Benchmarking Testnets

This is a place to track the performance of testnets, specifically the upload speed.

To help out, please follow these steps:

  • find a file about 20 - 25 MB (preferrably one that’s available somewhere on the internet)
  • your current upload / download speed as measured on
  • upload it several times to the testnet as private data
  • make a post with the following info:
    • the file size (and url if it’s a public file)
    • the amount of time taken for each upload
    • the specific testnet being used (eg test15, alpha1 etc)

This will help track consistency and performance across each testnet from a variety of locations and connections. Consider it a form of market research :sunglasses:

Aggregated Results:

Date          Network    Average Speed (Mbps)
2017-03-24    Test 15    1.447

Test 15 Aggregated Results

Avergae Speed (Mbps)
1.026   trenti
0.797   draw
1.337   digipl
3.323   dask
1.200   polpolrene
1.000   mav

File Size: 24.3 MiB (download)
Testnet 15
Speedtest Download / Upload speed: 26.93 Mbps / 15.21 Mbps
Timings (seconds):


I think you also need to know the user’s upload/download link speed too, as this may have a bearing.

Your tests show speeds of around 110K+ Bytes/sec and for many ADSL2 users this is the maximum upload speed their link can achieve (110 * 8 = 880K bits/sec + overheads) with a 1Mbits/sec upload speed. And if they are doing other things then they may not even be able to achieve 1 Mbits/sec

EDIT: this is mainly a concern for when they upload a file to their account.
EDIT2: @mav, nice upload speeds you have. Maybe one day us Aussies can get such real speeds


I’m sorry to say I’m in Australia on the NBN :smiley: those speeds are lower than normal

1 Like

If the file is stored as immutable data, public or private is the same. The only difference is the datamap encrypted or not.

First upload test: 94.5KB/s. (70MB)
Second upload test; 112,4KB/s (91,6MB)
Both as public data (MP3 Albums)

Resend second test as private, so chunks stored: 138KB/s

Fiber: 300/300Mbps.

I think that the relay node is the bottleneck in several cases.


really, where I am - its slow as hell. Can’t get to 2Mbps - sad state of affair really. :slight_smile:
Now even Kenya is faster than us poor aussies internet wise.

1 Like

My view is not that different from previous testnets. I’m on cable with quite good speeds here in Holland. Upload bumps between 100 KB/sec. and 220 KB./sec. I always make a dummy file around 475 MB. and split it up in 23 MB chunks.

I’m not running the Vault at home at the moment, so this is me connecting to a Vault somewhere. Will test downloads later on.


(Edit cause replied to a user rather than main thread)

24.2 MB / 61 sec average = 398 KB/s upload from home

100Mbit up/down fiber

1 Like

I made several test rebooting the Launcher. From the third test I change the order of the Hard Coded IP trying to connect to different contacts.

24MB 7zip files.

First try: 167KB/s
Second try: 167KB/s

Third try: 162KB/s
Fourth try: 170KB/s
Fifth try: 178KB/s
Sixth try: 159KB/s

Quite stable…


File size: 20.1 MB = 160,8 Mbits
Testnet 15
Download/Upload speedtest: 192/19 Mbps

1st upload: 211 seconds -> 0,762 Mbps (= 95.3 KB/s)
2nd upload: 210 seconds -> 0,766 Mbps
3th upload: 214 seconds -> 0,751 Mbps


I wonder this kind of benchmarking can only work if the files are random data?.. maximising the number of fragments that require uploading.

Also, I wonder that much of the time appears to be local fragmenting and encrypting and only the last seconds is the actual upload?.. but I don’t know if it’s more simple than that.

Useful still to get a sense of time/MB averages… but I wonder if two different files would see different results.

To create a random roughly 10MB file in linux:

dd if=/dev/urandom of=a.log bs=1M count=10


File size: 24.9 MB (mp4,
Testnet 15
Download/Upload speed: 93.98 Mbps / 30.45 Mbps
Private upload tests:

  1. 208 s
  2. 204 s
  3. 190 s
  4. 190 s
  5. 183 s

Just started a Vault. for resource_proof I had a 2MB/sec. upload speed. When I was done my Vault started to fill up:

It peaked in the beginning with 6.2 MB/sec. Thats 6.2 MegaByte per second. Thought it was quite amazing :clap:


File Size: 19.8 MiB (download)
Testnet 16
Speedtest Download / Upload speed: 2.31 Mbps / 0.80 Mbps (slower connection to my usual NBN one!)
Timings (seconds):

About 0.5 Mbps upload speed

My usual test (Mp3 Albums)

Fiber: 300/300Mbps

Friday: 180KB/Seg.
Saturday new account same data: 201KB/Seg. (My account was lost but, al least some chunks, possibility not)
Today new data: 154KB/Seg

1 Like

I’m currently crawling at about 0.4 Mbps on clearnet so my input for Testnet17 won’t be much use from this connection. Anyone got some stats for upload speed on the new testnet?


In a couple of test with the Web Hosting Manager, uploading about 250KB/Seg.

Downloading is quite fast, more than 10MB/Seg, but seem that the WHM use some kind of cache (unfortunately after upload a directory of 150MB. cannot access neither the WHM or the SAFE Mail).


Testnet 19:

19.8 MiB file took 2m8 to upload for a rate of 1.3 Mbps - which is pretty close to the speed of previous testnets.

Only did the one upload so not especially scientific this time :wink:


I have found 19 to be slower to upload large files but generally quicker to upload smaller ones. At the same time GETs seem quicker with websites loading faster and with file downloads being quicker too. It’s pretty subjective and not very scientific, and any differences may well have to do with my system and connection rather than SAFE, but I wonder if any tweaks have been made to optimise performance in certain areas?


Alpha 2, 19 MiB file uploaded in 3m 34s

So a rate of 0.74 Mbps


  • this is just 3h after the release of the network so it’s probably a fairly busy place
  • I’m on a shitty connection that maxes out at 0.9 Mbps upload so… I need to test from a better connection!

Any other speed tests for alpha 2?

edit: got an upload speed of 1.73 Mbps on a different connection (tested at the time as capable of about 37 Mbps upload so SAFE nowhere near saturated this connection)