Brerrabbit,
Thanks for replying.
> What kind of connectivity details can you relay?
Primary backup copies go to a Quantum DXI7500 with 22TB of raw storage. It's a
dual controller array with two fully-populated 16-slot disk enclosures. Not
sure if the disks are SATA or FC, I suspect SATA.
I average about 220MBPS when backing up to this unit, the throughput fluctuates
between 200-300MBPS when monitoring in realtime from the Networker console and
the DXI console which reports the realtime throughput to/from each virtual tape
drive and for the whole VTL.
Pretty satisfied with this throughput and it's what Quantum said to expect,
there is a throughput anaylsis tool in the DXI7500 management console which
does a 5-minute throughput test and reports the unit is capable of 380MBPS
write speeds and over 600MBPS read speeds, although the test doesn't make clear
if this is for deduplicated or un-deduplicated data. To be honest I think the
results from this tool are pretty optimistic but the unit is clearly capable of
ingesting 300MBPS of writes and maybe more if I had more storage nodes writing
to it concurrently.
I also have a Quantum i500 with three LTO4 drives (HP) to clone from the
DXI7500.
Backup software is Networker 7.4 SP2.
Networker server is Windows 2003 SP2 and eight storage nodes, all Windows 2003
SP2 except for one Solaris 8 server.
All fibre-channel connected (Brocade switches) all ports on all devices are
running at 4-Gbps.
The Networker server, SNs, DXI7500 and i500 library are connected across two
switches which are separate fabrics no ISL between them.
I have half the storage nodes in one Brocade switch with two ports from DXI7500
+ two LTO4 drives, and the other storage nodes in second switch with two ports
from DXI7500 + 1 LTO4 drive.
My script kicks off 3 instances of nsrclone to try to use all 3 physical LTO4
drives. I do all the cloning from the Networker server (to two LTO4 drives) +
one of the Windows storage nodes in the other fabric (1 LTO4 drive) One issue
with my script is I don't get equal amounts of data in each clone job, I cloned
about 2TB on Thursday which took 5 hours but one of my clone jobs contained
about 100GB of data and the remaining 1900GB was split nearly evenly across the
other two clone jobs so only using two LTO4 tape drives for most of the
cloning. I am separating out the SSIDs per volume so I don't get two clone jobs
trying to use the same volume but I regularly seem to get two clone jobs with
over 90% of the data to be cloned and the third clone job with a very small
proportion, which is a bit odd as the same number of volumes are assigned to
each clone job.
Anyway I really need to improve the speed of cloning, obviously I need to work
more on my script to equalise the clone data across all 3 nsrclone instances
but even when only using two tape drives I average about 45MBPS. Watching the
Networker and DXI consoles during cloning I see bursts up to 120MBPS and never
see speeds below 30MBPS, however there are 10-second periods where the
throughput drops to 0 then fires up again, the DXI7500 reports the virtual tape
drive status as moving but the same tape remains in the drive, I think this is
the DXI7500 moving the virtual tape to the correct position for the next
saveset whihc I thought t would be pretty instant with a VTL but waiting for
Quantum to confirm this and what the moving status on the virtual tape drive
actually means.
The DXI7500 allows a maximum of 40 virtual tape drives so I have presented 16
to the Networker server, 16 to the other Windows SN doing cloning and one each
to the other six storage nodes.
> I'm cloning from various models of EMC Disk Libraries to a Quantum i2k with
> LTO3 drives and see sustained speeds between 100 MB/s and 120 MB/s, but then
> sometimes I see 10-15 MB/s.
Yes I guess your lower speeds are due to the limits on your lower performing
arrays?
I am no Networker expert (mainly used Commvault, Netbackup, and TSM) and it's
the first time I have used Networker cloning so I was wondering if there's any
specific performance tweaks to improve cloning throughput and any known
limitations with Networker cloning?
Basically it looks to me as if my hardware is capable of sustaining 100MBPS
plus but I am getting delays from Networker in between each clone of a saveset
which drags down my average throughput and I wondered if others saw the same
delays with regular 10-second periods with 0-MBPS throughput.
If I can average 70MBPS or thereabouts across all 3 tape drives I'll be pretty
happy and so will my customer.
> The reason, as every consultant will tell you, is it depends. For example, if
> your multiple instances of nsrclone are trying to read virtual tapes that are
> on the same RAID group (or equivalent), then you're going to get lower
> performance than if they are on widely different RAID groups. That's just one
> example, there are tons of others. Give us some more details and maybe
> something will leap out.
+----------------------------------------------------------------------
|This was sent by gerard.robinson AT fordway DOT com via Backup Central.
|Forward SPAM to abuse AT backupcentral DOT com.
+----------------------------------------------------------------------
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 with 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
|
