Filesystem using Oracle and mirroring VG ?

Hello,

I have a filesystem which has Oracle EBS 11.5.9 running and database 9i

# df -g

Filesystem    GB blocks      Free %Used    Iused %Iused Mounted on

/dev/fslv00      270.00     69.68   75%   746263     5% /oratec

which is in oraclevg

# lsvg oraclevg
VOLUME GROUP:       oraclevg                 VG IDENTIFIER:  00c82b6e00004c0000000133f8a8c64c
VG STATE:           active                   PP SIZE:        256 megabyte(s)
VG PERMISSION:      read/write               TOTAL PPs:      1092 (279552 megabytes)
MAX LVs:            256                      FREE PPs:       11 (2816 megabytes)
LVs:                2                        USED PPs:       1081 (276736 megabytes)
OPEN LVs:           2                        QUORUM:         2
TOTAL PVs:          2                        VG DESCRIPTORS: 3
STALE PVs:          0                        STALE PPs:      0
ACTIVE PVs:         2                        AUTO ON:        yes
MAX PPs per VG:     32512                                    0
MAX PPs per PV:     1016                     MAX PVs:        32
LTG size (Dynamic): 256 kilobyte(s)          AUTO SYNC:      no
HOT SPARE:          no                       BB POLICY:      relocatable

Now if I mirror oraclevg, will the oracle EBS run slow after mirroring of oraclevg ?

anyone having any similar experience.

depending on the location of the second storage (if it is a second storage) the response time may increase
that's not easy to say

create a unmirrored filesystem, and run a tool like iozone on it
then mirror the vg, and run it again with the same parameters

AIX 5L Open Source Packages | Main / IOzone

don't forget to turn off the filesystem cache

@funksen

I looked at IOzone :
Ozone is a filesystem benchmark tool. The benchmark generates and measures a variety of file operations. IOzone has been ported to many machines a nd runs under many operating systems.
IOzone is useful for performing a broad filesystem analysis of a vendors computer platform. The benchmark tests file I/O performance for the following operations: Read, write, re-read, re-write, read backwards, read strided, fread, fwrite, random read, pread, mmap, aio_read, aio_write.

so what would the benchmark indicate, i mean how can we read it. If it indicates high I/O then after mirroring it would be even higher ? right ?

filosophizer, it is impossible to teach someone the job over the net. funksen has given you some useful pointers already but if you lack the basics to even understand what he is telling you we can't help you. How to benchmark I/O is learned by understanding how storage technology works, how Unix kernels work, how disks are accessed, by which strategies relational databases (in your case Oracle) optimize disk access and probably some other areas of competence you seriously lack.

To learn all this funksen has probably read several books, a lot of documents plus had some years of experience. How is he supposed to transfer this accumulated amount of knowledge to you here in a few articles? This is impossible.

It is not ill will from our side, but there is probably only one way for you: learn starting from the basics, the same way as he did. Over time you will arrive at the same point as funksen, but neither without effort nor in short time. It took him some years of experience and it will take you the same time to know what he knows.

You come across like a first grader, eager to solve math problems involving advanced calculus and now expect the teacher to tell you how to solve it. In order for you to understand his explanation he would have to explain so many things which in turn will need explanation too, etc., ad infinitum, that you are probably best off pursuing a sound education in systems administration. This, alas, is beyond our scope here.

To finally answer your question:

No, not right. It depends. It depends on so many things you can only find out by first benchmarking the system and then interpreting the values produced by these benchmarks that an answer is simply impossible.

It is not even clear your system is really I/O-bound as you claim: how have you analyzed that and by applying which methods did you arrive at this conclusion? Have you run filemon? vmstat? iostat? What is the OS level (different AIX versions need different tuning parameters)? What is the contents of "/etc/tunables/lastboot"? Which size is the SGA? What is the cache hit/miss statistics in Oracle?

The answers to all these (and a lot more) questions are (or could be) factors if it comes to the I/O-performance of a database. Given the info you have presented until now we could as well toss a coin and answer "friday".

bakunin

Bakunin is right, this depends on many things
large sequential ios would suffer less from a mirrored disk than a lot of small random ios

you didn't even say what kind of storage it is, local scsi, iscsi, san attached
I'll give you a short answer, because I'm short in time today

if your system is important enough to mirror the volume groups for data safety and to prevent a downtime due to a storage outage, then go for it

I do it this way:
cluster: mirror rootvg, mirror datavgs (mostly over two locations)
non cluster system that boots from san: no mirror at all
non cluster system that boots from local hdisk: mirror rootvg

if I would be using iscsi disks, I would mirror some of the more critical non cluster systems as well, but that's because our ip network is not as stable as our san

no, you must take a look at increased latency, and decreased overall data throughput in bytes/second