HP MSL 2024 SCSI 1840 drive throughput

[KEN_SYD]

Thanks Curtis,

I will check with ArcServe and report back.
Hava a good day.

[KEN_SYD]

Hi Curtis,

Can you please see attached image? Is that the last compression ratio you mentioned about?

I greatly appriciate your help. Much quicker than tape support team.

[Curtis Ballard]

Yes that is the average write compression ratio achieved on the 153GB written to that tape.

It is rather convenient that the compression ratio achieved with your data exactly matches one of the built in compression ratios in the L&TT performance test so you already know that the hardware is capable of writing at 177MB/s with that compression ratio when writing from a memory buffer.

[KEN_SYD]

Hi Curtis,
So in LTT Dev perf, the option random, Zeros, 1.6:1, 2:1 and so on means LTT will send a data that can be compressed at that respectively ratio?

When I tested with Zeros, I also got 177 MB/s. That is a bit confuse. I check help page but there is no explaination.
Can you please help?
Many thanks

[Curtis Ballard]

Yes the L&TT options specify the average compression ratio of the data that will be sent during the test.

The "all zeros" test is a way to test for the absolute maximum rate that your system can possibly achieve when sending data. All zeros compress to almost nothing so the tape drive is just pulling bytes off the bus as fast as it can but it doesn't have to do anything with them except add to the count of zeros received.

Getting 177MB/s for both all zeros and 1.6:1 compression means that 177MB/s is the hardware limit of your system and that the drive can accept data at that rate when it is compressible at 2:1.

You could have seen 177MB/s for 1.6:1 compression but 320MB/s for all zeros if your system was capable of pushing data at the maximum possible speed a U320 SCSI bus can support but the drive hardware peaked at 177MB/s with 1.6:1 data. When both values are the same the drive hardware could accept data faster than the server can send it.

[KEN_SYD]

Hi Curtis,
From your information about zero testing mean, I check further I found that tape library connected to backup server using Ultra 3 cable (Dell P/N: 8948X). Maximum bandwidth for Ultra 3 is 160 MB/s. That could explain why I always get 177 MB/s for zero, 1.6:1 and higher compression testing.
That makes sense right?

[Curtis Ballard]

The HBA will determine the maximum speed possible more than the cable. I don't see anything in the thread that would let me determine what rate your HBA is running at.

An L&TT "support ticket" would have that information. There is a tab for "support ticket" once you have selected the drive and that will take you to a page where you can "extract device data" and then save the ticket. Inside the ticket it will say exactly what speed the HBA and tape drive have negotiated.

[KEN_SYD]

Thanks Curtis,

In support ticket, negotiated burst data rate is 320 MB/s (see attachment). Thus still a mistery why limited at 177 MB/s.
So cable is not important in this case. Thank for the information.

[Curtis Ballard]

No mystery why the speed is limited to 177MB/s - that is all that the hardware in your system is capable of pushing the data. Almost nobody has a server that can actually push an U320 SCSI bus at 320MB/s. I think 177MB/s is quite good and really close to the maximum possible for the drive hardware with the data compression ratio that you get for your data so I wouldn't spend any time trying to improve that.

[KEN_SYD]

Thank you everyone, especially Cutis.

I now know what is maximum hardware ability. And found LTT is a very useful tool for drive testing.

I'm currently working with arcserver because the estimate native speed when using arcserve is just around 50-80 MB/s when compression ratio is from 1.3 to 1.5.

Thank again for all valueable information from you guys.