To generalize this, has anyone looked at how the system requirements (disk usage, memory, i/o rate, network usage, ... , etc) of machines running gmetad need to scale as the number of nodes increase. More specifically, has anyone come up with a (rough) algorithm to estimate system requirements for a top level gmetad monitoring a hierarchical grid (many clusters reporting to ultimately one collection point) configuration considering for example scalability flags? -john
On Mon, 9 Feb 2004, Federico Sacerdoti wrote: > Yes, this is the case. The problem is you dont know how many nodes you > will have. What if a large number are added? Everyone's RRD disk > requirements will go up. > > For Rocks clusters, we allocate a Linux tmpfs with a maximum size of > 25% available memory. Only the "frontend" node (not a compute node) > keeps the RRD databases. > > -Federico > > On Feb 5, 2004, at 6:30 PM, Brooks Davis wrote: > > > Does anyone know how to go about estimating ganglia disk requirements > > assuming one knows the number of nodes and the number of metrics per > > node? It looks like we use about 12k per-metric per-node plus some > > summary info. Is this alwasy the case? > > > > I'm thinking about this because I'm looking at adding support to the > > FreeBSD startup script for more or less automatic handling of memory > > files system backed storage of data since the disk IO load of writing > > directly to disk cripples the performance of even machines with > > hardware > > raid. To do that, I'll need a good way to estimate disk usage based on > > some parameters the users can figure out easily. > > > > -- Brooks > > > > -- > > Any statement of the form "X is the one, true Y" is FALSE. > > PGP fingerprint 655D 519C 26A7 82E7 2529 9BF0 5D8E 8BE9 F238 1AD4 > > > Federico > > Rocks Cluster Group, San Diego Supercomputing Center, CA > > > > ------------------------------------------------------- > The SF.Net email is sponsored by EclipseCon 2004 > Premiere Conference on Open Tools Development and Integration > See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. > http://www.eclipsecon.org/osdn > _______________________________________________ > Ganglia-developers mailing list > Ganglia-developers@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/ganglia-developers > > > --------------------------------- John Hicks - HPCC Engineer TransPAC, Indiana University [EMAIL PROTECTED], 317-278-1083 "Let the wild rumpus start!" - MS
