Fast Track to Solaris 10 Adoption: ZFS Technology
Functionality & Usability Issues
Please click on a question below or download a pdf version.
- Can you have separate pools per system?
- I'm still confused. Suppose I have three E450s. Would ZFS allow me to integrate storage across all three boxes into one big "poor man's SAN"?
- Will a GUI be available for the administrative functions?
- Can disks be designated as spares to be used only in case of disk failure, or is it like RAID5 where a disk failure results in no data loss, but lessened performance?
- I hope this is relevant. Using SVM on the Solaris 9 OS, all disk objects are named with digits. Will we be able to use text labels in the Solaris 10 OS?
- Can users see into the snapshots, and can they copy files out to perform their own "restores"?
- Are there caps on how large a filesystem can grow? I would hate to see one filesystem run away and not only fill itself, but also eat all the other disk space connected to the system.
- Can I use the command vmstat to analyze performance of the disk in the Solaris 10 OS?
- In a shared storage type environment how would I move a ZFS volume from one host to another? Like exporting a Veritas disk group?
- Will ZFS have a backup command like ufsdump?
- Is there a GUI to manage the storage pool?
- Are there any limitations or special considerations concerning ZFS and x86 fdisk partitions? Does ZFS exist in an x86 fdisk partition so it is possible to have ZFS, NTFS, FAT, and ext3 filesystems all on the same disk?
- What degree of control does an administrator have over how the pool is allocated? For example, can I direct certain filesystems to stay on certain hardware devices?
- What happens when I need to grow/shrink a ZFS?
- Where is the storage pool managed? At the host? On the arrays? Somewhere in-between?
 |
Q: Can you have separate pools per system?
A: Yes, you can have multiple pools attached to one host. Typically though, you'd only need one.
 Back to top
Q: I'm still confused. Suppose I have three E450s. Would ZFS allow me to integrate storage across all three boxes into one big "poor man's SAN"?
A: No, ZFS is a local filesystem (for the time being). To access storage attached to a different host, use NFS.
Back to top
Q: Will a GUI be available for the administrative functions?
A: GUI will not be available in the initial release. It will be available later. But administration with ZFS is so simple that all one needs to do is state the intent and ZFS will take care of the rest.
Back to top
Q: Can disks be designated as spares to be used only in case of disk failure, or is it like RAID5 where a disk failure results in no data loss, but lessened performance?
A: with ZFS, you'll be able to designate "hot space," which is reserved for use in the case of a disk failure. Rather than leaving an entire disk unused, thus reducing performance, ZFS reserves a little bit of each disk in the pool.
Back to top
Q: I hope this is relevant. Using SVM on the Solaris 9 OS, all disk objects are named with digits. Will we be able to use text labels in the Solaris 10 OS?
A: It is on the roadmap.
Back to top
Q: Can users see into the snapshots, and can they copy files out to perform their own "restores"?
A: Yes.
Back to top
Q: Are there caps on how large a filesystem can grow? I would hate to see one filesystem run away and not only fill itself, but also eat all the other disk space connected to the system.
A: Short answer: yes. Each filesystem has a hard cap, which can be dynamically changed by the administrator.
Back to top
Q: Can I use the command vmstat to analyze performance of the disk in the Solaris 10 OS?
A: Yes; in fact with DTrace in the Solaris 10 OS, you can do more advanced analysis than with vmstat.
Back to top
Q: In a shared storage type environment how would I move a ZFS volume from one host to another? Like exporting a Veritas disk group?
A: You can use HA Storage+ to enable automatic fail-over. Additionally, you can easily export the pool from one host and import it to another without re-cabling if it is attached to multiple hosts (e.g., in a SAN).
Back to top
Q: Will ZFS have a backup command like ufsdump?
A: You'd back up ZFS at a higher level, with any POSIX-compliant backup utility.
Back to top
Q: Is there a GUI to manage the storage pool?
A: Not at this time, but it's simple enough that there doesn't appear to be a need for it.
Back to top
Q: Are there any limitations or special considerations concerning ZFS and x86 fdisk partitions? Does ZFS exist in an x86 fdisk partition so it is possible to have ZFS, NTFS, FAT, and ext3 filesystems all on the same disk?
A: ZFS will recognize any style partitions that the Solaris OS does. That said, we greatly prefer EFI labels because they support large disks and partitions and are much better designed than most other disk labeling schemes.
Back to top
Q: What degree of control does an administrator have over how the pool is allocated? For example, can I direct certain filesystems to stay on certain hardware devices?
A: With the ZFS pooled storage model, all filesystems in a pool draw storage from all disks in the pool. This way, ZFS can make optimal use of all available disk bandwidth.
Back to top
Q: What happens when I need to grow/shrink a ZFS?
A: It's extremely simple: issue a single command that changes the size of the filesystem.
Back to top
Q: Where is the storage pool managed? At the host? On the arrays? Somewhere in-between?
A: The storage pool is managed in the Solaris OS, using simple, straightforward command-line tools.
| 
