Hi friends,

I just got an idea ... out there is a lot (a very large lot) of devices that run based on some kind of linux, permanently running and most of the time idling lazy around. What I mean are i.e. such devices like DSL modems (i.e. AVM Fritz!Box), routers, and other thingies pretending to be simple..

These boxes are around in numerous quantities, stay online most of the time - almost all of the 24/7 I presume - and most of the time they are lazy. And in the case of DSL modem boxes, they are online 24/7 ! Did anybody thing about hooking a BOINC client into them ?

The only obstacle I find - besides of low ressources in one or another box - is the lack of any display ... that would require a BOINC manager able to manage a remote client or a web browser interface within the client.

Any comment to that concept ?

Seneca

Hi friends,

I just got an idea ... out there is a lot (a very large lot) of devices that run based on some kind of linux, permanently running and most of the time idling lazy around. What I mean are i.e. such devices like DSL modems (i.e. AVM Fritz!Box), routers, and other thingies pretending to be simple..

These boxes are around in numerous quantities, stay online most of the time - almost all of the 24/7 I presume - and most of the time they are lazy. And in the case of DSL modem boxes, they are online 24/7 ! Did anybody thing about hooking a BOINC client into them ?

The only obstacle I find - besides of low ressources in one or another box - is the lack of any display ... that would require a BOINC manager able to manage a remote client or a web browser interface within the client.

Any comment to that concept ?

Seneca

At one time, I considered running BOINC on my son's PS3, but just in playing video games it gets uncomfortably warm.

I can just imagine how hot it would get running at 100% processor utilization.

Burning up a PS3 and having to buy them another, or buying a quadcore that is designed for this purpose wasn't a difficult choice.

---

If everyone is thinking alike, someone isn't thinking. - General George Patton

...
The only obstacle I find - besides of low ressources in one or another box - is the lack of any display ... that would require a BOINC manager able to manage a remote client or a web browser interface within the client.
...

BOINC Manager is able to manage a remote client. You only have to set up your remote_hosts.cfg and gui_rpc_auth.cfg files.

Gruß,
Gundolf

---

Computer sind nicht alles im Leben. (Kleiner Scherz)

Hello,

...
The only obstacle I find - besides of low ressources in one or another box - is the lack of any display ... that would require a BOINC manager able to manage a remote client or a web browser interface within the client.
...

BOINC Manager is able to manage a remote client. (...)

Ooops - I've not been aware of that.

So the only thing in the way would be to create a client for i.e. AVM's Fritz!Box and find a way to install it on top of the firmware. Any volunteers around ?
:-)

Seneca

Hello,

Burning up a PS3 and having to buy them another, or buying a quadcore that is designed for this purpose wasn't a difficult choice.

PS3 etc. - OK. But a flatrate DSL/VoIP modem box usually idles around 24/7 while using electric power, a new quad core would cost money AND extra energy.

Utilizig such a box would add computing power for free (besides of the possible extra power consumption for running 100%) - which is one of the main features of BOINC. And many boxes reside in cool basment rooms - no risk of overheating 'em.

Greetings

Seneca

BTW: Anybody out there who wants to join my seti@home team NATHAN ? :-)
O=O

So the only thing in the way would be to create a client for i.e. AVM's Fritz!Box ...

No, there's also the little matter of creating a science application for it.
BOINC itself doesn't do any crunching, as has been pointed out many times before.
And science apps need computing power and storage.

Utilizig such a box would add computing power for free (besides of the possible extra power consumption for running 100%) - which is one of the main features of BOINC. And many boxes reside in cool basment rooms - no risk of overheating 'em.

I think, that you get very fast a overheating, because if the CPU ran at a higher load. And the FritzBox has no fan !
My FritzBox is at up to 65 % CPU, if I am surfing and 50 % in idle. So I am not think, that there would be enught free resources for any science appliation.
Also, it's a MIPS EL CPU, not shure how fast is it and if it's posible, if the WU finish within a deadline.
There are still BOINC and SETI ports for Linux MIPS EL available. But please use it at your own risk !

Edit: Searched a little bit around. The MIPS 4K CPU in the FritzBox is about 25 to 30 times slower as the CoreDuo 2 GHz CPU. So a WU could need about 5 to 7 days - if the CPU gets the full utilisation. So about 2 weeks/WU would be normal, because the FritzBox is Utilized by 50 % by it's normal work.
Also if you ran into overheating, you need to throttle down the application (for example 25 % for BOINC science) to prevent overheating, the WU needs about 8 weeks/per WU. - SETI deadline is 4 weeks for the typical WUs.

I've played a little bit around with my Fritzbox....
The FritzBox 7170 has only 32 MB memory installed. About 6 MB free memory. It's absolutly to less for BOINC and any science application. The BOINC client needs about 15 to 20 MB memory, SETI about minimum 32 MB, SIMAP is about 10 MB required memory.

No the real hurdle is that most all BOINC DC sciences need an Intel instruction set CPU and sufficient numbers to make it economical. There's a billion untapped PC's around, so who's going to invest in running it on speciality devices and support it?

No. There are a lot of "devices" or small/very small systems which are a great market for BOINC and a small or throtteled science application. Also you need not need any Intel CPU, BOINC rans very beautyfull on other CPUs (PARISC, IA64, PPC, Power,...) - The HPUX, AIX,... ports makes a very good RAC at the other projects...
Also a mayory of mobile or home entertainment devices did ran on non Intel CPUs, mostly PPC, MIPS and ARM.
There are intentions from the BOINC side to port BOINC on small devices. It was discussed on the devs mailinglists and also the first attempts was succesfull.

PS3 etc. - OK. But a flatrate DSL/VoIP modem box usually idles around 24/7 while using electric power, a new quad core would cost money AND extra energy.

Running BOINC on a DSL modem (which has too little CPU power anyway) *would* use more energy. CPUs use a lot more power while they're in use. Or more correctly: any modern CPU is advanced enough to lower its power usage a lot while they're not in use, and a processor in a DSL modem would be in such low-power mode most of the time.

I have tried compiling BOINC for my iPod touch. It got warmer than I like it during the CPU-intensive process of compiling C++ code, so I gave up; since even *if* I succeeded at compiling it (and porting some project science app), it would get too hot while running it too.

But at least it was interesting to see :)

---

I think as projects get bigger and require more CPU power you can forget about using anything other then the latest fastest PC's to run BOINC with most project on anything.

It would be nice to run BOINC on small PC's or even older ones runing DOS that only required the fan for the PSU. They would take less power but you could only run small WU's.

Someone would have to make a copy of BOINC that can run a project and require less then 512MB.

Some of these PC had less than a 200 Watt PSU. and a very slow CPU!

---

I think as projects get bigger and require more CPU power you can forget about using anything other then the latest fastest PC's to run BOINC with most project on anything.

There are and hopefully will be projects, which are not CPU-intensive - such would even run on a mobile phone ;-)

Someone would have to make a copy of BOINC that can run a project and require less then 512MB.

BOINC client alone needs just a bunch of MBs memory.
There are projects, which require as much as 1.5 MB of memory.
Satisfied? ;-)

Peter

I would love to be able to run BOINC on my mobile devices like my 2 HTC Wizards and my iPod Touch. For the 2 smartphones, I no longer use them at all, and on the iPod Touch, I use it very sparingly. Would be cool if I could throw these devices into the mix.

Boincoid (boincoid.sf.net) have been developing a Java port. It's still in the pre-alpha stage, but the potential is there.

The crunching would not be as efficient as compiling for the architecture, but a Java client would make it possible to run on just about any machine.

The problem I have with Boincoid (or Java Boinc) at the moment is that it has been designed specifically for Java SE, and as those who have anything to do with mobile applications will know, there should be some focus on Java ME - which is what most smart phones run.

However, as I say, the potential is enormous, and this is only the very early stages after all.

I see no problem in crunching WUs on my mobile while it's charging. :-)

It would be nice to run BOINC on small PC's or even older ones runing DOS that only required the fan for the PSU. They would take less power but you could only run small WU's.

Old computers don't use less power if you think about computation speed per watt (or, like some BOINC users call it, credits/watt :P). You can be sure old computers will really lose there.

---

Hi... just wondering... what do you think about an ARM architecture port? It would run on nokia smartphones, and some of the latest netbooks(and not-yet released notebooks), maybe routers too :) ... ARM is getting popular because of the low power consumption and relatively good computing power (even MS is considering an ARM port of Windows 7). Anyone interested (I could use some help...)?

I am experimenting with a ARM science applications port in the moment. Hopefully I can release the BOINC client and one or more science applications soon.
The problem on the ARM CPUs is, that it have a good integer perfomance, but a very low to low Floating performance...

In a way, porting BOINC to "unusual" devices is almost a revival of the earlier days of distributed computing (e.g., Seti@home) when the platforms we used were less, powerful than the PDAs and embedded processors we use everyday.

I first learned of Seti@home when my home PC was a 386DX25 running OS/2 Warp. My first several thousand WUs were crunched on that machine. Over time, as the average PC became more powerful, the projects concomitantly increased the memory and CPU requirements for their WUs. If SETI could make small enough WUs back then, why can't they now? Before any port to an "unusual" platform can do any real work, we need a renaissance in tiny WUs-- perhaps they could be restricted by platform so that the bladeservers don't suck them all up in one clock cycle.

While the bottom-line contribution to the projects would be almost meaningless, the availability of tiny WUs would provide the raw material for programmers to do some interesting things with interesting hardware.

---

I am experimenting with a ARM science applications port in the moment. Hopefully I can release the BOINC client and one or more science applications soon.
The problem on the ARM CPUs is, that it have a good integer perfomance, but a very low to low Floating performance...

Anybody interested to help test the BOINC client and some science applications on the ARM CPU ? - Please contact me...

I have a Brother Word Processor and one of those old Mail Station... be nice if there was a way to get them to run BOINC!

---