| Title: | LSM |
| Moderator: | SMURF::SHIDERLY |
| Created: | Mon Jan 17 1994 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 803 |
| Total number of notes: | 2852 |
Hi,
We are currently working on a 3 node TruCluster configuration. Following
is the diagram.
----------------------
| Memory Channel Hub |
----------------------
/ / \
------------------- / ----------------
| / |
-------- -------- ------------ --------
| 8400 | | 8400 | | SW800 | | 4100 |
| | | | | | | |
|KZPSA |------- |KZPSA |------- | HSZ40(2) |--------|KZPSA | scsi bus 1
| | | | | | | |
|KZPSA |--------|KZPSA |--------| HSZ40(2) |--------|KZPSA | scsi bus 2
| | | | | | | |
|KZPSA |--------|KZPSA |--------| HSZ40(2) |--------|KZPSA | scsi bus 3
| | | | | | | |
-------- -------- ------------ --------
NODE A NODE B NODE C
| \ /
\ \ /
-----------\ --------- ---------------
---------- | |
| | |
------------------
| Memory Channel |
------------------
Node A will run application A and Node C will run Application c. Node B will
failover node incase either Node A or Node B fails. Applications A & B will
be using Oracle. We will be using raw devices for the database.
The configuration calls for no point of failure, thus the 2 memory
channel hubs.
To implement redundancy on the SCSI buses, we will be using LSM to mirror
disks across the buses; such that a LSM mirror volume may contain a RAID5 disks
located on scsi bus 1 and scsi bus 2; another volume may contain RAID5 disks
from scsi bus 2 and scsi bus 3 and so on. By the way, we will be implementing
HW RAID5 on the HSZ40 level.
Questions are :
1. What limitations are there in doing LSM (mirror) across HSZ40 disks ?
Such that how many gigabytes of data can be LSM mirrored ?
2. Would there be any problems with performance on doing LSM(mirror) ?
3. Is there a better way of implementing this based on the customer
requirement of no point of failure configuration ?
Looking forward to your replies.
Thank you.
/diana
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 760.1 | KITCHE::schott | Eric R. Schott USG Product Management | Fri Feb 28 1997 10:18 | 42 | |
> >To implement redundancy on the SCSI buses, we will be using LSM to mirror >disks across the buses such that a LSM mirror volume may contain a RAID5 disks >located on scsi bus 1 and scsi bus 2; another volume may contain RAID5 disks >from scsi bus 2 and scsi bus 3 and so on. By the way, we will be implementing >HW RAID5 on the HSZ40 level. > >Questions are : >1. What limitations are there in doing LSM (mirror) across HSZ40 disks ? > Such that how many gigabytes of data can be LSM mirrored ? You may want to run BCL in LSM to ensure mirror re-sync times are short...if you do this, ensure the HSZ WBC is enabled on the BCL HSZ devices > >2. Would there be any problems with performance on doing LSM(mirror) ? There should not be other than the above concern. > >3. Is there a better way of implementing this based on the customer > requirement of no point of failure configuration ? This is probably ok for now...some things to keep in mind: Depending on your I/O load, you may not want to be using RAID5 on the HSZ, as the LSM mirroring has already protected the data. The concern here is performance as RAID5 write performance and RAID1 write performance combined will be slower than just RAID1. In the future, you could go to an all hw raid configuration once pathing is available in Digital UNIX, and clusters...but this is out in CY98. > >Looking forward to your replies. > >Thank you. > >/diana | |||||