Parallelism: Default number of threads

General Topics
1

The FAQ says parallelism is supported, but doesn’t specify the default number of threads Kopia uses. Is the default 1, i.e. no parallelism? I back up to a hard drive, so multithreading is undesirable, as it leads to fragmentation, slower transfer speed (head jumping for no reason) and increased wear. Just wanted to make sure I’m not doing those things unknowingly.

2

Quoting from the FAQ you linked to:

By default, Kopia sets this setting to the number of logical cores your machine’s CPU has

Out of curiosity: what hard drives do you use and how much data do you read/write that you need to care about these things?

3

Noted. Should’ve paid more attention. That’s for the second parallelism though. What about the first one?

Just regular externals. 2.5, 3.5. Mostly WD. Doesn’t matter though. All spinning drives by their nature are worse off with multithreading than without it.

Edit: Seems the default is 1 re:First parallelism.

> kopia policy show --global
Uploads:
  Max parallel snapshots (server/UI):   1   (defined for this target)
  Max parallel file reads:              -   (defined for this target)
  Parallel upload above size:      2.1 GB   (defined for this target)

Weirdly, Max parallel file reads is set to -. Isn’t it supposed to be the number of CPU threads?
Also, would setting Parallel upload above size to - or 0 set both parallel snapshots and parallel file reads to never? I want Kopia to be entirely sequential when it comes to writes.

4

I’m not so sure about that. I’d be surprised if those drives had anything less than 64MB cache but that’s more a question of age. I’d look @ the respective data sheets & run some benchmarks… & check that the SMART diagnostics are clean.

I’d be more concerned with not enough reserved blocks remaining for future bad block reallocation.

Thanks for posting this thread. It helped me sort a quandary.

5

Spinning drives are usually quite good at taking data, as there are so many caches in the pipeline even before the data hits the drive. Reading, however, is a different thing. If you’re using a single disk for backing up, I actually wouldn’t reduce the number of threats, as longs as I am not dealing with few but big files. Kopia will pack smaller files into larger blobs anyway, so having multiple, concurrent reads threads on the source is a good thing.