Home
| Calendar
| Mail Lists
| List Archives
| Desktop SIG
| Hardware Hacking SIG
Wiki | Flickr | PicasaWeb | Video | Maps & Directions | Installfests | Keysignings Linux Cafe | Meeting Notes | Linux Links | Bling | About BLU |
>> There are a few types of clusters..... >> >> 1) Where all the machines run the same code >> 2) where the job is divided across multiple machines >> 3) mix and match combinations of the above two >> > > In any case there are some things that you should learn regardless of > the type of cluster when dealing with multiple *nix machines. You'll > need to learn something like Puppet or Chef to manage configurations, > installed packages, and have these things backed up by revision > control to manage changes on the system. You'll need to learn how to > deal with NIS or LDAP for centralizing users across the computers. > Although you can use Puppet/Chef to do that as well but NIS or LDAP is > a better solution. > > As for how to go about doing this. I would just set up a bunch of > virtual machines to learn the ropes of these things. I wouldn't actually look at *any* of the applications in your list. First thing I'd do is learn ssh/sshd/agent setup. If you are dealing with a cluster, that's your first priority. Second, if it is a "compute" cluster, where a job is divided across multiple computers, figure out what mechanism they use, i.e. MPI, LAM, PVM, etc. If it is a redandancy cluster, find out which mechanism is used for distribution, i.e. nginx, LVS, etc. Know the mechanism by which the cluster operates as a "cluster." Know the mechanisms by which they are managed. > -- > David >
BLU is a member of BostonUserGroups | |
We also thank MIT for the use of their facilities. |