Availability Cores Talk for PBA@UCB #632

BradleyOlson64 · 2023-06-06T19:40:30Z

Ready for final review

nuke-web3 · 2023-07-31T04:19:00Z

paritytech/polkadot#805 for your review targeting this PR 🙏 😁

* fiixes img with closing `/>`, fmt, spelling * mv files into 7-Polkadot mod * fix `yarn build` for img paths * fit code on screen better

…/pba-content into brad-execution-cores

nuke-web3

Trust this is good when you are ready! Let's just do a final bug check in #814 to fix in that PR if needed.

Overkillus

Overall I like the direction. Love the methaphor and would love it even more if it was expanded upon. Left a bunch of comments on specific slides

syllabus/7-Polkadot/Availability_Cores/Deep_Dive_Availability_Cores_Slides.md

Overkillus · 2023-08-02T19:55:02Z

syllabus/7-Polkadot/Availability_Cores/Deep_Dive_Availability_Cores_Slides.md

+
+<img rounded style="width: 1500" src="../assets/exotic_core_scheduling.svg" />
+
+Notes:


If you'll be going with the loading bay methaphor expand on it here. Say that a business can lease to carriages or maybe 1 carriage every other train. It works flawlessly with it so let's exploit it.

Overkillus · 2023-08-02T19:57:46Z

syllabus/7-Polkadot/Availability_Cores/Deep_Dive_Availability_Cores_Slides.md

+
+---
+
+# How Cores Gate Each Step of the Parachains Protocol


Don't you only explain how it gates/controls backing? I don't see how it connects to every step and I still don't see the connection to approval voting or disputes.

I have one slide explaining that since approval and disputes occur after inclusion they are implicitly gated. Gated in this case meaning that blocks will be approved or disputed in proportion to how much core time you buy and fill.

Overkillus · 2023-08-02T20:01:19Z

syllabus/7-Polkadot/Availability_Cores/Deep_Dive_Availability_Cores_Slides.md

+
+<pba-flex center>
+
+1. Improved liveness


How would you defend this claim?

Ethereum doesn't have cores and has no liveness issues. What in our design requires cores for liveness? What do we get out of it?

Feel like we HAVE TO mention sharded execution. Not everyone computes everything and cores are a medium a subdivision that allows for this division of labour.

relevant https://github.com/Polkadot-Blockchain-Academy/pba-content/pull/632/files#r1282368554 - "predictability" is probably a better framing than liveness, agreed.

An Ethereum transaction will get in eventually, so it is live, though can get crowded out in the near-term. but guaranteeing predictable near-term access to resources and smoothing out short-term volatility is the main benefit here.

recommend reworking this section for two benefits:

predictability (of parachain making progress in the near-term)

resource allocation (execution sharding division of labor)

I take all points given about rephrasing.

Though I have a couple word use quibbles around the interpretation of "improved liveness":

Improved - My use of improved is relative to a version of Polkadot with similar sharding but no mechanism to rotate backing groups. Though I suppose there are plenty of ways to rotate backing groups without cores.

Liveness - It seems I draw a different boundary between liveness and latency than perhaps the industry does. How long does a chain need to be inactive before it can be termed a liveness issue?
It seems to me that any unintentional degradation of 'time since last block' is a minor hit to liveness. The longer the 'time since last block' the worse the liveness issue is. So if we cut that time from 1 session to 100 blocks I feel like that improves our worst case liveness at least.

Overkillus · 2023-08-02T20:07:50Z

syllabus/7-Polkadot/Availability_Cores/Deep_Dive_Availability_Cores_Slides.md

+
+1. Improved liveness
+1. Flexibility of market structure
+1. Flexibility of allocation


Cores fundamentally limit our allocation flexibility. It limits the unit of allocation to 1 parablock per 1 relaychain every 6s for 2years in only a single market model with auctions. That is pretty much by definition a terrible allocation flexibility no?

Only with parathreads we solve part of the problem with the 2 years part but until we get the exotic scheduling our flexibility in other regards continues to be heavily constrained by the unit we have chosen. I cannot view it as a flexibility benefit. True flexibility is selling of totally arbitrary amounts of space/computation. Maybe a contentious topic...

Cores vs core allocation mechanisms is the distinction here. Slot auctions are a clearly bad core allocation mechanism, but this isn't baked into the definition of core itself.

We talked a lot in other lectures about how we can use cores more effectively with better allocation mechanisms.

Some discussion on the "totally arbitrary" aspect: https://github.com/paritytech/polkadot/issues/7441 , at least a step towards that direction.

It's true, probably not better for flexibility overall.

If anything, I'd frame it as a "predictability" benefit. Cores allow us to build mechanisms that reserve future resources for particular processes, giving them better guarantees of continuity and less exposure to short-term volatility. We basically trade off flexibility for predictability.

This is a bit of a definition thing again. I'd define it as inflexible when short periods of exorbitant prices on Ethereum prevent average users from interacting in a reasonable way. Predictability can act as a kind of flexibility in preventing such scenarios.

Also there is a bit of an apples and oranges comparison going on when we compare Polkadot, a sharded scalable network, to just Eth as an L1. If we include Eth L2s, then we could factor in reduced flexibility due to slow withdrawal periods de-facto partitioning the network.

rphmeier · 2023-08-03T03:59:43Z

syllabus/7-Polkadot/Availability_Cores/Deep_Dive_Availability_Cores_Slides.md

@@ -0,0 +1,560 @@
+---
+title: Deep Dive, Availability Cores


Could be confusing as we've referred to Execution Cores a lot so far

I’ll make a slide addressing the two names

rphmeier · 2023-08-03T04:00:45Z

syllabus/7-Polkadot/Availability_Cores/Deep_Dive_Availability_Cores_Slides.md

+```rust[1|3-7|4,9-13]
+pub(crate) type ParathreadQueue<T> = StorageValue<_, ParathreadClaimQueue, ValueQuery>;
+
+pub struct ParathreadClaimQueue {


This is actually dead code from early 2020 that is being replaced, not sure it's worth going over

…/pba-content into brad-execution-cores

BradleyOlson64 added 10 commits May 26, 2023 08:47

Stub of execution core slides

74b791a

Define blockspace and cores

76f7e64

Started cores and backing

58969d8

Adding detail

879d688

Add graphics

349fab1

Pre Dry Run Commit

17a7fc7

tweak note

2545fef

Nearly there

99d3d9f

Revision and extension

e8a300f

Blockspace market reason

dc32c77

BradleyOlson64 self-assigned this Jul 30, 2023

BradleyOlson64 requested review from Overkillus and nuke-web3 July 30, 2023 17:53

Merge branch 'main' into brad-execution-cores

17eec89

nuke-web3 mentioned this pull request Jul 31, 2023

Execution cores polish, fix build #805

Merged

nuke-web3 and others added 5 commits July 31, 2023 18:23

Execution cores polish, fix build (#805)

4878dd2

* fiixes img with closing `/>`, fmt, spelling * mv files into 7-Polkadot mod * fix `yarn build` for img paths * fit code on screen better

Merge branch 'main' of https://github.com/Polkadot-Blockchain-Academy…

70bb1e7

…/pba-content into brad-execution-cores

Fixing content reverted by Nukes correction

2d7c39f

fmt

9e9f949

Less text on slide

cc4556b

nuke-web3 approved these changes Aug 1, 2023

View reviewed changes

Overkillus reviewed Aug 2, 2023

View reviewed changes

Core time added fragments to slides

541bc0f

rphmeier reviewed Aug 3, 2023

View reviewed changes

BradleyOlson64 added 3 commits August 2, 2023 21:08

Partially address comments

3aebc59

Final touch up

e696395

Merge branch 'main' of https://github.com/Polkadot-Blockchain-Academy…

ccf2e90

…/pba-content into brad-execution-cores

BradleyOlson64 merged commit c595003 into main Aug 3, 2023

BradleyOlson64 deleted the brad-execution-cores branch August 3, 2023 05:49

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Availability Cores Talk for PBA@UCB #632

Availability Cores Talk for PBA@UCB #632

BradleyOlson64 commented Jun 6, 2023 •

edited

Loading

nuke-web3 commented Jul 31, 2023

nuke-web3 left a comment

Overkillus left a comment

Overkillus Aug 2, 2023

Overkillus Aug 2, 2023

BradleyOlson64 Aug 3, 2023

Overkillus Aug 2, 2023

Overkillus Aug 2, 2023

rphmeier Aug 3, 2023 •

edited

Loading

rphmeier Aug 3, 2023 •

edited

Loading

BradleyOlson64 Aug 3, 2023 •

edited

Loading

Overkillus Aug 2, 2023

rphmeier Aug 3, 2023

rphmeier Aug 3, 2023 •

edited

Loading

BradleyOlson64 Aug 3, 2023

BradleyOlson64 Aug 3, 2023

rphmeier Aug 3, 2023

BradleyOlson64 Aug 3, 2023

rphmeier Aug 3, 2023 •

edited

Loading


		<img rounded style="width: 1500" src="../assets/exotic_core_scheduling.svg" />

		Notes:

Availability Cores Talk for PBA@UCB #632

Availability Cores Talk for PBA@UCB #632

Conversation

BradleyOlson64 commented Jun 6, 2023 • edited Loading

nuke-web3 commented Jul 31, 2023

nuke-web3 left a comment

Choose a reason for hiding this comment

Overkillus left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rphmeier Aug 3, 2023 • edited Loading

Choose a reason for hiding this comment

rphmeier Aug 3, 2023 • edited Loading

Choose a reason for hiding this comment

BradleyOlson64 Aug 3, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rphmeier Aug 3, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rphmeier Aug 3, 2023 • edited Loading

Choose a reason for hiding this comment

BradleyOlson64 commented Jun 6, 2023 •

edited

Loading

rphmeier Aug 3, 2023 •

edited

Loading

rphmeier Aug 3, 2023 •

edited

Loading

BradleyOlson64 Aug 3, 2023 •

edited

Loading

rphmeier Aug 3, 2023 •

edited

Loading

rphmeier Aug 3, 2023 •

edited

Loading