Updates on Test Data

Announcements
623

Awesome. Thats good to know. I thought it looked at your stored data qty at the end of the month.

624

I believe the satellite tracks the time from the moment your node sais it stored the piece until the client send delete command, or the TTL is reached. I donā€™t know whatā€™s the minimum time frame (unit) for messurement. Maybe itā€™s 1 hour, or 1 min, or less, but you get paid for the above mentioned time period.

625

I see. You are correct, I missed the point, Iā€™m sorry.
ā€¦not enough sleep than not enough coffeeā€¦

and @nerdatwork is right, I didnā€™t see what you were referring to and that confused me.

3 Likes
626

Do you know the granularity of the calculations?
Hours? Minutes? Seconds? Something else entirely?

627

Nanosecods! (or maybe even more precise)

4 Likes
628

Ugh.
Search is your friend. Sorry I was lazy and thank you :slight_smile:

629

I only found it fast, because it was my question being answered, otherwise it would have been difficult

630

Please tell me, is the function of saving the current scanning position when the filewalker is running already working? Or is work still underway?

1 Like
631

We are going to test 3 alternative node selections today. It would be great to get some feedback how each of them impacts your nodes.

  1. choice of 2 (test run finished a few minutes ago)
  2. bestofn with an factor of 2 (API pods currenty restarting. Test should start in about 10 minutes)
  3. choiceofn with n=3
  4. filterbest (not sure about the exact config for this one yet)
6 Likes
632

Awesome to see more options being tested.
Could you elaborate a little? 1 and 2 sound effectively the same. And what does filterbest mean?

2 Likes
633

My plan was to describe them later. For now we just need to find out which one works best :slight_smile:

4 Likes
634

Understood, Iā€™ll try to be patient. :joy:

3 Likes
635

If you post between each testā€¦ so we have a timestampā€¦ then we can align each test with our router graphs :+1:

Edit: of all the times for Th3Vanā€™s setup to be b0rked: his graphs would have given you a great general-idea-of-how-100+nodes-perform.

4 Likes
636

bestofn with n=2 was a bit better. I have seen 60MBit/s on my storage node for a few minutes. We could maybe increase n and see how that works.

Up next choiceofn with n=4 (API pods are currently restarting. Test should start in about 10 minutes.

4 Likes
637

One data point for you, the second test (I think that would be bestofn with n=2 (whatever that means)) saw about a 10%-15% increase in downstream throughput across all my IPs without a significant increase in CPU or IOWait.

4 Likes
638

As far as I can tell, the previous one was better, as @ACarneiro said.

1 Like
639

45-75Mbps right now for me

1 Like
640

It still is bestofn. I canā€™t get choiceofn to work. I will try to figure it out in the background. In the meantime lets try bestofn n=4

4 Likes
641

Whooooa! That pretty much doubled my throughput.
Not sure what itā€™s doing for the success rate, mind youā€¦

EDIT: Interestingly, for the first time during this stress test, not all IPs had a similar increase in traffic.

2 Likes
642

As far as load goes, everything seems normal on my end. Yea a bit higher RAM usage, but all within reasonable limits. Buffers + cache still good to go. I think the nodes are saying they can handle more than this :slight_smile:.

2 Likes