Submission + - Scalable, Resilient Storage and Benchmarking
Qoroite writes: I will shortly be tasked with specifying, testing, and implementing a Scalable, Highly Available storage system for a hosting company. The new system should provide for realtime redundancy across Locations, and provide access to multiple linux systems via something like NFS.
The system would be expected to provide storage for several thousand websites, home directories, and email mailboxes.
Can Slashdot suggest any possible solutions, what I ought to be looking out for, how should I benchmark, and accurately interpret the results. What to we think about Commercial solutions vs Open Source. The company would probably go for commercial option, so long as the costs are reasonable. Either way, the solution would need to be proven reliable and well trsted.
Would a NAS (NFS) fronted iSCSI SAN be appropriate, with the NAS bit running Linux-HA to supply redundancy.
The host is currently hosting with NFS on a 15k SCSI Raid5 system, using hardware raid, and LVM on Linux. They are reaching a point where this setup is no longer sufficient, the filesystem has already been split once into two arrays, however it appears that an IOPS limitation is still being hit during busy periors.
A design goal for the new system is that it can be extended to add capacity and performance, and ideally that management and integration of existing and new storage nodes shoud not become an adminstrative nightmare.
The system would be expected to provide storage for several thousand websites, home directories, and email mailboxes.
Can Slashdot suggest any possible solutions, what I ought to be looking out for, how should I benchmark, and accurately interpret the results. What to we think about Commercial solutions vs Open Source. The company would probably go for commercial option, so long as the costs are reasonable. Either way, the solution would need to be proven reliable and well trsted.
Would a NAS (NFS) fronted iSCSI SAN be appropriate, with the NAS bit running Linux-HA to supply redundancy.
The host is currently hosting with NFS on a 15k SCSI Raid5 system, using hardware raid, and LVM on Linux. They are reaching a point where this setup is no longer sufficient, the filesystem has already been split once into two arrays, however it appears that an IOPS limitation is still being hit during busy periors.
A design goal for the new system is that it can be extended to add capacity and performance, and ideally that management and integration of existing and new storage nodes shoud not become an adminstrative nightmare.
