Bug #9556

FAQ: does USB3 improve Tails performance?

Added by BitingBird 2015-06-10 20:20:52 . Updated 2018-02-08 08:03:42 .

Status:
Rejected
Priority:
Normal
Assignee:
Category:
Hardware support
Target version:
Start date:
2015-06-10
Due date:
% Done:

0%

Feature Branch:
Type of work:
End-user documentation
Blueprint:

Starter:
Affected tool:
Deliverable for:

Description

Many users ask that. I don’t know the answer.


Files

bonnie.tar (20480 B) hybridwipe, 2015-11-09 23:05:16
bonnie_jessie.tar (10240 B) hybridwipe, 2015-11-13 15:28:33

Subtasks


History

#1 Updated by BitingBird 2015-07-24 03:53:59

  • Assignee set to BitingBird

#2 Updated by intrigeri 2015-08-03 05:19:56

Ideally, one would benchmark the same USB 3.0 stick on a single computer that has both USB 2.0 and USB 3.0 ports.

#3 Updated by BitingBird 2015-08-05 08:07:33

  • Assignee deleted (BitingBird)
  • Type of work changed from End-user documentation to Test
  • Starter set to Yes

Yes. De-assigning from me, maybe somebody else wants to do the testing :)

#4 Updated by emmapeel 2015-10-19 00:12:55

One user has reported that Tails on USB 2.0 and 3.0 devices take the same time to boot from USB 2.1 and 3.0 ports.

#5 Updated by hybridwipe 2015-11-09 19:58:17

I tested this on one of my laptops (Acer Aspire), which has 3 USB ports, 1x2.x & 1x3.0. I tested with tails 1.7, then enabling an admin password and installing bonnie. I’ll attach the results for the 3 ports (1 is the 2.0 port, 2 & 3 are 3.0). In short, 3.0 is about twice as fast for some metrics:
E.g., Sequential output rewrite goes from 15k to 30/29k. Latency drops from 15k/ms to 11k/ms. Sequential input goes from 36k with 126ms latency to 109k with fractions of a millisecond.

Boot times (from a cold boot, until the login screen appears and is usable):
1: 1:04.35
2: 1:07.28
3: 1:04.45

So yeah, no difference on boot.

#6 Updated by hybridwipe 2015-11-09 23:05:22

#7 Updated by hybridwipe 2015-11-09 23:06:15

Forgot to add, this was using a usb 3.0 key (ADATA UV128/32GB)

#8 Updated by sajolida 2015-11-10 11:24:21

Thanks for investigating all this! So the short answer is: USB 3 doesn’t improve startup time but it can improve read and write performances to the persistent volume. Is that it?

#9 Updated by hybridwipe 2015-11-10 11:43:50

sajolida wrote:
> Thanks for investigating all this! So the short answer is: USB 3 doesn’t improve startup time but it can improve read and write performances to the persistent volume. Is that it?

Correct.

#10 Updated by intrigeri 2015-11-11 09:01:25

I’d be curious to know if startup time is improved on USB 3.0 for Tails based on Jessie (it might be that some more startup parallelism in there allows us to benefit from more I/O throughput).

#11 Updated by hybridwipe 2015-11-13 08:54:23

intrigeri wrote:
> I’d be curious to know if startup time is improved on USB 3.0 for Tails based on Jessie (it might be that some more startup parallelism in there allows us to benefit from more I/O throughput).

Boot times using 20150111:
1: 1:11.49
2: 1:09.58
3: 1:10.48

I also saved the bonnie output but don’t have it handy with me atm. I can upload later if there’s interest.

#12 Updated by hybridwipe 2015-11-13 15:28:46

Bonnie files / html results for jessie testing.

#13 Updated by sajolida 2015-11-14 01:48:41

So Tails Jessie is a bit slower to start in general and USB 3.0 still makes no clear difference.

#14 Updated by intrigeri 2015-11-14 08:02:48

> Boot times using 20150111:
> 1: 1:11.49
> 2: 1:09.58
> 3: 1:10.48

Does this compare USB 2.0 vs. 3.0 on Tails/Jessie, or something else?

#15 Updated by hybridwipe 2015-11-15 21:16:27

intrigeri wrote:
> > Boot times using 20150111:
> > 1: 1:11.49
> > 2: 1:09.58
> > 3: 1:10.48
>
> Does this compare USB 2.0 vs. 3.0 on Tails/Jessie, or something else?

Yes, same setup as in comment #5, but with Jessie:

I tested this on one of my laptops (Acer Aspire), which has 3 USB ports, 1x2.x & 1x3.0. I tested with tails 1.7, then enabling an admin password and installing bonnie. I’ll attach the results for the 3 ports (1 is the 2.0 port, 2 & 3 are 3.0).

To summarize:

Tails 1.7:
Port 1 (usb 2.0): 1:04.35
Port 2 (usb 3.0): 1:07.28
Port 3 (usb 3.0): 1:04.45

Tails 1.9 (Jessie/20150111):
Port 1 (usb 2.0): 1:11.49
Port 2 (usb 3.0): 1:09.58
Port 3 (usb 3.0): 1:10.48

#16 Updated by intrigeri 2015-11-16 03:55:11

Thanks!

#17 Updated by cypherpunks 2016-03-02 02:46:53

From what I know, the vast majority of the boot time is taken up by enabling AppArmor, and to a lesser extent, other services. But when I boot with the toram option, I notice an incredible speedup when using USB 3.0. The difference is more than 100 MB/s vs 20 MB/s. That doesn’t matter when only a few MB have to be read at boot but it matters a lot when 900 has to be read.

However whether or not this also speeds up runtime performance is another matter. How fast can the “average machine” read from a highly compressed squashfs image? Can it decompress faster than USB 2.0 can stream the data into RAM, or slower? If it can decompress faster, then USB 3.0 should show a real improvement at runtime.

#18 Updated by Anonymous 2018-01-19 14:28:13

  • Assignee set to sajolida
  • QA Check set to Info Needed
  • Type of work changed from Test to End-user documentation
  • Starter deleted (Yes)

@sajolida / tech writers: do you want to document this in the FAQ:

“USB3 doesn’t improve startup time but it can improve read and write performances to the persistent volume.”
If yes, fine, if not please reject this ticket and leave it here for reference.

#19 Updated by sajolida 2018-02-08 08:03:42

  • Status changed from Confirmed to Rejected
  • Assignee deleted (sajolida)
  • QA Check deleted (Info Needed)

I feel like it’s not worth documenting because the performance gains are not clear. People also probably confuse USB3 sticks and USB3 ports on their computer. So being precise would take a bit more explanation to in the end say “sorry but not really”.

I say we reject this ticket and we can open it back later on if the amount of people asking about USB3 is bothering us, despite us not having very exciting things to answer them.