LTO2 vs 9940 Which way to go??

[phaase]

Looking at moving up from 9840 20 Gig level technology to 200/300 gig tapes to backup AIX & NT and was looking for some views on the pro's/con's of LTO2 vs 9940.

The management lean is towards LTO2 as it is cheaper and more "politically correct" at the moment.

I came across the following worrying statement:

"LTO tapes will not read past errors once they are encountered on a tape. That means that one error on one tape can ruin your entire restore process. 9840 / 9940 technology is designed to read past errors and to use extensive error correction technology to recover the data and continue with the restore process."

Given the above, has anyone seen/experienced any problems like this or otherwise with LTO/LTO2 technology and what has been peoples experience with LTO/LTO2 as a backup media choice. Speed? Reliability? Rehitch delays? Anything else?

Paul

 

[comtec17]

This would all depend on how much data and how fast you need it to recover. I have experience with both products and the LTO2 is very fast 30MB/s native but takes about 3 minutes to load and mount. I have never experienced any read errors that stopped me from restoring data.

The 9940 tape is great if you are using some type of HSM or need to mount your tapes within 12 seconds.

The biggest difference between the 2 technologies is the amount of data that they will hold and native throughput.

Will these drives be SCSI or Fibre attached?

 

[phaase]

Drives will be Fibre attached. No HSM activity, the closest being TSM(ADSM) migrations and SAP log archiving. Recalls(restores) are few and far between.

Lots of TSM space reclamation and offsite copies for DRP are the main tape read applications. Writing is mainly backups from disk, tape-to-tape copy, and TSM migrations.

Throughput i'm guessing is going to be more input constrained and as such 30MPS drive capacicity makes either drive suitable throughput wise.

The 3 minutes to load and mount is pretty sad In fact...thats awefull!! Are all LTO2 drives this slow?

I could probably factor in 1-2 more LTO2 drives to allow for slow mount times and still come out cheaper. Being spoilt by quick mount times in the past - Mandatory with HSM, but not as critical nowdays. I'll have to check the mount profile to see how much this would impact.

 

[phaase]

I got to thinking about the 3 minute load time as most vendors talk about sub 1 minute. STK link below shows IBM and HP drives sub 1 minute. What causes the extra 2 minute delay?

http://www.storagetek.com/products/product_page41.html#specifications

 

[comtec17]

The delay is up to 3 minutes. I have seen them load in 45 seconds on a tape that is scratch. The load time is identical to DLT due to the forwarding and reading of the blocks to position the tape.

 

[phaase]

OK...No high speed locate and is relative to file position on tape.

1/ I could possibly keep a few 9840's for daytime write processing and then copy them to LTO's when full to overcome the speed delay......

2/ I could use a scratch pool of LTO's for daytime output and merge them when it's quieter later....

Other than the locate speed issue, any other reason not to go LTO2?

Is rehitching an issue?

 

[comtec17]

I currently installed on a client site 10TB of disk for a disk pool with 20 LTO Gen1 tape drives in a STK L700e with TSM. When I migrate from disk to tape it is very fast as long as you have your TSM Server performance tuned. The slow load time is seldom felt.

The 9940/9840 difference is that the tape is real to real with intellegent reading. The tape rewinds to the middle before unloading. As with LTO it is a single thread of tape with the very last of the tape the beginning and the very front of the tape when you open the sleeve being the end.

 

[phaase]

TSM server performance is being fixed at the moment with a new box so that should be OK. Data to back up is about 8Tb's.

If we go LTO2, we are possibly looking at 4-6 LTO2's in a new L700e or a second hand L700. Will also see what else is available, but the L700/e seems best fit. Not sure what the "e" option gives though.

Thanks for the help!

 

[comtec17]

The e option is for a pass-thru. You can connect 2 L700-e's together to make for more drives and slots using 1 robot.

 

[comtec17]

Oops, it is 2 robots.

 

[dmayes]

To me the number one factor in STK-9xxx technology vs any LTO and/or DLT type is reliability, which includes the "duty-cycle" you'll be placing on the drives.

IMHO...IBM (magstar) and STK (9xxx) have the best drives to support a heavy workload.

As long as your management's main concern is cost...then you might remind them that "you get what you pay for"...and do they want to risk backups/recoveries on the "cheapest" technology.

I'll climb off my soap-box now.

 

[johnny99]

We use 9940B and it's a great drive.
When we moved away from DLT LTO was not on the market so we went for 9940. And I am happy about that.

What I would expect is that in an havy loaded installation 9940B would be the best around (if you can live with drives that are a little more slow then 9840C)

Up-front the cost of LTO2 is lower, but depending on your installation 9940B could have a lower TCO.

/johnny

 

[drpaine]

dmayes is absolutely correct.

there is a huge difference in reliability and duty cycle for 9940 vs LTO

 

[drpaine]

does anyone know where i can get a manual for a 9730 timberwolf in pdf format ?

please send to drpaine10@yahoo.com

 

[comtec17]

sent

 

[comtec17]

drpaine,
I tried to email the manual to you but your account blocked it stating it was too large.

 

[phaase]

How big is the duty cycle difference???

How much is too much?
20 mounts per day?
40 mounts per day?
100 mounts per day?

What starts happening? Tape I/O errors? Machine checks?

I need to be able to explain what the difference is and what happens when the drives are pushed too much.

 

[comtec17]

For 9940B:

Mean time between failures (MTBF)
Power on: 290,000 hr @ 100% duty cycle

Tape load: 240,000 hr @ 10 loads/day (100,000 loads)

Tape path motion (TPM): 196,000 hr @ 70% TPM duty cycle

Head life: 8.5 yr @ 70% TPM duty cycle

Uncorrected bit error rate: 1x10(-18)

Undetected bit error rate: 1x10(-33)

For 9940A:

Power on: 290,000 hr @ 100% duty cycle

Tape load: 240,000 hr @ 10 loads/day (100,000 loads)

Tape path motion (TPM): 196,000 hr @ 70% TPM duty cycle

Head life: 8.5 yr @ 70% TPM duty cycle

Uncorrected bit error rate: 1x10(-18)

Undetected bit error rate: 1x10(-33)


For LTO Gen2

Mean time between failures (MTBF)
Power on: 250,000 hr @ 100% duty cycle

Head life: 60,000 hr

Uncorrected bit error rate: 1x10(-17)

Undetected bit error rate: 1x10(-27)

For LTO Gen1:

Mean time between failures (MTBF)
Power on: 250,000 hr @ 100% duty cycle

Head life: 30,000 hr

Uncorrected bit error rate: 1x10(-17)

Undetected bit error rate: 1x10(-27)

 

[johnny99]

I found some links that could be of use to you:

http://www.storagetek.com/products/tape_storage/tape_drives.html

http://www.storagetek.com/upload/product_collateral/current/38/289.pdf

http://www.storagetek.com/upload/product_collateral/current/38/362.pdf

http://www.storagetek.com/products/product_page38.html#specifications

http://www.storagetek.com/products/product_page38.html#specifications

A very interesting number you can get out of the two last links are:

LTO2:
Power on: 250,000 hr @ 100% duty cycle
Head life: 60,000 hr
Uncorrected bit error rate: 1x10(-17)
Undetected bit error rate: 1x10(-27)

9940B
Power on: 290,000 hr @ 100% duty cycle
Tape load: 240,000 hr @ 10 loads/day (100,000 loads)
Tape path motion (TPM): 196,000 hr @ 70% TPM duty cycle
Head life: 8.5 yr @ 70% TPM duty cycle
Uncorrected bit error rate: 1x10(-18)
Undetected bit error rate: 1x10(-33)

Uncorrected bit error rate and Undetected bit error rate is very interesting in the LTO2 vs. T9940B.
The difference is huge, Factor 10 on uncurrected bit errors and a factor 1.000.000 on Undetected bit errors

Another thing to remember is that the only other tape drive that StorageTek sees as "in the same liga" is IBM3590E11

 

[phaase]

LTO1's have a 100,000 load/unload duty cycle which is the same as the 9940's. Guessing LTO2 is the same or better.

Power on hours look similiar enough not to be significant.

Head life of 60,000 hours @ 24 hours per day = 2000'ish days = 6 years....plenty.

I'm not seeing the significant difference in duty cycle performance....Am I missing something?