IRC logs of #boinc for Thursday, 2012-11-08

01:16 *** zombie67 has quit IRC

03:22 *** microchip_ has joined #boinc

03:22 *** microchip_ has quit IRC

03:22 *** microchip_ has joined #boinc

05:33 *** desti has joined #boinc

05:37 *** desti_T2 has quit IRC

05:54 *** Vobine has joined #boinc

05:54 *** Psychoholic has quit IRC

05:54 *** Vobine is now known as Psychoholic

06:54 *** Caterpillar has joined #boinc

07:16 *** microchip_ has quit IRC

07:34 *** RBecker has quit IRC

08:05 *** RBecker has joined #boinc

08:05 *** RBecker has quit IRC

08:06 *** RBecker has joined #boinc

08:35 *** desti_T2 has joined #boinc

08:37 *** desti has quit IRC

09:17 *** Aeternus has joined #boinc

09:17 *** Aeternus has quit IRC

09:17 *** Aeternus has joined #boinc

10:20 *** efc has quit IRC

10:20 *** microchip_ has joined #boinc

10:20 *** microchip_ has quit IRC

10:20 *** microchip_ has joined #boinc

10:48 *** MarcoFe has joined #boinc

10:54 *** CoderForLife has quit IRC

12:14 *** Aeternus has quit IRC

13:04 *** mtrg has joined #boinc

13:04 <mtrg> is this the official boinc channel?

13:05 <microchip_> yes

13:06 <mtrg> is there any general grid computing channel here in freenode?

13:08 <mtrg> i'm planning to run a computing grid across georgraphically scattered comodity hardware that i own, and a general discussion on grid computing would be great for me

13:08 <microchip_> no idea

13:13 <Tank_Master> some teams have channels here, there may be some projects as well

13:14 *** si has joined #boinc

13:19 <PovAddict> mtrg: explain your task and I may be able to tell you if BOINC is suitable

13:21 <mtrg> PovAddict: performing an exhaustive search for some machine learning stuff based on a space of problems that are divided into parts where each participating computer takes care of a few parts

13:22 <PovAddict> do you own all the machines that will run as clients? (ie. it's not really "volunteer" computing?)

13:22 <mtrg> my employer does

13:22 <mtrg> some of them are laptops, owned by employer too

13:23 <PovAddict> ok, question was if they are from the "general public" or not

13:23 <mtrg> they are staff, not any random people from outside (so far)

13:24 <mtrg> but that might be an interesting addition although it is not a priority.. we got 100s of computers that are setting idle locally that harnessing their power could be great

13:25 <PovAddict> bbiab

13:25 <mtrg> one of the issues why i'm not confident in deciding what i need to know is that i'm not fully sure about the general arch of a grid. all components that i can think of are: a scheduler and a middleware

13:25 <PovAddict> dealing with a more urgent matter + need ice cream

13:25 <mtrg> tyt

13:35 <mtrg> anyone knows a good source that explains the general arch. of grids?

13:42 <Tank_Master> not right off

13:43 *** yoyo[RKN] has joined #boinc

13:43 <Tank_Master> but basicly you have a project server that hosts connect to via the boinc app

13:44 <mtrg> a server that keeps track of all other vulanteering computers?:

13:44 <Tank_Master> yes, the server logs hosts, users, apps, and some other things

13:45 <mtrg> isn't this kinda centerlized rather than distributed?

13:45 <Tank_Master> i think its configerable on how long some of that info stays, like the histroy of results for returned work

13:46 <Tank_Master> the server doesnt runs any of the work, it just keeps track of it

13:46 <Tank_Master> the work is what is distributed

13:46 <mtrg> does the server also collect back the solved problems?

13:46 <Tank_Master> yes

13:46 <Tank_Master> when the host finishes the work, it is uploaded to the server

13:46 <mtrg> so it's only good if the time taken to solve a problem is usually larger than the time taken to send and recieve the request/answer respectively

13:47 <PovAddict> yes

13:47 <PovAddict> and if you use volunteered resources, the time to process has to be *MUCH* larger than the time to send/receive

13:47 <Tank_Master> its more ideal for the type of work thats not dependent on previous work

13:47 <mtrg> yeah. fits my case fairly well

13:47 <PovAddict> since they are on normal broadband connections, and either way wouldn't want BOINC to use their internet connection all the time

13:47 <Tank_Master> and work that can be hightly paralized

13:48 <mtrg> yeah

13:48 <Tank_Master> there are a few projects who's work is based on returned results, but thats not common

13:49 <yoyo[RKN]> moin

13:49 <PovAddict> I have seen two general cases of that in practice

13:49 <Tank_Master> mtrg, yoyo[RKN] runs the yoyo project

13:49 <mtrg> "returned results"? is there any non-returned results project? (i'm assuming that a returned result is a returned answer from a vulanteer computer

13:50 <Tank_Master> PovAddict has written and managed some projects in the past as well

13:50 <PovAddict> mtrg: no, all projects have the clients return data :P

13:50 <mtrg> data != answer?

13:50 <PovAddict> mtrg: Tank_Master means some projects create further work based on the result of previous tasks

13:50 <mtrg> err, a/answer/result/

13:50 <Tank_Master> i ment work they issue to clients is based onn the results of work previously sent and returned

13:50 *** si has quit IRC

13:51 <mtrg> i see

13:51 <Tank_Master> you will find some results will return incorrect sometimes

13:51 <mtrg> how does boinc decide which workstation to get which task? is there any sort of scheduler?

13:52 <yoyo[RKN]> the server doesn't send results

13:52 <PovAddict> yoyo[RKN]: nobody implied that :P

13:52 <PovAddict> mtrg: it's mostly arbitrary (first come first serve)

13:52 <yoyo[RKN]> the client requests works

13:52 <Tank_Master> thats why boinc has a built in redundency where you can send the same work unit out to multible hosts and then the server compaires the results

13:52 <yoyo[RKN]> :D

13:53 <mtrg> PovAddict: i see. i guess it's because tasks are so many that eventually all of them will get busy anyway

13:53 <PovAddict> mtrg: the scheduler logic is to make sure workstations get tasks they can actually process (eg. enough disk/memory), and a reasonable amount of them

13:53 <mtrg> oh so there is a scheduler

13:53 <PovAddict> yes, I just mean it doesn't exactly pick and choose where to send each task

13:54 <Tank_Master> it just determans if the work unit avalibe for process is acceptible to run on the host requesting work

13:54 <mtrg> does the boinc server run that scheduler, or is it the vulanteering computers that run the scheduler and based on it request jobs?

13:54 <PovAddict> mtrg: communication is initiated by the clients, so the server can't decide when and how to send tasks to clients

13:54 <yoyo[RKN]> scheduler runs on the server

13:55 <PovAddict> a client asks for tasks, the server gives it some

13:55 <mtrg> so the server needs to keep track of which vulanteering workstation has which capacity

13:55 <Tank_Master> like if theres no work for the linux app, and a linux host requests work, the schedular wont send that host with work, but ti will send a message of "no work for your type of computer"

13:55 <PovAddict> the client has a local queue

13:55 <PovAddict> and processes the tasks it got

13:55 <mtrg> would be nice if the client says "give me al ot of task as i'm a tough guy"

13:56 <yoyo[RKN]> this is the case

13:56 <Tank_Master> the host will request enough to buffer work for a specified amount of time, max of 10 days

13:56 <PovAddict> mtrg: clients have a local queue, so they usually do ask for multiple tasks

13:56 <mtrg> yoyo[RKN]: why does the server do the queueing then?

13:56 <yoyo[RKN]> the client requests X seconds of works and tells it capabilities

13:56 <mtrg> yoyo[RKN]: err, why does the server do the scheduling*

13:56 <PovAddict> both client and server have some kind of queue and some kind of scheduling :P

13:57 <yoyo[RKN]> the scheduler decides, based on the client request, which work it sends

13:57 <Tank_Master> only the server knows what work is avalible to run

13:57 <mtrg> i see

13:57 <PovAddict> there's three parts that you may think as "schedulers"

13:57 <mtrg> so a client requests problems by announcing its abilities, and a server attempts to find the best match

13:57 <Tank_Master> the host asks the server whats avalible for itself, and the server looks up in the db for any work that matches the host's capibilities

13:58 <mtrg> Tank_Master: got it

13:58 <PovAddict> - the server has a process that chooses how many and what tasks to send to a client when the client asks for tasks

13:58 <Tank_Master> there ase some fallback, like a 64bit host will ast for work, and the server may instead send 32bit work

13:59 <Tank_Master> are*

13:59 <PovAddict> - the client has a 'scheduler' that chooses what task in the local queue to run next, and maybe even pause a running task to start or resume another

14:00 <yoyo[RKN]> &wx sxf

14:00 <PovAddict> - and the client also has algorithms that decide when to ask for more work, how much work to ask for, and from what projects

14:00 <Romulus> yoyo[RKN]: Temperature: 48°F / 9°C | Humidity: 76% | Pressure: 29.95in / 1014hPa (Steady) | Conditions: Mostly Cloudy | Wind Direction: WSW | Wind Speed: 14mph / 22km/h ; Chance of Rain. High:51 ° F.; Chance of Rain. Low:44 ° F.; Chance of Rain. High:50 ° F.; Chance of Rain. Low:44 ° F.; Chance of Rain. High:51 ° F.; Scattered Clouds. Low:41 ° F.;

14:00 <PovAddict> (since you can attach a client to multiple projects)

14:01 <Tank_Master> you can also set limits on the server as to how much work hosts are alowed to have at one time

14:01 <Tank_Master> to prevent one host from hording all the work, and thus the other hosts dont get any

14:02 <Tank_Master> or to prevent a host from corrupting all the work (if there is a computation error detected, the max alowed for that host is reduced)

14:03 <Tank_Master> down to 1 or 2 a day, then as the host returns valid work, that limit is raised back to the upper limit defined by the server config files

14:03 <mtrg> yoyo[RKN]: what does that bot really do? is it backed by some sort of grid?

14:03 <Tank_Master> its someone's personal bot

14:04 <PovAddict> mtrg: what bot, Romulus? it's just a generic IRC bot that we mostly use for weather :)

14:04 <yoyo[RKN]> PovAddict do you have write permissions to boinc git?

14:04 <PovAddict> yoyo[RKN]: no

14:04 <Tank_Master> you think it would be the mess it is if he did? :P

14:04 <mtrg> PovAddict: heh. i see

14:05 <Tank_Master> &seen god

14:05 <Romulus> Tank_Master: god was last seen in #boinc 6 years, 49 weeks, 0 days, 8 hours, 1 minute, and 13 seconds ago: <god> why have you summoned me from my holy grail?

14:05 <Tank_Master> yeah, its been here a while

14:05 <mtrg> so it seems that grid computing is a pretty broad and a generic term for a pretty big space of possible arch. that i might have

14:05 <PovAddict> mtrg: all we said so far is specific to BOINC :P

14:06 <PovAddict> other systems, for example, support having the server immediately give a task to a client and track its progress in real time

14:06 <Tank_Master> with enough time and determonation, you can use just about any piece of hardware to run an app for boinc

14:06 <mtrg> how does grid computing differentiate than utility computing?

14:07 <mtrg> is it just the billing?

14:07 <PovAddict> or completely different approaches like mapreduce in computers with fast interconnections

14:07 <mtrg> i'm a bit lost on what grid computing really is

14:07 <PovAddict> mtrg: I think they are orthogonal

14:08 <PovAddict> wikipedia says "Utility computing is the packaging of computing resources, such as computation, storage and services, as a metered service. This model has the advantage of a low or no initial cost to acquire computer resources; instead, computational resources are essentially rented."

14:08 <yoyo[RKN]> PovAddict would Boinc recognize a system which has 256 threads? Is there some limit in Boinc client?

14:08 <PovAddict> yoyo[RKN]: you mean cores that it can run tasks on?

14:09 <yoyo[RKN]> yes

14:09 <mtrg> PovAddict: i see. probably grid computing, with utility computing, can give us cloud computing (corrrect me if i'm wrong)

14:09 <PovAddict> mtrg: "cloud" is like "web 2.0"

14:09 <PovAddict> "let's call it web 2.0! what will it be about?"

14:09 <PovAddict> everyone uses it to mean whatever they want it to mean, or for marketing -.-

14:09 <mtrg> i hate the term cloud computing

14:10 <PovAddict> just mentally replace "in the cloud" with "for suckers" when reading stuff

14:11 <mtrg> but what do suckers understand when they read "cloud computing"?

14:11 <mtrg> i happen to deal with many suckers and understanding them is good. they happent pay me.

14:11 <yoyo[RKN]> "cloud computing" is just for marketing. It only says there is something in the net.

14:12 <mtrg> what a shitty way to categorize things

14:12 <wdsmia> asks mtrg to please watch his or her languageĀ… Thanks, your friendly channel op.

14:12 <mtrg> k

14:12 <PovAddict> as I understand it, there are two different and orthogonal meanings

14:12 <PovAddict> one is from the user perspective

14:12 <PovAddict> instead of having data (or even the application itself) on your computer, it's on some server that you access through the internet

14:13 <PovAddict> for example using Google Docs instead of Microsoft Office

14:13 <PovAddict> the other meaning is on the server side

14:14 <PovAddict> where you don't have specific servers with specific roles that you can identify

14:14 <PovAddict> you just have a big amount of servers with no single point of failure

14:14 <mtrg> yeah.. but so is grid computing with utility computing, right?

14:15 <PovAddict> and as an admin or app developer, you don't know or care in what specific server your stuff is running

14:15 <PovAddict> mtrg: utility computing is usually implemented that way

14:15 <PovAddict> mtrg: for example, you don't get access to a fileserver; you just get "storage" in some unspecified place

14:15 <PovAddict> your data may get moved to another set of disks

14:16 <PovAddict> a disk may die and the system automatically replicates the data elsewhere

14:16 <mtrg> can cloud computing not be distributed

14:16 <PovAddict> Someone Else(tm) may move things around and replace hardware

14:16 <PovAddict> and all that is transparent, you're never aware of individual servers

14:16 <mtrg> i'm not aware of inidvidual workstations in grid computing either

14:16 <PovAddict> Google is *very* much engineered that way, developers inside Google don't even know where their code is running, they just deploy it into the network

14:17 <mtrg> i think  historically the term "grid computing" was meant to be "as easy and on-demand access just like power grids"

14:17 <PovAddict> mtrg: I dunno, in the first meaning (from the user perspective), one could argue a FTP server on a single physical server means "your data is in the cloud" :P

14:21 <mtrg> generally, what are the components of almost all grid computing implementations. i suspect: scheduler, middleware

14:22 <PovAddict> just try to solve your goal... trying to understand the real meaning of vague terms like "grid computing" seems like a wild-goose chase

14:23 <Tank_Master> what are you refering to as middleware?

14:23 <Tank_Master> for boinc, theres only ther server and the host, nothing inbetween

14:24 <PovAddict> Tank_Master: yet some people say BOINC *is* middleware

14:25 <Tank_Master> id ask "in the middle of what?" then

14:25 <mtrg> Tank_Master: stuff like gLite http://en.wikipedia.org/wiki/GLite

14:25 <Romulus> Title: gLite - Wikipedia, the free encyclopedia (at en.wikipedia.org)

14:25 <PovAddict> in the middle of project applications and the raw computer?

14:25 <mtrg> boinc seems to be a "middleware" by wikipedia

14:27 <Tank_Master> meh, i dunno

14:29 <mtrg> as far as i understand it, a middleware's job is making the grid transparent to its developer/user. so that they submit a job to the middleware, which in turn takes care of distributing it as a black box, and then return the final answer to the developer/user

14:29 <mtrg> i'm just almost guessing though. i could be wrong.

14:37 *** Aeternus has joined #boinc

14:37 *** Aeternus has quit IRC

14:37 *** Aeternus has joined #boinc

14:41 *** microchip_ has quit IRC

15:08 <desti_T2> http://i.imgur.com/TmoH3.gif

15:09 <PovAddict> ouch

15:44 <mtrg> interesting how the cat is perfectly ok

16:19 *** yoyo[RKN] has quit IRC

17:15 *** whynot has joined #boinc

17:32 *** RBecker has quit IRC

17:32 *** RBecker has joined #boinc

17:40 *** MarcoFe has quit IRC

18:19 *** whynot has quit IRC

18:22 *** PovAddict has quit IRC

18:23 *** pppingme has quit IRC

18:23 *** nonMOOSE has quit IRC

18:24 *** nonMOOSE has joined #boinc

18:24 *** Aeternus has quit IRC

18:25 *** PovAddict has joined #boinc

18:27 *** pppingme has joined #boinc

19:08 *** RBecker has quit IRC

19:10 *** RBecker has joined #boinc

19:50 <valer> gLite appears to have middleware between the developer and the "cloud" whereas BOINC developers typically would control the server/scheduler themselves, requiring no such middleware.

19:51 <valer> Unless you call the BOINC server middleware. ;P

19:52 <valer> Like httpd is middleware for your PHP programs...

19:55 <mtrg> do you have any document that explains these basics about middlewares archetictures?

19:55 <mtrg> its hard for me to pick the right middleware

19:58 <desti_T2> it's best to pick no middleware

19:59 <valer> It's a general term meaning almost anything.

19:59 <PovAddict> you're not in a neutral channel

19:59 <PovAddict> in my opinion, and probably everyone else here, use BOINC unless you really find you can't :P

20:02 <valer> One definition of a grid is a loose set of heterogenous and widely geographically distributed hosts (that can perform some task in a parallel manner). A cluster on the other hand is set of similar hosts located in the same facility.

20:02 <valer> So grid computing is anything using a grid.

20:08 <mtrg> valer: what is a grid

20:08 <valer> I'm getting an AI bot feeling here.

20:09 <mtrg> i'm serious, you can send me a captcha challenge and i'll slove it for you

20:09 <valer> Hehe

20:09 <desti_T2> http://www.news.com.au/breaking-news/world/no-talking-to-aliens-at-the-bbc/story-e6frfkui-1226513392145

20:09 <Romulus> <http://tinyurl.com/b6tgmeb> (at www.news.com.au)

20:10 <mtrg> valer: what is a grid then? (other than vertical/horizontal lines)

20:10 <PovAddict> mtrg: why do you care about those vague terms? just go and build your thing

20:10 <mtrg> i want to understand what valer said. he said "grid computing is anything using a grid."

20:11 <valer> In the previous line I defined a grid.

20:12 <mtrg> thus line 2 was redundant (i thought that may be you had a better reason)

20:29 <valer> It comes to my mind that some "future BOINC" could use a P2P network to serve and schedule data. Dividing and sub-dividing the data into work units might also happen in the grid. Thus the main server would just inject the raw data and receive the results.

20:30 <mtrg> is there any more basic grid implemenmtation? i'm thinking to study the traffic if boinc via wireshark

20:30 <PovAddict> why?

20:31 <mtrg> learning how they work from a network point of view to gain confidence in the things i'm installed in 100s of computers

20:31 <PovAddict> BOINC sends XML over HTTP; the format isn't documented, and I don't see what you gain from knowing it

20:31 <PovAddict> have you already installed BOINC and attached to a public project to watch it work?

20:31 <mtrg> installing

20:32 <mtrg> btw, is SETI@home, FOLDING@home, are all using BOINC with the difference being only type of problems they solve (i.e. aliens vs. protient folding)

20:33 <PovAddict> SETI@Home uses BOINC, and so do another hundred projects

20:33 <PovAddict> Folding@Home doesn't use BOINC; it has its own software

20:33 <mtrg> interesting..

20:34 <mtrg> any particular language that i should code in to let my problems distributed via BOINC?

20:34 <PovAddict> preferably something that compiles into native code and can call functions in a C library

20:35 <mtrg> C?

20:35 <Romulus> hmm... C is for Cookie :), mtrg

20:35 <PovAddict> such as C :P

20:36 <mtrg> is there any general use public BOINC server that allows me to push my tasks into?

20:37 <PovAddict> no, that would be insecure as hell

20:38 <mtrg> does it mean that my boinc client will compile the C/C++ code into my native machine's language, and run it as the user where boinc is running as?

20:40 <mtrg> seems to be something neat. boinc is the server, and boinc_client is the client, right?

20:40 <mtrg> i wonder if the default boundle of boinc is set to work with any particular project, or if the default config is fairly secure such that it doesn't do anything

20:50 <mtrg> why do i need to give boinc manager my email address when joining (Say) SETI@home?

20:54 <RBecker> That's how it registers the account

20:57 <PovAddict> mtrg: the client downloads compiled binaries from the server

20:58 *** efc has joined #boinc

21:05 <valer> Individual users can compile their own binaries but it has some problems: you need to trust that the hosts produce correct results.

21:08 *** Caterpillar has quit IRC

21:16 *** NudgeyNR|2 has quit IRC

21:16 <valer> On my system, boinc_client is the core client and boinc is a wrapper script for it.

21:16 *** NudgeyNR has joined #boinc

21:51 <valer> I find the premise of sudoku@vtaiwan very funny.

22:05 *** Psychoholic has quit IRC

22:39 *** efc has quit IRC

Generated by irclog2html.py 2.4 by Marius Gedminas - find it at mg.pov.lt!