Performace, or How IPFS will be better than BitTorrent

[DonaldTsang]

Here are a list of issues that is plaguing IPFS: Performace. It has to beat existing system to gain a wider adoption. One factor is "the default should be semi-optimal", where once installed, without much parameter tweaking, it will perform better than BitTorrent.

• Improve transfer speed #2111 (transfer speed)
• Massive DHT traffic on adding big datasets #2828 (DHT management)
• IPFS consumes a large amount of network traffic #2917 (network traffic usage)
• Excessive bandwidth use #3429 (another bandwidth usage)
• Adding file to repo with many files is slow #3545 (file adding)
• Importer flow needs a redesign #3638 (import issue)
• ipfs add hangs and stalls with large folders #3885 (hanging system)
• Pinning is slow when there are many pins #5221 (slow pinning)
• IPFS Repo Storage And Files Over A Huge Amount（30G+）, "ipfs add" action slow down in a high concurrency environment #5438 (another)
• Bitswap Improvement Plan #5723 (Bitswap)

Here are some tests

• Performance test - Add/Cat on a single IPFS instance #3131 (speed testing)
• Improve time to add 10,000 files serially. #3316 (adding files)
• Investigate Performance characteristics of 10GB LAN Transfer #3317 (high-speed ethernet)
• IPNS very slow #3860 (IPNS)
• Slow transfer over LAN #5037 (another speed test)
• IPFS Performance #5226 (newer test)
• Goroutine count, Node Performance, and correlation with peer count #5420 (further analysis)

[Stebalien]

Yes we have performance issues and we're working on them. But this issue doesn't really add anything to the discussion.

[DonaldTsang]

@Stebalien I was expecting this to be a meta-issue on performance

[Stebalien]

Ah, got it. My concern with this issue is that it's kind of a grab bag of different unrelated performance issues and we already have a perf label.

There appear to be four categories:

• IPNS is slow: IPNS very slow #3860
• Adding files to go-ipfs is slow. Could probably use a good meta issue, want to make one?
• Bitswap transfer speed is slow: Bitswap Improvement Plan #5723
• IPFS uses a lot of background bandwidth: Excessive bandwidth use #3429 (may need a better meta issue?).

[DonaldTsang]

@Stebalien In that case making a meta-issue for all four topics would be better, it would make reading easier.

[jasonzhouu]

@DonaldTsang
suggest a new item:

• What is the reason for very slow #6382 (latency of DHT resolving)

change a item:

• IPFS Performance #5226 (delay test)

[DonaldTsang]

@jasonzhouu are you suggesting to combine those two issues?

[jasonzhouu]

@DonaldTsang
I'm suggesting to:

• add the first one
• and change the description of the second one, which already exist in the original list.

[Stebalien]

I've created a meta issue for resolution delay #6383.

[jasonzhouu]

I've created a meta issue for resolution delay #6383.

Awesome!