One interesting/cool management feature that I would love to see is to be able to view the stats for one host across all projects... ie, what contribution is a given CPU making across all of my projects? It'd be useful since sometimes remote machines running BOINC crash, and cool to see how well a work/home computer does across a variety of projects. I'm not sure if it's possible, but it should be, right?

This one is answered in the F.A.Q.:
[url=http://www.boincstats.com/forum/viewtopic.php?t=203' target='_blank]http://www.boincstats.com/forum/viewtopic.php?t=203[/url]

Thanks! Sorry for missing that in the FAQ.

Great work with the site!!!

Would it be possible to just have a list of hosts x projects and their credit day/week/month? In other words, the page would show each host for each project all on one page and what the host has been up to lately.

It might be of a lot less usefulness, but it solves the problem of the lack of database support for having one host across all projects.

Just a thought. I love the site and keep up the great work!

DT
[url=http://www.davidcthompson.com' target='_blank]http://www.davidcthompson.com[/url]

<span style='font-size:14pt;line-height:100%'>(edit: formatting)</span>

Dave, I don't understand what you mean I think.

There is no CPID. I cannot create a page for a host in all projects. If I could, I could also make stats for this.

Please clarify what you mean.

Sorry, I think my comment was a little unclear. Again, thanks a lot for all your hard work on this and I think the site is GREAT!

My idea was to simply list all of the data you have about each host in each project on one screen. So, let's say that I have three computers ("A", "B", and "C&quot each of which works on three projects ("1", "2", and "3&quot. There is no way to match up the computers, I agree with that. So, instead, the page just lists all of the data you do have about all the hosts. The page would display:



The site wouldn't "know" that comptuer "A" is the same across all the projects, so it would just list it independently each time.

Hmmm. I don't think that's a great explanation either, but maybe it makes a little more sense. The idea would just be to put all host information on one page so that a user could look and intuit whether there are some hosts doing better or worse than others. The site can't match hosts across projects, but a user probably could.

This data is available becuase all it does is take the "See all hosts for this user" information that's available across several projects and put it onto one page.

In terms of implementation, I'd select all projects for the user and then select all hosts from that user, regardless of what project. I don't know what your database looks like, but I'd imagine that the query would resemble:



The result should be a list of all host infos across all project that a user is a part of. I'm sure your DB isn't shaped exactly that way, but that's the idea.

edit: corrected psuedo-SQL to add project_name so that we can have the name of the project displayed

Maybe I'm the only one interested in this. Thanks again for all your hard work and let me know if there's anything I can ever do to help out!

@Willy:

I do have my hopes, regarding host cpid - or at least I got them as I noticed an empty cpid field in the xml - so maybe that will come at a later time

That'd be great if it ever happened. Might that also solve some of the problems with merging hosts?

I hope the BOINC people do implement it.

The next server version of BOINC will support this. But I don't know if my server will be able to cope with this. SETI hosts already takes over 30 minutes to process (from XML to sorted stats with history). It will be a challenge to do this while not killing access to the site.

Hmm. Yes I see your point, especially if it implemented in the incremental updates... In the end it shouldn't differ that much from individual cpid, when it comes to the coding, but host xml are just so much bigger than user xml

Are there any optimizations that could be squezed out of the code to cut the 30 min down? Is most of the slowdown in the DB work or the processing work? I'm happy to help in a general sense to see where the work is being done and try to optimize further.

Bottleneck is the I/O to the harddisk. It's a single SCSI U320 harddisk. I'm already planning for a server upgrade (but have to get money for that first) and that will contain at least SCSI-RAID (if possible RAID-5) and an AMD X2.

Memory is not a problem (3G, nor should the CPU (PIV 3,4GHz) be a problem if it keeps it head cool enough.

The code is as optimized as possible.

I'm thinking about pruning the history of users/teams/hosts that are not accesses/looked at by anyone for over 2~4 weeks. This will save a lot of time as the history is a big part of the DB.

Sounds like you're on top of things! Keep up the great work!

There is a host CPID in the client state.xml file. If it is not getting into the stats.xml file, that is a different problem. For the machine I am typing on, I have an entry in the client_state.xml file that reads as follows: <host_cpid>xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx</host_cpid> where I have removed the actual cpid (just in case it is there is an internal and an external CPID). I was under the impression that this should be making it into the stats xml file for the hosts. <!--emo&-->

A lot (but not all) of the hosts now have a CPID in the exported files.

But the CPID is not the only problem. The BOINCstats server has to make the stats, and that's a tough job. SETI hosts stats takes over 30 minutes, just imagine what combined hosts stats would take.

Something has to give then, like dropping the history for hosts (saves a lot of time).

Or donations should flow in, and I could upgrade the server with more CPU power and a RAID configuration

There is a host CPID generated on the newer clients. The surprise is that it is not being exported in the hosts stats XML yet.