DepSpid
Es darf gevotet werden:
Currently, a DepSpid task consists of two phases:
Phase One (0-50%): This phase mainly uses internet network resources and doesn't use much CPU time (in fact, everything higher then a few minutes during this phase is unusual).
Phase Two (50-100%): This phase is CPU-intensive and doesn't need an internet connection.
The current design of the spider application ties both phases together. However, this is going to change in the near future.
The goal is to have two applications instead of one, meaning one application for each phase. This has multiple advantages:
1) The BOINC client will be able to handle resources much better. Currently it has to handle the second phase like being non-cpu-intensive which can cause much trouble.
2) If the task fails during phase two, the work done during phase one is currently lost, too. With two separate applications the second phase could be reprocessed or even being calculated multiple times to ensure the correctness.
3) The current way must allow long cpu times for the whole task. That's why it's difficult to abort tasks using too much CPU time during phase one. With a network only task, the maximum CPU time can be set very limited to avoid problems.
4) The cpu-intensive application used for phase two could be used for backend processing tasks, too. These tasks don't exist yet but will soon have to be created for further data processing. (The calculation of dependencies between single pages within one domain will be almost equal to the calculation between whole domains.)
5) The cpu-intensive part would be easier to be ported to Linux, too. However, there should be Linux applications for both parts in the (far) future.
6) Having a part without network access will give users with low internet connection speeds or users which are not allowed to run a networking project (like at work) to participate in the project, too.
The current way BOINC is designed doesn't allow a project to send non-cpu-intensive and cpu-intensive tasks together. That's why a second URL for attaching the client will be required (at least until BOINC changes one day). There will be one URL for non-cpu-intensive tasks and one URL for cpu-intensive tasks. This is just a way for the client to distinguish between the two types. User accounts, web-pages, stats etc will stay singular. You will just have the choice to attach your clients to the cpu-intensive part, the network-intensive part or both parts.
I'm currently trying to find out if there will be enough participants for both parts. Please use the vote to tell me what you're going to do once there are two project parts.
Leider wieder eins aufgegeben.Sun, 26 Oct 2008 --- Project closed
During the last two years, the project has made some big steps forward. However, it seems that every step forward has lead two steps backward. The list of things that need to be done has grown immense during this time. Ongoing technical problems combined with organizational and personal reasons have left only one conclusion: To close the project. It's currently completly unsure if or when the project may be continued (in whatever way). Please detach your clients from the project. The domain is going to stay alife at least for a while but forums will be closed soon. Sorry for the bad news. This was the only remaining option. It was a hard decision but it had to be made.