Average replication on Storj

  1. What is the average replication for files on Storij?

  2. Where do you see the stat on this data?

  3. How is the replication like for small files vs large files?

You should check out the whitepaper: lots of good info in it. Such as:

“Storj does not rely on replication to achieve durability and distribution, but instead uses Reed Solomon17 erasure coding. Storj is able to achieve over 11 9s of durability with an effective expansion factor of 80/29= 2.7”

4 Likes

Related blogpost: Replication is bad for decentralized storage, part 1: Erasure codes for fun and profit

3 Likes

And as for the stats (to see how well the network is doing, maintaining a minimum of 29 pieces of each piece of data to ensure it’s recoverable)… I believe that’s what the "Healthy Pieces Median/Min) refers to in the dashboard.

2024-01-30_eu1

(If I’m interpreting it wrong, somebody let me know…) but I think that’s saying “For the data being managed for the EU1 Satellite… for every piece of data originally encoded and uploaded as 80 pieces… a median of 66 and minimum of 54 pieces were found across all SNO’s monitored”

If people always start by uploading 80 pieces… why not always 80? SNOs are coming and going all the time (for maintenance, connection outages, permanently, whatever…) so the Satellites are constantly initiating repairs to cover any SNOs who aren’t online.

2 Likes

We do try and get back up to 80. There is a repair queue. Let’s say some piece drops down to 50 so it triggers 30 repair jobs. It is possible some of those fail, giving us less than 80, such as 75. The other 5 go into the repair queue again and are reprioritized. The goal is to get to 80.

3 Likes

There is no replication and no users’ files, only pieces of encrypted files, see

and

1 Like