Martijn
09-15-2006, 01:11 PM
I send WCG an email a few days ago, on the optimized client issue. Let me quote the mail over here:
Hello,
XtremeSystems has recently left Rosetta@Home and joined the WCG project. We left from Rosetta because we were called cheaters. The local staff did nothing about it. There had been many flame wars in which we participated. We later realized that it had been a wrong movement, as a lot of people from the community joined against us. I want to say something very important over here: we definitely aren't cheaters, we are just a bunch of dedicated, crazy overclockers. We care about the science being done, and that is why we chose WCG as our next project.
As we were moving on to this project, people from Rosetta contacted others at WCG, telling them about 'why we were cheaters'. Yet again a flame war has started at your forums, in the general chat area. We decided not to react to any flames in the thread, though. We are a dedicated team and we won't get ourselves being called cheaters again, as is happening now in this thread (read from page 3 on):
http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=8594&offset=0
Because we are so dedicated, we are using the optimized cruncher files. It modifies the benchmark used by the boinc client (by using the SSE/SSE2 instruction sets, depending on what CPU you are running). In our opinion it justifies the difference between Intel and AMD systems (as BOINC prefers AMD cpus). If you consider this cheating, we are willing not to use them, as long as there is a clear statement at your forums (or anywhere else on your site) where we can read this. If you do not want us (and anyone else) to use them, make this clear and we will delete them straight away.
Our goal is to get to the first place in your project, and we don't want to be called cheaters on the way getting there. As for now, there has never ever been any project in which we weren't able to get to first place when we wanted to. We have the firepower to do that. Our greatest member still has to get to this project (he had a break after leaving Rosetta), and he had an average output higher than that of the team in second place. At the moment, we are having an output of about 100000 credits a day (boinc credit). We used to have a steady output of more than 500000 credits per day at Rosetta (yes, that's half a million). This means, that if we can get to full throttle, which we definitely will if we stay, we can achieve 4 million WCG credits per day. We are glad to help your project, but if we don't know whether we are allowed to use optimised clients and are being called cheaters, we can still leave your project, as we haven't yet done much work. When we left Rosetta, we were begged by the project administrator to please come back. We didn't. Please don't make us decide to leave to another project again.
We really like your project and the science being done, but if we can't crunch because of members calling us cheaters and nothing is done about it, I'm afraid to say we'll probably leave. I therefore ask you to take appropriate action by either removing the thread, banning the members that are flaming and by officially stating whether optimized clients are allowed. If you want more info on them, please contact me or any other members of the team.
You don't want to lose this team, as what happened at Rosetta. You have not yet seen nothing of what this team is capable of. I and every other member of my team hopes that you will take appropriate actions.
Sincerely,
Martijn Kruit from XS (XS_Martijn)
And I got a reply!! :woot:
Martijn,
Movieman has also contacted us and I will copy to you what I sent to him
below. Let me know if you have additional questions. We have responded in
the forums as well and the issue appears to have died down.
I have to ask - you plan to become #1? Easynews is an awfully big team :-)
You would make my year if you guys could overtake Easynews.
Let me know if you have additional questions beyond what is below.
"First, I would like to thank you and your team for joining World Community
Grid. You have added a lot of power to our research and we sincerely
appreciate it.
Second - as I believe you and your team appear to agree, we believe that
distributed computing is about the science. The competition is fun, sparks
enthusiasm and helps maintain interest but it is secondary to the science.
Having said that let me explain something about World Community Grid that
you may already know, but is necessary for what I will write below. World
Community Grid does not run any science of its own. On our website we
state: "World Community Grid is making technology available only to public
and not-for-profit organizations to use in humanitarian research that might
otherwise not be completed due to the high cost of the computer
infrastructure required in the absence of a public grid. As part of our
commitment to advancing human welfare, all results will be in the public
domain and made public to the global research community. " We accept
applications from scientists who wish to run their research on this grid.
You can read more about this process here:
http://www.worldcommunitygrid.org/projects_showcase/viewSubmitAProposal.do
Once a project is selected we help modify the science application to be
grid enabled, do security audits of the applications (and maintain the
versions we run in our own source code libraries to ensure only changes
that we authorize/approve get implemented), conduct testing of the
application and develop processes for putting work on the grid and
delivering results back to the scientists. We also are the only ones who
have access to the servers that distribute the work and run the website.
In essence, we are here to make sure that the members can trust that the
science being run is valid and safe. We are also here to be a resource
that the scientific community can tap to run research that would otherwise
be infeasible for them to perform.
I mention all of this because one of the things we won't be doing is
working to fully optimize each science application to take advantage of the
latest performance optimizations on new chips. The reason that we won't be
doing this is practical. Because of our mission, many of the projects we
run will have a limited amount of work. The time it takes to fully
optimize the application will not be proportional to the length of time the
project will run. This is very different then on something like Seti@Home
where they run a couple of applications for many years. In that case it
does make sense to spend the time to get every last bit of performance out
of the processors and deal with the complexities of validation from
different platforms. Additionally, the time it would take us to fully
optimize and test the science application on the different chip
architectures is time that we don't spend working on adding the next
project. We have chosen to make sure that we keep a steady stream of new
projects coming into World Community Grid so that we have work for our
members to run when our current projects end. This stream of new projects
coming on-line and other projects finishing is something that is different
about us from other DC projects and it impacts how we approach things.
One of the things that we have been able to do is that we use homogenous
redundancy at the operating system level (i.e. workunits are only sent to
the same OS (except for Mac where we have to differentiate between Intel
and PPC)). At this level and with the amount of chip optimization we have
done all results return are identical. This allows us to have high
confidence in the results and reduces the complexity of testing which in
turn allows us to board more projects and run more research.
We are very interested in making sure that members are contributing
effectively - even if we are not able to fully optimize the applications
for their computer. For example, one thing we have done is that we have
done some work to increase the efficiency of the BOINC workunit
distribution system. You have come from Rosetta@Home where they not using
redundancy (they can confirm results on their own servers - we don't have
that luxury) so I'm not sure if you are aware of this or not. However,
projects such as Seti@Home and Einstein@Home send out four workunits and
then attempt to validate when they get three back. We have spent some time
changing the way the system works so that we only send out three and then
validate when we get three back. We had to do some changes to ensure that
workunits would still be validated within a timely basis (all workunits are
validated within in a few days and all are validated within two weeks).
This change increased the efficiency of the grid by about 23% (I say this
to make sure that you understand that we do value our members computer
time). We have returned this code back to the BOINC community.
Now - how this relates to the optimized clients. We have no intention of
limiting what client anyone uses provided that it returns valid science
data. We also think that trying to define a 'fair' amount of credit is
very tricky and it can create ideological discussions that are difficult to
resolve. We have been using the 'stock' BOINC system so far and it has
served good enough. However, once large disparities between the claimed
credits for different hosts in the same workunit appear then there becomes
a need to make sure the credit is assigned fairly between different
workunits. This is what we have now started to do. We are discussing what
changes to make internally now. Additionally some of us will be at the
BOINC conference in Geneva next week and we will discuss with some of the
other project leads how they have responded to this. Once we have made due
consideration and spent some time examining what impact this will have we
will announce the changes, if any. Please feel free to add any input you
have especially since you have experienced this issue on other projects.
Our goal is to make the system both actually fair and perceived as fair.
thanks,
Kevin"
Thank you for participating.
Hello,
XtremeSystems has recently left Rosetta@Home and joined the WCG project. We left from Rosetta because we were called cheaters. The local staff did nothing about it. There had been many flame wars in which we participated. We later realized that it had been a wrong movement, as a lot of people from the community joined against us. I want to say something very important over here: we definitely aren't cheaters, we are just a bunch of dedicated, crazy overclockers. We care about the science being done, and that is why we chose WCG as our next project.
As we were moving on to this project, people from Rosetta contacted others at WCG, telling them about 'why we were cheaters'. Yet again a flame war has started at your forums, in the general chat area. We decided not to react to any flames in the thread, though. We are a dedicated team and we won't get ourselves being called cheaters again, as is happening now in this thread (read from page 3 on):
http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=8594&offset=0
Because we are so dedicated, we are using the optimized cruncher files. It modifies the benchmark used by the boinc client (by using the SSE/SSE2 instruction sets, depending on what CPU you are running). In our opinion it justifies the difference between Intel and AMD systems (as BOINC prefers AMD cpus). If you consider this cheating, we are willing not to use them, as long as there is a clear statement at your forums (or anywhere else on your site) where we can read this. If you do not want us (and anyone else) to use them, make this clear and we will delete them straight away.
Our goal is to get to the first place in your project, and we don't want to be called cheaters on the way getting there. As for now, there has never ever been any project in which we weren't able to get to first place when we wanted to. We have the firepower to do that. Our greatest member still has to get to this project (he had a break after leaving Rosetta), and he had an average output higher than that of the team in second place. At the moment, we are having an output of about 100000 credits a day (boinc credit). We used to have a steady output of more than 500000 credits per day at Rosetta (yes, that's half a million). This means, that if we can get to full throttle, which we definitely will if we stay, we can achieve 4 million WCG credits per day. We are glad to help your project, but if we don't know whether we are allowed to use optimised clients and are being called cheaters, we can still leave your project, as we haven't yet done much work. When we left Rosetta, we were begged by the project administrator to please come back. We didn't. Please don't make us decide to leave to another project again.
We really like your project and the science being done, but if we can't crunch because of members calling us cheaters and nothing is done about it, I'm afraid to say we'll probably leave. I therefore ask you to take appropriate action by either removing the thread, banning the members that are flaming and by officially stating whether optimized clients are allowed. If you want more info on them, please contact me or any other members of the team.
You don't want to lose this team, as what happened at Rosetta. You have not yet seen nothing of what this team is capable of. I and every other member of my team hopes that you will take appropriate actions.
Sincerely,
Martijn Kruit from XS (XS_Martijn)
And I got a reply!! :woot:
Martijn,
Movieman has also contacted us and I will copy to you what I sent to him
below. Let me know if you have additional questions. We have responded in
the forums as well and the issue appears to have died down.
I have to ask - you plan to become #1? Easynews is an awfully big team :-)
You would make my year if you guys could overtake Easynews.
Let me know if you have additional questions beyond what is below.
"First, I would like to thank you and your team for joining World Community
Grid. You have added a lot of power to our research and we sincerely
appreciate it.
Second - as I believe you and your team appear to agree, we believe that
distributed computing is about the science. The competition is fun, sparks
enthusiasm and helps maintain interest but it is secondary to the science.
Having said that let me explain something about World Community Grid that
you may already know, but is necessary for what I will write below. World
Community Grid does not run any science of its own. On our website we
state: "World Community Grid is making technology available only to public
and not-for-profit organizations to use in humanitarian research that might
otherwise not be completed due to the high cost of the computer
infrastructure required in the absence of a public grid. As part of our
commitment to advancing human welfare, all results will be in the public
domain and made public to the global research community. " We accept
applications from scientists who wish to run their research on this grid.
You can read more about this process here:
http://www.worldcommunitygrid.org/projects_showcase/viewSubmitAProposal.do
Once a project is selected we help modify the science application to be
grid enabled, do security audits of the applications (and maintain the
versions we run in our own source code libraries to ensure only changes
that we authorize/approve get implemented), conduct testing of the
application and develop processes for putting work on the grid and
delivering results back to the scientists. We also are the only ones who
have access to the servers that distribute the work and run the website.
In essence, we are here to make sure that the members can trust that the
science being run is valid and safe. We are also here to be a resource
that the scientific community can tap to run research that would otherwise
be infeasible for them to perform.
I mention all of this because one of the things we won't be doing is
working to fully optimize each science application to take advantage of the
latest performance optimizations on new chips. The reason that we won't be
doing this is practical. Because of our mission, many of the projects we
run will have a limited amount of work. The time it takes to fully
optimize the application will not be proportional to the length of time the
project will run. This is very different then on something like Seti@Home
where they run a couple of applications for many years. In that case it
does make sense to spend the time to get every last bit of performance out
of the processors and deal with the complexities of validation from
different platforms. Additionally, the time it would take us to fully
optimize and test the science application on the different chip
architectures is time that we don't spend working on adding the next
project. We have chosen to make sure that we keep a steady stream of new
projects coming into World Community Grid so that we have work for our
members to run when our current projects end. This stream of new projects
coming on-line and other projects finishing is something that is different
about us from other DC projects and it impacts how we approach things.
One of the things that we have been able to do is that we use homogenous
redundancy at the operating system level (i.e. workunits are only sent to
the same OS (except for Mac where we have to differentiate between Intel
and PPC)). At this level and with the amount of chip optimization we have
done all results return are identical. This allows us to have high
confidence in the results and reduces the complexity of testing which in
turn allows us to board more projects and run more research.
We are very interested in making sure that members are contributing
effectively - even if we are not able to fully optimize the applications
for their computer. For example, one thing we have done is that we have
done some work to increase the efficiency of the BOINC workunit
distribution system. You have come from Rosetta@Home where they not using
redundancy (they can confirm results on their own servers - we don't have
that luxury) so I'm not sure if you are aware of this or not. However,
projects such as Seti@Home and Einstein@Home send out four workunits and
then attempt to validate when they get three back. We have spent some time
changing the way the system works so that we only send out three and then
validate when we get three back. We had to do some changes to ensure that
workunits would still be validated within a timely basis (all workunits are
validated within in a few days and all are validated within two weeks).
This change increased the efficiency of the grid by about 23% (I say this
to make sure that you understand that we do value our members computer
time). We have returned this code back to the BOINC community.
Now - how this relates to the optimized clients. We have no intention of
limiting what client anyone uses provided that it returns valid science
data. We also think that trying to define a 'fair' amount of credit is
very tricky and it can create ideological discussions that are difficult to
resolve. We have been using the 'stock' BOINC system so far and it has
served good enough. However, once large disparities between the claimed
credits for different hosts in the same workunit appear then there becomes
a need to make sure the credit is assigned fairly between different
workunits. This is what we have now started to do. We are discussing what
changes to make internally now. Additionally some of us will be at the
BOINC conference in Geneva next week and we will discuss with some of the
other project leads how they have responded to this. Once we have made due
consideration and spent some time examining what impact this will have we
will announce the changes, if any. Please feel free to add any input you
have especially since you have experienced this issue on other projects.
Our goal is to make the system both actually fair and perceived as fair.
thanks,
Kevin"
Thank you for participating.