If I get it right you backup the 4 groups at the same time and they are on the
same (super sized) disk.
is that a raid5 by any chance?
if so your paralell datastreams may interfere on the disk level.
any read action of 1 stream moves all disk arms on raid3 or 5
4 streams may dramatically reduce throuhput on the disk side.
I expect that the oracle backups run alone ?
I adressed this issue by defining my 1Tbyte disk box as 5 independant raid1
volumes (but my data sets allow that). I donot need 1 large volume (yet).
Maarten
On Wednesday 01 March 2006 09:27, Gaddy wrote:
> Hi NetWorkers,
>
> in our installation we run both online and offline
> backups of our Notes (V6.5.2) servers.
> While online backups work very well (most of the time)
> (main reason for doing this is: parallelism)
> we are facing some performance issues with the offline
> < plain file system > backup of these servers.
>
> All notes data resides in one - quite large - filesystem,
> directory structure is almost flat.
>
> For ~450GB Notes file system it takes more than 15 hrs to finish the
> offline backup, which is much more it should.
>
> Environment:
> all data on CX700, FC-2Gbps attached to client and server,
> details: Meta-Lun of 2x RAID-5 Luns = 2x [4+1]
> LTO-2 FC-2Gbps attached to server
> 1 Gbit Ethernet
> no CPU shortages on client and server
> client compression is OFF
> everything is on Win2k+3
>
> Fore even more details see savegroup completion below.
>
> To speed up the backup we decided to parallelize the file system backup
> manually. 'save -M' would be the best choice, but does not realy
> support restores (BTW, who cares about restores :-)
> So we configured 4 groups with 4 different directives,
> each directive skips some portion of the filesystem,
> ie:
> << E: >>
> +null: [a-k]*.nsf [r-z]*.nsf <-- backup l*.nsf ... t*.nsf
> +null: ... more file types here
>
> This works well but has not improved backup speed significantly.
> So I'm a bit clueless now:
> - servers have enough resources
> - all pipes (LAN, SAN) are fat enough
> - drives (disk, tape) should have enough power
> but still this is very slow :-(
> (BTW, no problem of doing Oracle backups >50MB/sec in the same environment)
>
> Just before I start to rack "my" brain,
> I'd like to ask for some hints where to start the investigation first.
> How are you doing backups of large filesystems?
> (sure we can do LANfree, Snapshot, etc. but have decided to speed up
> the filesystem backup 1st)
>
> Tips are much appreciated.
>
> Thanks -sg-
> --
> Steffen Gattert; VISIOPLANT Hamburg
>
> -------- savegroup completion msg of one group ------
> pls. note the backups of "E:\" on x01 & x02 & x03
> backup of "E:\" is about 25% of whole file system
> 4 groups backing up "E:\" on x01 & x02 in parallel
> 1st: [a-f]*.nsf, ...
> 2nd: [g-k]*.nsf, ...
> 3rd: [l-r]*.nsf, ...
> 4th: [s-z]*.nsf, ...
> with 5 min. timeshift between each group
>
> NetWorker savegroup: (notice) Lotus_Notes_Server_A_F completed
> Start time: Fri Feb 17 00:33:03 2006
> End time: Sat Feb 18 06:48:09 2006
> --- Successful Save Sets ---
> xxxxxxxx01: C:\ level=full, 1426 MB 01:49:19 17770 files
> xxxxxxxx01: D:\ level=full, 4238 MB 02:01:31 689 files
> xxxxxxxx01: E:\ level=full, 141 GB 16:04:51 10370 files <--!!!
> xxxxxxxx01: SYSTEM_STATE:\ level=full, 13 MB 00:00:21 17 files
> xxxxxxxx01: SYSTEM_DB:\ level=full, 874 KB 00:00:13 11 files
> xxxxxxxx01: SYSTEM_FILES:\ level=full, 227 MB 00:09:20 1848 files
> index:xxxxxxxx01 level=full, 32 MB 00:00:33 164 files
> xxxxxxxx02: C:\ level=full, 2468 MB 03:46:28 18397 files
> xxxxxxxx02: D:\ level=full, 4844 MB 02:20:18 1137 files
> xxxxxxxx02: E:\ level=full, 124 GB 22:19:15 5128 files <--!!!
> xxxxxxxx02: SYSTEM_STATE:\ level=full, 13 MB 00:02:02 15 files
> xxxxxxxx02: SYSTEM_DB:\ level=full, 895 KB 00:02:23 11 files
> xxxxxxxx02: SYSTEM_FILES:\ level=full, 226 MB 00:10:50 1841 files
> index:xxxxxxxx02 level=full, 57 MB 00:00:25 166 files
> xxxxxxxx03: C:\ level=full, 1389 MB 00:52:05 15325 files
> xxxxxxxx03: D:\ level=full, 8792 MB 03:11:47 1314 files
> xxxxxxxx03: E:\ level=full, 118 GB 22:13:57 4532 files <--!!!
> xxxxxxxx03: SYSTEM_STATE:\ level=full, 12 MB 00:02:15 15 files
> xxxxxxxx03: SYSTEM_DB:\ level=full, 889 KB 00:02:06 11 files
> xxxxxxxx03: SYSTEM_FILES:\ level=full, 226 MB 00:11:00 1800 files
> index:xxxxxxxx03 level=full, 49 MB 00:00:24 122 files
> xxxxxxxx04: C:\ level=full, 746 MB 00:42:11 11025 files
> xxxxxxxx04: D:\ level=full, 917 MB 00:32:32 9819 files
> xxxxxxxx04: E:\ level=full, 18 GB 04:40:51 1755 files
> xxxxxxxx04: SYSTEM_STATE:\ level=full, 12 MB 00:02:11 15 files
> xxxxxxxx04: SYSTEM_DB:\ level=full, 877 KB 00:02:06 11 files
> xxxxxxxx04: SYSTEM_FILES:\ level=full, 226 MB 00:10:36 1820 files
> index:xxxxxxxx04 level=full, 11 MB 00:00:10 96 files
>
> To sign off this list, send email to listserv AT listserv.temple DOT edu and
> type
> "signoff networker" in the body of the email. Please write to
> networker-request AT listserv.temple DOT edu if you have any problems wit this
> list. You can access the archives at
> http://listserv.temple.edu/archives/networker.html or via RSS at
> http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
--
Maarten Boot,
Compuware Europe B.V.
Hoogoorddreef 5
1101 BA Amsterdam
Tel: +31 20 312 6511
To sign off this list, send email to listserv AT listserv.temple DOT edu and
type "signoff networker" in the
body of the email. Please write to networker-request AT listserv.temple DOT edu
if you have any problems
wit this list. You can access the archives at
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
|