This needs to be *primarily* handled in BOINC scheduler/feeder itself. BOINC knows how much RAM, what CPU the PC has etc.

I just flet that "click the start button below, wait 30 seconds and we'll see how well suited your PC is for Rosetta" sounded a lot better than "Just go to another website, download software that appears unrelated to Rosetta, install it, and then have it determine what's possible and what's not, and maybe Rosetta won't work for you after all".

I'm all for the targetted WUs ideas. But they aren't here this week, so they aren't viable for the oncoming BBCers. I'd also point out that altering the type of WUs sent results in numerous questions from observant participants. The more predictable and consistent the better. If anything changes, ideally it would be by user request in preferences changes.

The British are coming! The British are coming!

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

since CASP is starting soon, and many of the proteins will be larger, we wanted to do some calculatoins on a broader range of sizes. before pursuing this much further, we need some way of ensuring that these jobs are only sent out to machines appropriate for them, which is difficult with the current BOINC setup; we hope Rom can help us with this.

If time to tweak BOINC is just too short, perhaps it'd be best to create a new project (a la CPDN's Seasonal Attribution project, clearly asking for hosts with 1GB RAM or more) for the needs of CASP and upcoming vaccine runs.

---

Best UFO Resources
Wikipedia R@h
How-To: Join Distributed Computing projects that benefit humanity

The British are coming! The British are coming!

I can assure you that a lot of non-British people, from all over the world, watch / listen /read BBC.

I know for a fact that many people in Greece do (and were sad that greek language section of BBC http://www.bbc.co.uk/greek/ closed down a few weeks ago)

So, this BBC thing could be really big. I'm a bit suprised though that the previous article from LiveScience, picked up by MSNBC, FoxNews and Yahoo had little effect on TeraFLOPS. If we had the Web-stats for the homepage, we might be able to tell if those LS/MSNBC/FoxNews/Yahoo/etc visitors came to learn more but found the joining process to hard or something.

It also gives you an indication of how hard it is to attract users of different demographic groups, because the LiveScience article was very compelling IMO, making direct reference to R curing cancer.

Yet, overall project TeraFLOPS are back at 20-22 where they were back in January...

---

Best UFO Resources
Wikipedia R@h
How-To: Join Distributed Computing projects that benefit humanity

One of my machines is a 500mhz with 250 Ram and does just fine. It rarely has an error. It's OS is Windows ME.

Both these boxes crunch for rosetta

Box 1

Box 2

Box 2 is slower and I aborted the last wu so that it wouldn't time out because I was testing the box with another project. Both crunch multiple projects.

---

Here are some of my boxes too: they are well under the requs, but they never fail and always report on time. But then of course, they are dedicated to rosetta 100% (headless linux machines, CLI only).

box1
box2
box3
box4

Hope theses boxes will remain usable... maybe a "small box" flag in the preferences?

-Trib'

---

I would assume a typical 'BBC watcher' would not be building headless crunchers or having dedicated crunching machines.
The specs have to be aimes at normal home usrs who use their computers for the internet, office and maybe the odd game. Not for a dedicated power cruncher ;-)

Hence there is no point in pointing out 'dedicated' crunchers. Yes it can run on lower spec, but then so can BBC-CCE. It's what is a feasable spec for the average home usrs. (and since it's a UK major site a typical UK PC Home user, which is general either dial-up... I think about half the connected population. Or they have low capped broadband.)
Hopefully this will change as companies get less stingy.

---

Team mauisun.org

I would assume a typical 'BBC watcher' would not be building headless crunchers or having dedicated crunching machines.
The specs have to be aimes at normal home usrs who use their computers for the internet, office and maybe the odd game. Not for a dedicated power cruncher ;-)

Hence there is no point in pointing out 'dedicated' crunchers. Yes it can run on lower spec, but then so can BBC-CCE. It's what is a feasable spec for the average home usrs. (and since it's a UK major site a typical UK PC Home user, which is general either dial-up... I think about half the connected population. Or they have low capped broadband.)
Hopefully this will change as companies get less stingy.

Don't forget is about a 30Mb download for initial setup (10MB BOINC, 10MB Sci-App, 10Mb helper files) then 2Mb for each job every 4Hrs crunched.
Of course that's if they have left their connection/computer on long enough to get the initial rosetta files and subsequent jobs.

---

Team mauisun.org

Comparing the specs from BBC climateprediction and Rosetta I doubt there will be any computers which failed cpdn-specs. RAM is 512 MB both (actually CPDN uses less than Rosetta) and traffic for Rosetta is higher than for CPDN (so Broadband required as well). The only advantage of Rosetta is that WUs are shorter. So a 1 GHZ P3 is still sufficient (given it has 512 MB RAM). Although one should keep in mind that the recent large WUs need several hours to reach the first checkpoint - not good for the average computer user who leaves the PC running only a few hours.

I fear as long as there is no "big WU" - Flag there is no machine below the BBC-Climate-Spec capable of running Rosetta. :-(

---

Comparing the specs from BBC climateprediction and Rosetta I doubt there will be any computers which failed cpdn-specs. RAM is 512 MB both (actually CPDN uses less than Rosetta) and traffic for Rosetta is higher than for CPDN (so Broadband required as well).

Don't give up so easily. Some of those that will consider R@H DO meet the BBC minimum specs, but they had other considerations that caused them not to want to crunch it. Like perhaps they only run their PC 8hrs a day and didn't want to wait for a year to complete a single WU (or didn't feel they'd finish before the deadline). Perhaps they tried the BBC project and found it conflicted too much with the operation of their PC. Or that it used the hard disk too much swapping. BBC needs 600MB of disk space, R@H is less than a third of that. And I believe the initial WU download is extremely large compared to Rosetta. THIS is likely the key reason for their broadband req. 30MB, in manageable pieces is very doable on dial up. And if you get them to set their WU preference to 24hrs, then they won't have to download so many WUs.

The big WU flag, and creating another project for big WUs, or small WUs are all fine ideas. But, what can we do in a week? I think these are beyond the scope. Even if they were attempted, they would be untested, and unproven. Not to mention the burden on the project staff to maintain it all.

I think we just need to focus on the existing project, with it's existing imperfections, and ask ourselves how to make it easy for SOME of the BBCers to come to R@H. And I think our best avenue to do that is to stick to the project guidelines (that's why they are there), and to make the ones that DO come happy that they did. This will payoff many months down the road as those happy participants share their experience with their friends, and further compelling results are produced from the project. We won't get them all, and to try and do so would potentially lead to negative experiences and word-of-mouth which harm the project many months down the road.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

I'm not at all trying to stop this btw ;), just saying be careful on what is said.

BBC CCE download wasn't so bad (see edit after the paragraph) from what I remember and since after that there is only uplads, though large at around 4 to 5 MB are infrequent and anyway my Rosetta uploads are at 2.5MB for only 8hrs work <eek>
Mind I can actually upload them on Rosetta, BBC CCE seems to have a problem with firewalls being itn the way and cannot acknowlege reciept of a task after upload.

Their broadband requirement comes from the BBC-CCE (BOINC) Client download (10Mb) and the 5Mb initial start (yes only 5Mb) and the 4 to 5 MB infrequent uploads
They ZIP all their files, send it, have it unZIPped at our end (which inflates to some 200Mb + in size directories, actual sci apps are only a couple of Mb insize) then the results are also ZIPped up and sent.
THIS IS SOMETHING WE HAVE BEEN TRYING TO TELL ROSETTA FOR QUITE SOME TIME!
Rosetta could cut the sci-app 4.98 down to 2.5Mb from 6.5Mb and maybe use it for better compression on the uploads.
As for the other files although the downlaod would be smaller, i.e. a recompression of the whole structure for the .gz it would take up more space once uncompressed as i'm not sure you can read from the .zip like you can the .gz, hence CPDN/BBC-CCE directory size. Though for uploads and th sci.exe it would work. but we're getting off topic and bug fixing first

Use .ZIP, it is afaik built into BOINC OR just ask CPDN how it is done.


Anyways,
So

- Give a landing page
- Breif easy to understand, no techy stuff (that can be linked to)
- easy step by step installation (maybe with a video ?) QMC very top of page/BBC CEE maybe good sites to get ideas from
- Give Req's (512Mb etc BROADBAND needed)
- Mention Windows/Linux (x86) and MacOSX (PPC) available
- FAQ Page
- ????

I assume rosetta site & forum will be upgraded to 5.5 soon since there has major reported problems at Ralph, maybe time to move the forum struce as well ?

---

Team mauisun.org

FWIW, about the installation instructions of the landing page, I've just added some formatting (nesting, screenshot of BOINCMgr, reference to R@h's FAQ etc) to the Wikipedia page:

http://en.wikipedia.org/wiki/Rosetta%40home#How_to_join

I didn't find the QMC installation instructions particularly clear, because they don't mention BOINC anywhere. The distinction grid-software vs science software should be somewhat clear to people.

PS: I'd like to add another section explaining the need for volunteer power (citing the 5.000.000 CPU-hours on BlueGene awarded to Rosetta for 2006) and another about the intellectual property and answering the Q's like "will someone make money (or build bio weapons) with my CPU time donation?".

Also, on the science front, some material about Rosetta's successes (in CASP or anywhere else) and references, publications etc. Wikipedia is supposed to be an encyclopedia afterall.

---

Best UFO Resources
Wikipedia R@h
How-To: Join Distributed Computing projects that benefit humanity

The thing that puts people off of CPDN most has nothing to do with downloads, uploads or bandwidth. The problem with CPDN is the level of dedication it requires. To complete a single model requires at least a month on a fast dedicated machine. Moreover, the models fail frequently, and do so after people have spend hundreds of hours running them. Most of the discussions on forums and Team sites from people who have left CPDN mention specifically the failing model wasting their CPU time as the reason they left the project.

Personally I left when they decided not to support Mac platforms. For me this was a matter of principal. I have a number Windows machines available to me and could apply them to any project, but I will not support a project that does not support the standard platforms supported by BOINC.

I think we can assume that people who sign up for BOINC projects have already demonstrated a level of commitment to the concept of DC. Part of that commitment is downloading the project and the WUs. While the landing page might mention the bandwidth required to load and run the project, I would not make a big deal out of that. If people connect and later find that the bandwidth requirement is beyond their level of commitment, then they have the option of disconnecting.

The Rosetta time setting if properly used can reduce the loading of WUs to a factor far less than other BOINC projects when measured over time.

I say invite the people in and let them decide if they want to stay. The big issue the project needs to worry about is fixing the last of the bugs, not compressing files. Most people will run the application if it will run without errors.

Despite my reservations about the lower specs necessary for Rosetta I agree with Feet1st that we should just concentrate on the current issue. Now would be a good time for a link on the BBC site since CPDM just announced _all_ (!) the models downloaded for the BBC experiment are faulty and need to be aborted (ouch). I'm sure there are quite a few people out who would rather try another project where faulty WU do exist as well but don't take 2 months to be discovered.

So the specs are:

Machine not older than 6 years + broadband.

Specifically:

Proc 600 MHz and above
512 MB Ram
Broadband or dial-up with long initial download (~30 MB) and daily traffic of up to 10 MB.

Then a link and that's it. What do we need to do?

---

What do we need to do?

That's the "big question". I believe the answer is that the project would like some help putting a project description in to lay-persons language. And if there's any analogies that can be made or tie-ins specific to the BBC project that would help folks already familiar with BBC to more readily understand what Rosetta is, and is not, then so much the better.

So far as simple explaination of the project, there's an entire thread on the subject. I'm not sure we struck on one yet that really hits home. I put together a physical demonstration, but I don't think it's as effective reading it, as it would be DOING it. Perhaps a video of that physical demonstraition would be useful?

So, one avenue of what we need to do is post simple 1 paragraph descriptions of what Rosetta is. The more the marrier, and we can vote on them (using the plus and minus signs under each post), and the project team will adjust them as they need to (we lay-people sometimes have a way of misstating things in ways that would actually misrepresent the project's goals).

Another avenue is the line of questions about any interactions with the BBC application that us user's might be aware of, that the Rosetta folks would not be familiar with.

I think the input on bandwidth is useful. I'm not clear how we'll resolve that one (a week doesn't afford rolling out enhanced compression and some other ideas that have been brought up elsewhere). I think perhaps saying broadband is recommended, but dial-up is possible too, and then layout the specific download sizes people should expect, and when (initial, vs. each Rosetta update, vs. day-to-day WU exchange).

Tralala can you post a link to description of the WU problems they're having?

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

What do we need to do?

... Another avenue is the line of questions about any interactions with the BBC application that us user's might be aware of, that the Rosetta folks would not be familiar with.

I have run Rosetta/Ralph and CPDN on the same machines for some time. There really is no interaction between the BOINC CPDN and Either Rosetta or Ralph. In fact they may be more compatible in some ways than other projects. CPDN runs about 15 min between checkpoints and it is recommended to keep application in memory set to yes, just as with Rosetta. The largest difference is that CPDN stores a lot of data on the users machine. All the trickles and model outputs are kept and must be manually removed by the user. That data storage requirement can become quite large. In any case application swapping and other elements of the BOINC environment work well between the projects. The only issue is that the BBC CPDN is not using a standard BOINC package.

I think the input on bandwidth is useful. I'm not clear how we'll resolve that one (a week doesn't afford rolling out enhanced compression and some other ideas that have been brought up elsewhere). I think perhaps saying broadband is recommended, but dial-up is possible too, and then layout the specific download sizes people should expect, and when (initial, vs. each Rosetta update, vs. day-to-day WU exchange).

This seems a reasonable approach. Might I also suggest that I can prepare a short list based on updated versions of the FAQs, and you could place a link to those on your "Welcome" page. These could be focused on issues such as getting started, running the project, and finding help if they need it. I would be happy to coordinate this with The Rosetta System administrator if you decide to go that way.

I am not certain why the Work Unit downloads shock people so much. On the systems I have observed the compressed total for a Work Unit averages just under 3 MB. While this may take a few minuets on a modem, it is certainly not a major problem unless people are running large queue sizes, or short time settings. With the new user adjustable time setting, running a large queue will not be necessary.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact

I am not certain why the Work Unit downloads shock people so much. On the systems I have observed the compressed total for a Work Unit averages just under 3 MB. While this may take a few minuets on a modem,..

2.5Mb results (these largescale_full atom tasks) takes about 20mins of uploading and afaik not resumable, yet to test that. Similar sized downloads are of course quicker, but not by that much (I have a perfect dial-up connection connect straight-ish though one of the fastest backbones in Briton). The biggest problem is that it pretty much kills off your web browsing as BOINC doesn't transfer only in idle bandwidth. I know you can set max up/downspeeds but by default these are wide open. Not somthing your average joe home user would like.
Though as soon as they find a way to make the 200/400_v1.3.gz files sticky and make it use that for all subsequnet jobs till it's no longer needed (or even keep a good bunch of them sticky) then we'll probably be laughing as they are the killers. That'll be similar to what Einstien does.
For instance I've just sent in a job then obtained two new jobs
The downloading time for them both from start to finsh at around 3300bytes/sec, dedicated purley to that task (as surfing is v.painful hence the length of this post :-D ) too a couple of seconds under 30mins.
Anyways going of topic. Nuff said it's a boradband only project, except for the dedicated ;-) Though there's no way I can run my other computers on here as well.


We are not talking about seasoned DCers here, the TV shows are on the BBC and generally aimed at the average home users.

Anyways on with attracting them :-)

Interaction between CPDN & Rosetta is fine (though we're not talking about that)
Interaction between BBC CCE & Rosetta is fine (from a BOINCers point of view, i.e. me & you)
Interation between BBC CCE & Rosetta is a mess (from a BBC CCE users point of view)

We'll need something like
Q) I already have BBC CCE installed on my computer, how can I join Rosetta
A) You'll need to 'detach' from BBC CCE as per their instructions (link to that in BBC CCE FAQ), then uninstall the BBC CCE program. Then you can follow the instructions as normal.
Note: you can then attach back to BBC CCE if you want (instruct to use bbc.cpdn.org instead of rosetta.bakerlab.org when following rosetta attachment and use email and password used there)



sorry about the QMC thing I was thinking of something different (like the top part was a 1 2 3 ... to installing, my bad) miss remembered.

P.S. Yeah the dodgy work units is unfortunate, though to most users will be transparent as they send a kill and get another unit. Well untill they look at the progress in the screen saver ;-)

---

Team mauisun.org

We are not talking about seasoned DCers here, the TV shows are on the BBC and generally aimed at the average home users.

Amen to that.

Please allow me to be "devil's advocate" for a moment.

The issue of bringing more TeraFLOPS to Rosetta@home has been discussed for at least 1.5 month now (arbitrarily picking the date of DB's 1st post to the science journal). There've been several threads with ideas, a mailing list, a Wikipedia page, 1st page feature in UW's homepage and some very compelling coverage in major news sites, e.g.

LiveScience
picked up and published on:
Yahoo! News: Armchair Scientists Can Work on Medical Breakthroughs
MSNBC
FoxNews

Yet, despite a big increase in # hosts which has exceeded 110.000, we're still at "only" 22 TeraFLOPS, roughly where we were in Jan-06 (and I don't know how these TFLOPS # is calculated, if it comes from BOINC benchmarks it could be highly overstated).

Some food for thought IMHO.

---

Best UFO Resources
Wikipedia R@h
How-To: Join Distributed Computing projects that benefit humanity

"devil's advocate"...we're still at "only" 22 TeraFLOPS

So, taking what you say optimistically, you are saying we aren't likely to retain many BBCers. Maybe 1% of their current active host count of 76,000 and of their body count of 21,000 hosts no longer crunching BBC, and of how many viewers that just weren't compelled by a climate change experiment?

Growth is exponential. Every user you get today leads you to growth of another 3-5 over the next 2 years. So, while 1,000 new active hosts from BBC may only SOUND like 2 days worth of additions, it's a lot more than that. A lot of the "new hosts" counted on any given day are people detaching and attaching again, or moving to a new PC etc. They are also existing DCers and they only allocate 10-20% resource share to Rosetta.

So, 1,000 from BBC is real new hosts, and 1,000 new hosts represents 2% of the current 50,000 active Rosetta hosts, but possibly 10% more TFLOPS.

If we get 20% of the BBC hosts, well, then it MIGHT be a 40% bump in hosts, and a 100% bump in Rosetta TFLOPS, just for putting together some BBC-friendly materials. And the better we make those materials, the higher that conversion percentage will be.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

Feet1st, in your calculations of "real" new hosts from BBC, keep in mind that the news articles on LiveScience/Yahoo/MSNBC/FoxNews were published on 5/6-Apr-2006. That was 10 days ago. Any "instant adoption" effect would be evident during the first 3 days (from publication).

Frankly I'm a bit suprised about the (lack of) effect of the news coverage on overall TeraFLOPS.

What makes you think that BBC effect will be vastly superior over LiveScience/Yahoo/MSNBC/FoxNews ?

To the rest you said about DCers, I agree.

---

Best UFO Resources
Wikipedia R@h
How-To: Join Distributed Computing projects that benefit humanity

I'm a bit suprised about the (lack of) effect of the news coverage on overall TeraFLOPS.

What makes you think that BBC effect will be vastly superior over LiveScience/Yahoo/MSNBC/FoxNews ?

Well, to my knowledge none of those news agencies specifically target the computer audience. We all saw the news on the internet, and so we have a computer somewhere (although perhaps at work where it's not ours to devote to a DC project). The BBC host count definately has a computer, and internet, and an interest in DC projects. So, to some extent I think the audience there is an even better fit with the R@H target audience.

I have no clue if the BBC effect will be superior. But I believe those news articles probably only get read for a week or so before they are "old news". The BBC relationship will last for an extended time. And perhaps lead to some air-time on BBC television.

My point, and please try to take this if nothing else, is that it MIGHT be a really big thing! And even if it is not, by preparing for it, we've enhanced our materials and experience with recruiting more crunchers.

You mentioned the ideas threads and the Wikipedia, etc. But keep in mind that the Wikipedia changes just came out in the past 2 weeks, and very few (if any) of the other ideas have seen the light of day at this point. They all have merit and all will help expand the audience and the host count.

I was the membership chair of a local club. I devised probably 20 different ways to recruit more people in to the club. When I saw new people showing up, I'd ask them "so, how did you hear about the club" (market research, AND conversation starter). The net result was that no one thing brought on the 30% increase in membership... but ALL of them combined did so. Some of the ideas showed ZERO results... until the next year! "...Well, I heard about it a year ago and did nothing..."

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/