This is not necessarily a problem, more of an observation, but if anyone else agrees that these results are disappointing and has any advice or a magic fix, it would be much appreciated. I am new to using BE2010 and have just implemented a LTO5 single tape backup unit system. Perhaps there is a setting somewhere that I have missed in setup...
I am doing a standard backup of 2 files of around 300gb each to new LTO5 media, and getting an average job rate of ~ 1750MB/Min. With verify on, this takes around 12 hours.
To be honest, I was expecting something significantly faster than this both in terms of tape write speed and the subsequent verification. The theoretical maximum for LTO5 is 140MB/sec. This gives a potential job rate of 8400Mb/Min, obviously this is just on paper and not "real world". I have however run IBM Tape Diagnostic Tool and have a confirmed write speed of 132.74MB/s, so pretty close to the theoretical max.
With BE2010 I always get roughly the same job rate rate, regardless of where the files are located - I have tried backups with the source files on Raid 1 or Raid 5 arrays (the R5 obviously has a much faster sustained read capability). To me this suggests that the hard drives/their controllers are not a bottleneck.
However... when a backup job is running, I always see roughly the same read disk activity from beremote.exe of 30Mb/Sec on these files. Multiply this by 60 and you get almost exactly my Job Rate as reported in BE2010. Why such a low disk reported throughput? The arrays I tested on are quite capable of 100-600Mb/min sequential read, and the server is doing virtually nothing else. CPU utilisation is ~ 10% (SQL Server ticking over mainly, DEV system so no users connected 90% of the time), beremote.exe is usually around 1% CPU util whilst a backup job is running. It is almost as those there is a cap on BE2010 source file disk read speed.
The relevant specs of the hardware are below.
- IBM System X3630M3
- Dual Xeon E5506 @ 2.13GHz, Quad Core, 4Mb Cache, QPI 4.8GT/s
- 32Gb RAM (8Gb free, SQL Server memory capped)
- Windows 2008 R2 X64 SP1
- ServeRAID M5015 SAS Controller, 512Mb (hard drives)
- 2*IBM 2TB SATA 3Gbps (Raid1)
- 4*IBM 2TB SATA 3Gbps (Raid 5)
- IBM/LSI 6Gb SAS HBA (dedicated for Tape Drive)
- IBM HH LTO Gen 5 Model TS2250
The media is all new LTO5 media, the tape drive is new too.
As far as I can see, I am only achieving about 20% what is possible with LTO5. Basically I am getting less that LTO2 speed.
I have done a few more tests using a smaller single test file (around 20Gb) for convenience, but still seeing the same scenario.
I have also had a look around Symantecs website regarding slow tape performance. The articles I have read are more geared towards tweaking for a bit more performance. I can only think that there must be something fundamentally wrong with my setup as I am not after a tweak for a few Mb/sec, but looking for at least doubling my performance which I don't think is an unreasonable expectation. Given that my scenario of a few large files located on the same server this should be close to an ideal backup scenario in terms of speed, so achieving only 20% of the maximum throughput is pretty disappointing, and to be honest inconvenient.
Apologies for the long post!
Cheers
Paul
Accepted Solutions
