static-void
asked on
Redhat GFS redundancy
Hi there
Im looking into RedHat GFS as a filesystem for a research paper. I was wondering if anyone knew how redundancy is implemented in GFS? Ive been doing some reading up on it but all i can find is that journeling can do something to do with recoery of data and that when a node fails the data from the node is fenced so it doesnt get accesses until the node comes back online. Im really interested in if it provides some kind of mirroring or block based raid across the storage nodes to maintain avaliability of all parts of the filesystem in the event of a failure. Does anyone know if this is possible. If not how does GFS do its high avaliablity configurations and how can it allow fulfilment of a SLA?
Thanks
Luke
Im looking into RedHat GFS as a filesystem for a research paper. I was wondering if anyone knew how redundancy is implemented in GFS? Ive been doing some reading up on it but all i can find is that journeling can do something to do with recoery of data and that when a node fails the data from the node is fenced so it doesnt get accesses until the node comes back online. Im really interested in if it provides some kind of mirroring or block based raid across the storage nodes to maintain avaliability of all parts of the filesystem in the event of a failure. Does anyone know if this is possible. If not how does GFS do its high avaliablity configurations and how can it allow fulfilment of a SLA?
Thanks
Luke
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
So basically what your saying is that in order to have redundancy it needs to be implemented at the block level IE use GFS arch over top of SAN arch and provide all failover and redundancy at that low level not at a high level GFS redundancy. The reason im asking is that we are looking at deploying a HPC cluster for research purposes at the university im at, this cluster is implemented using comodity hardware and we would like to implement a clustered filesystem running across the local drives running in the cluster nodes for storage. Were currently looking at writing some extra code for GlusterFS to improve performance and redundancy across the cluster for this type of application. Im doing some reasearch into what this would involve, and im looking at how other systems do redundancy as GlusterFS seems to have the same issues with redundancy where it relies on redundancy at the device level which on comodity hardware is obviously out of the question.
If you know of any other systems which would implement a clustered file system across a number of nodes and which do do some level of redundancy i would really like to hear about it :)
Thanks
Luke