| [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] |
On 8:03:24 pm 2004-11-29 Richard Prescott <richard@xxxxxxxxx> wrote: > Hello, > > I am a newbie in OGFS so forgive my ignorance. Although, berfore > posting I read almost everything I could found on the subject but I > didn't understand everything. > > I planning to do a (production) setup moving our existing NFS/rsync to > something a little bit more appropriate. I have two requirements : I wouldn't consider OGFS in a production situation at this point (although Red Hat's GFS is worth considering since it's being actively maintained). GNBD isn't really a good idea in a production environment either. > > 1) replacing the actual rsync setup. This is the purpose of GFS so no > big deal here. I might setup a something involving FC or using a > software RAID over an network drive I don't know yet. FC will work for you. Unfortunately software raid doesn't get along well with OGFS/GFS (well, clusters really since metadata updates from nodes don't get updated on all the nodes) > > 2) replacing the actual NFS service. The fun begin here. All our > workstation act more or less as part of a (SPICE) calculus farm. I > want to move forward in that direction. A SPICE simulation as we run > it produce a lot of information. Close to 'cat /dev/random > > ~/sim/result.raw'. So any networked fs will struggle on the > bandwidth. The only way I see to deal with this issue is to optimized > bandwidth utilisation by : > a) 'spice ... -o /dev/stdout | gzip -c > ~/sim/result.raw.gz' > b) caching write operation to local disk > c) caching read operation to local disk If you go with FC gear, you probably won't have that much problem with bandwidth (unless you have really slow disks) > > As I understand GNDB, caching isn't there yet. > > Am I wrong ? > > Is anybody know an alternative that could do it? > > Is anybody interrested in coding it? > > Note: by caching write operation, it is possible to schedule updates > uppons what are requested elsewhere. That is where the gain is. Might want to look at lustre. I don't suggest coda or intermezzo. (intermezzo was recently removed from the kernel, and to see coda go wouldn't surprise me much) > > Thanks for your time. > > Richard Prescott ------------------------------------------------------- SF email is sponsored by - The IT Product Guide Read honest & candid reviews on hundreds of IT Products from real users. Discover which products truly live up to the hype. Start reading now. http://productguide.itmanagersjournal.com/ _______________________________________________ Opengfs-users mailing list Opengfs-users@xxxxxxxxxxxxxxxxxxxxx https://lists.sourceforge.net/lists/listinfo/opengfs-users