LTO-5 tapes 1:1 compression, LTO-4 2:1

[bn8959]

Hi,

I am running Backup Exec 2010 R3 with an IBM TL3100 autoloader with an LTO-5 SAS drive. I am running backups to deduplication storage and then duplicate to tape jobs.

Any job that is written to an LTO-5 tape results in no compression (1:1), whereas if I write the same job to LTO-4 tapes, I get compression levels of around 1.8:1 to 2:1.

The drivers are the Symantec ones and are up to date and the library and drive firmware are also up to date.

Does anyone have any suggestions? My full backups are now tipping over to two tapes, which is annoying as we will have to double the number of tapes we have.

Interestingly on my duplicate policy jobs I can only select Hardware compression or None - so I cant test software compression.

Regards,

Ben

[Dev_T]

 Hi Ben,

"Compression" (Hardware OR Software) is the property of the File System (Sopftware) and the TAPE Drive (if Hardware) Backup Exec does not play any role in compression. I think you should run the Windows Backup tool on both the Drives & verify the results.

[ZeRoC00L]

Is your backup server fast enough to deliver the data to the tape streamer ?

[bn8959]

Hi, I assume it is, since it backs up fine to LTO-4. I know LTO-5 has a higher throughput, however I wouldnt expect the rate of data sent to the tape drive would affect its ability to compress.

[bn8959]

I will do some tests. Thank you.

[bn8959]

Ive just done a simple test job of backing up 500mb of data directly from the media server direct to tape. When written to LTO-5 I got a compression ratio of 1.1:1 and on LTO-4 is was 1.3:1 (430mb used on LTO4 and 511mb used on LTO5, for 543.4Mb of data).

I think that is really odd - I would expect for exactly the same data in the same drive, the same compression ratio.

[ZeRoC00L]

It can be that if the tape does not get the data fast enough, it disables compression to prevent stopping and rewinding the tape.

[bn8959]

Unfortunately I am running Windows Server 2008 R2, which does not support backing up to tape from the built in backup tool.

[bn8959]

Ah, that is interesting. Is there a way to change how backup exec sends the data to the tape (for example by buffering more data so it can be delivered in bigger chunks), or to change this behavior?

The source is a deduplication data store on an iSCSI mounted volume. It is true that it isnt the quickest data source (~50mb/sec).

[bn8959]

Although having said that, my test run from local storage (which ran at a rate of 4000mb/min), revealed the same difference in compression ratios - despite running at 'full speed'

[ZeRoC00L]

Still sounds strange, have you tried another tape or even another brand of LTO5 tapes ?

[bn8959]

It does dosent it. Unfortunately this batch are all the same brand. I am ordering some more so will test with a different brand.

I am now experimenting with different block and buffer sizes, and also trying the IBM drivers instead of the Symantec ones.

[Ken_Putnam]

let's go back to basics for a minute

Are you using software or hardware compression?

If software, from the Devices Tab, right click the LTO-5\Properties  and verify that "Enable Compression" is selected
 

If hardware, is there a switch on the drive itself to enable/disable compression?  (Check your user guide)

[pkh]

I have not used LTO5.  With LTO4, I find that using the same set of data, the compression ratio is roughly the same whether I use software or hardware compression.

I would do the following test with one set of data, from disk, not from the dedup folder.

1) ensure that no encryption is used, either hardware or software

2) use software compression and see what is the compression ratio.  From the job log, note the amount of data that BE said that it backed up.

3) use hardware compression on LTO4 and see what is the compression ratio.  From the job log, check the amount of data that is sent to the tape drive (this should be the same as in step 2) and the amount of data actually written to tape.

4) repeat step 3 with LTO5.

If the amount of data sent to LTO4 and LTO5 is the same and the compression ratio is lower with LTO5, then you should check with the LTO5 vendor.  When it comes to hardware compression, BE does not get involved.  It just output the data to the tape drive and the tape drive is the one which compresses the data.