Tape speed more than the max rate with multiplexing !? and a few more questions about it

[mhdganji]

Hi,

I'm using tape multiplexing jobs in netbackup and there are a few questions:

1- The network card speed is over 3 Gbps while the LTO-8 max speed is 350Mbps ! How can I enumerate this? Maybe because it's using some tape features like compression or accelerate or so. ha?

2- How can I find the actual tape writing speed while running the jobs? Now I'm just using estimates from network card transfer speed.

3- Is multiplexing safe and problem-free in restores? It makes me uncomfortable when I think the tape is sequential read/write device and multiple jobs are being written at the same time. No problem like pointers, restoring inconsistencies, low restore speed, de-fragmentation necessity or so?

4-Why multiplexing option is an option that should be enabled within schedules? Should it not be on storage, tape media or policy options?

Regards

[Nicolai · Accepted solution]

Hi @mhdganji 

1: Writing speed for tape drives are typical without compression applied. Appling compression will increase the amount of incoming data required to make the tape drive running without start/stop operations. Always ensure incoming network bandwidth is larger that tape drive bandwidth

2: Using the activity monitor to get a picture, multiplexing complicates the picture. But in general the writing speed you see in the activity monitor is spot on.

3: Multiplexing basically mean multiples backup streams are written concurrent to the tape drive, to use the full bandwidth of the tape drive. However what you gain during backup, you loose during restore. Let's say  backups are written with a MPX=4, and the backup are equally in size and delivered speed. then during restore 3/4 of the read data must be "throw away" during restore, causing increased restore speed. To make it worse, let's assume you have 3 really fast backup stream, and a really slow backup. Then the ration for data to be "throw away" for the slow backup, may be much worse. Bottom line, consider fragment size and MPX carefully. Large fragment sizes and large MPX is a really bad combo.
From a usage point of view, using multiplexed backup are safe to use. The technology has been in Netbackup for more than 22 years.

4: The storage unit configuration has a max number of stream, that goes hand in hand with the policy configuration.

[mhdganji]

So many thanks Nicolai,

Just to make things more clear to me

About Q2: Assuming there are 4 lines available for backup, why not the same available for restore from tape. As far as I can imagine, if we can have four channel/line/stream for writing let's say 4 VM's backup to the tape, then we will have all those for restoring the all 4 VM's together (Assuming we need to have all those back at the same time), What am I missing here?

Q4: I see that there are stream max number in storage but I cannot find why the multiplexing option is not in the hmmm let's say tape or policy general attributes. Although I can imagine that this is useful when having full backups in a special window and the incremental in another time at then we should not fill out the bandwidth

At last, does this MPX stream or storage stream is per VM or per disk per VM (each disk in each VM would count)

Appreciated

[Nicolai]

Hi @mhdganji 

Netbackup will restore all 4 steams at the same time, if the restore request are submitted within the MPX_RESTORE_DELAY setting.

https://www.veritas.com/support/en_US/article.100006440

The idea of MPX on schedule (I guess) is that you can "tune" the MPX setting depending on workload. E.g. Less MPX setting for full backup, and higer MPX for incremental backups. But no larger than the storage unit setting.

Epically for database backup, I found tuning MPX setting on schedule level really useful.

/Nicolai