxes). And if the rig can be alerted when the config file is changed, it could just fetch the new file and restart. In addition to the config file, you could offer a few basic web options, such as 1) shut down rig, 2) restart miners, 3) reboot rig, 4) fetch new config file and restart. So I think you could offer customers quite a bit of control without any direct connection to their rigs. Perhaps you could offer this as the "basic" package, and charge more for those who simply must have remote connection (or charge for remote connection on a per-use basis). Another advantage of this is that it makes software troubleshooting very simple. If a customer's rig isn't acting right, the first step is to revert to an earlier configuration that is know to be stable (they do this). If that fails, you yank the usb, put in a new one, and reboot. If that fails, then it is probably a hardware problem, and the customer can decide to have the rig shipped back to them, or you offer technical support at an hourly rate.
I don't have enough knowledge of linux to set up that complex of a system, but it's possible I could recruit a linux guru for the purpose at not too high a cost.
|
|
|
|
|
|