Decrease block time to 3s #3612

vncoelho · 2024-12-05T19:01:28Z

Following a recent request by TC and discussed by Core Devs, here we reduce block time to 3s.
Main modifications:

Modify config files to 3000ms
Modify MaxTransactionsPerBlock to avoid too much load in short time (optional but suggested)
Modify Extend By Factor to improve good behavior bonus avoiding ChangeViews on high latency (optional but suggested)
Modify GAS generation according to the new proposed block time

Positive aspects:

This block time of 3s is surely aligned with a NEO standards and what we can provide publicly. However, there will be more storage use, which happens in order to update unclaimed GAS (that happens every block, which could be improved in the future).

Negative aspects:

GAS Inflation can be reduced because of an increase on standard deviation on block time, consequently, reducing rewards of CNs, Committee and NEO holders.

vncoelho · 2024-12-05T19:02:40Z

Devcontainer is not working properly, I need it to run tests and adjust public void TestEconomicParameter().

Perhaps we need a Hard Fork at

                engine.SnapshotCache.Add(CreateStorageKey(Prefix_GasPerBlock).AddBigEndian(0u), new StorageItem(5 * GAS.Factor));
                // TODO - Add hardfork otherwise storage will be different
                // After HARD FORK

can you add this HF @shargon ?

roman-khimov · 2024-12-05T20:35:23Z

benchmarks/Neo.Benchmarks/config.json

-    "MillisecondsPerBlock": 15000,
-    "MaxTransactionsPerBlock": 512,
+    "MillisecondsPerBlock": 3000,
+    "MaxTransactionsPerBlock": 256,


MaxTransactionsPerBlock can be kept as is.

I think it need to be changed, I think that this comment here contradicts your comment below. I did not understand. You said to cut even more aggressively.

We need to reduce MaxTraceableBlocks, not MaxTransactionsPerBlock.

all can be reduced, each block wont be able to process that amount of transactions.

C# node can handle several thousands of simple NEP-17 transfers per second. It's all about MaxBlockSystemFee (#3552), not MaxTransactionsPerBlock.

I think that for a first release these proposed values are good.
They can be increased easily, the opposite is more complicated in my opnion.
So, I think we should just move forward.

src/Neo.CLI/config.fs.mainnet.json

src/Neo.CLI/config.json

roman-khimov · 2024-12-05T20:37:07Z

src/Neo/ProtocolSettings.cs

@@ -113,8 +113,8 @@ public record ProtocolSettings
            StandbyCommittee = Array.Empty<ECPoint>(),
            ValidatorsCount = 0,
            SeedList = Array.Empty<string>(),
-            MillisecondsPerBlock = 15000,
-            MaxTransactionsPerBlock = 512,
+            MillisecondsPerBlock = 3000,


I'm not sure we need to change defaults.

I think it is better, @roman-khimov , because it may, in the future, reflect in some of our tests that use default. Better to keep consistent.

Because we will change testnet only before, it's better to avoid default changes

roman-khimov · 2024-12-05T20:37:38Z

src/Neo/ProtocolSettings.cs

@@ -69,7 +69,7 @@ public record ProtocolSettings
        /// <summary>
        /// The maximum increment of the <see cref="Transaction.ValidUntilBlock"/> field.
        /// </summary>
-        public uint MaxValidUntilBlockIncrement => 86400000 / MillisecondsPerBlock;
+        public uint MaxValidUntilBlockIncrement => 86400000 / MillisecondsPerBlock; //TODO keep the same??


This one is configurable per-network, so I'd rather leave the default as is and configure particular networks.

roman-khimov · 2024-12-05T20:38:35Z

src/Neo.CLI/config.testnet.json

-    "MillisecondsPerBlock": 15000,
-    "MaxTransactionsPerBlock": 5000,
+    "MillisecondsPerBlock": 3000,
+    "MaxTransactionsPerBlock": 256,
    "MemoryPoolMaxTransactions": 50000,
    "MaxTraceableBlocks": 2102400,


Notice that it'd be ~73 days instead of current year. But I'd rather cut it down even more aggressively (see FS network settings).

I agree, let's say 64 or 128?

Opinions are welcome, your values remind me of E-letter chain. Nothing bad about that, but historic data (#2026) suggests our contracts are more used to somewhat longer history available. NeoFS networks use 17280 (3 days of 15s blocks).

roman-khimov · 2024-12-05T20:39:41Z

src/Neo/SmartContract/Native/NeoToken.cs

@@ -188,6 +188,9 @@ internal override ContractTask InitializeAsync(ApplicationEngine engine, Hardfor
                engine.SnapshotCache.Add(CreateStorageKey(Prefix_Committee), new StorageItem(cachedCommittee));
                engine.SnapshotCache.Add(CreateStorageKey(Prefix_VotersCount), new StorageItem(System.Array.Empty<byte>()));
                engine.SnapshotCache.Add(CreateStorageKey(Prefix_GasPerBlock).AddBigEndian(0u), new StorageItem(5 * GAS.Factor));
+                // TODO - Add hardfork otherwise storage will be different
+                // After HARD FORK
+                engine.SnapshotCache.Add(CreateStorageKey(Prefix_GasPerBlock).AddBigEndian(0u), new StorageItem(5 * GAS.Factor));


This also should be kept as is, it's a default and it affects block 0 state changes. And practically this doesn't matter, there is some default, it can be tuned for a particular network. Like people regularly run 1s networks with 5 GASPerBlock networks for tests and it's not an issue.

Maybe we need a hard fork for new default, but maybe not needed and can be kept simple as you said.

src/Plugins/DBFTPlugin/Consensus/ConsensusService.Check.cs

roman-khimov · 2024-12-05T20:43:34Z

src/Plugins/DBFTPlugin/Consensus/ConsensusService.Check.cs

-                // around 2*15/M=30.0/5 ~ 40% block time (for M=5)
-                ExtendTimerByFactor(2);
+            // around 4*3/M=12.0/5 ~ 80% block time (for M=5)
+            ExtendTimerByFactor(4);


This is a tricky one. In our tests we have not changed any of these parameters and they look scalable by their nature, if we use lower block times we expect lower waiting times in general as well. So I'd not change them unless we have some real proven need to.

This is needed, I have many tests on that for NeoCompiler.
I had also run in the past with server spread along globe.
I suggest these new parameters.
But if you guys do not want to change we can try with original. But this is my recommended for now.

@roman-khimov, you are right that this method is really tricky....I remember exactly why we created that....And that is really tricky...aheuaheua Luckly we had this idea during the creation of dBFT 2.0, otherwise (other changes would be necessary)......

roman-khimov · 2024-12-05T20:45:59Z

src/Plugins/StateService/Verification/VerificationContext.cs

@@ -27,7 +27,7 @@ namespace Neo.Plugins.StateService.Verification
 {
    class VerificationContext
    {
-        private const uint MaxValidUntilBlockIncrement = 100;
+        private const uint MaxValidUntilBlockIncrement = 100; // Change to 500!??


Whoa. Unexpected one. It's totally different from the "regular" MVUBI, but 500 here would let more ExtensiblePayloads with signatures float around the network, I doubt the intention here was to have it related to time.

We should related it with the timeblock

I think you guys should decide, that is why I left as comment.

docs/config.json.md

Co-authored-by: Christopher Schuchardt <[email protected]>

vncoelho · 2024-12-09T21:53:49Z

src/Neo/SmartContract/Native/NeoToken.cs

@@ -193,7 +193,10 @@ internal override ContractTask InitializeAsync(ApplicationEngine engine, Hardfor
                engine.SnapshotCache.Add(CreateStorageKey(Prefix_Committee), new StorageItem(cachedCommittee));
                engine.SnapshotCache.Add(_votersCount, new StorageItem(System.Array.Empty<byte>()));
                engine.SnapshotCache.Add(CreateStorageKey(Prefix_GasPerBlock).AddBigEndian(0u), new StorageItem(5 * GAS.Factor));


ping @shargon

I think that maybe it should be enabled with 1 instead of 5 after certain height.
So the hardfork will be triggered by modification the SnapshotCache withnew StorageItem(5 * GAS.Factor);

This should be moved to config.json?

We can't change it without affecting post-genesis state. And I see zero reason to do this, mainnet/testnet don't care, this will be adjusted by the committee, private networks are fine with whatever default and they can also adjust it if needed.

Jim8y · 2024-12-11T06:57:58Z

@vncoelho the base of this pr should be #3454

Jim8y · 2024-12-12T00:21:23Z

Statistics on the mainnet delay:

100ms Slots:
0-100 ms: 8.68%
100-200 ms: 8.07%
200-300 ms: 4.81%
300-400 ms: 3.34%
400-500 ms: 5.39%
500-600 ms: 6.39%
600-700 ms: 5.44%
700-800 ms: 4.43%
800-900 ms: 4.01%
900-1000 ms: 3.69%
1000-1100 ms: 4.67%
1100-1200 ms: 2.97%
1200-1300 ms: 2.14%
1300-1400 ms: 1.72%
1400-1500 ms: 1.18%
1500-1600 ms: 1.52%
1600-1700 ms: 1.86%
1700-1800 ms: 2.04%
1800-1900 ms: 2.26%
1900-2000 ms: 2.48%
2000-2100 ms: 4.58%
2100-2200 ms: 2.45%
2200-2300 ms: 1.46%
2300-2400 ms: 1.09%
2400-2500 ms: 0.67%
2500-2600 ms: 0.95%
2600-2700 ms: 0.99%
2700-2800 ms: 1.01%
2800-2900 ms: 1.08%
2900-3000 ms: 1.17%
3000-3100 ms: 2.40%
3100-3200 ms: 0.92%
3200-3300 ms: 0.48%
3300-3400 ms: 0.33%
3400-3500 ms: 0.17%
3500-3600 ms: 0.28%
3600-3700 ms: 0.22%
3700-3800 ms: 0.21%
3800-3900 ms: 0.22%
3900-4000 ms: 0.23%
4000-4100 ms: 0.51%
4100-4200 ms: 0.14%
4200-4300 ms: 0.07%
4300-4400 ms: 0.06%
4400-4500 ms: 0.03%
4500-4600 ms: 0.05%
4600-4700 ms: 0.04%
4700-4800 ms: 0.04%
4800-4900 ms: 0.04%
4900-5000 ms: 0.04%
5000-5100 ms: 0.08%

500ms Slots:
0-500 ms: 30.29%
500-1000 ms: 23.95%
1000-1500 ms: 12.69%
1500-2000 ms: 10.17%
2000-2500 ms: 10.26%
2500-3000 ms: 5.20%
3000-3500 ms: 4.31%
3500-4000 ms: 1.15%
4000-4500 ms: 0.81%
4500-5000 ms: 0.20%
5000-5500 ms: 0.14%

1000ms Slots:
0-1000 ms: 54.24%
1000-2000 ms: 22.86%
2000-3000 ms: 15.46%
3000-4000 ms: 5.47%
4000-5000 ms: 1.02%
5000-6000 ms: 0.22%

src/Neo.CLI/config.fs.mainnet.json

src/Neo.CLI/config.json

src/Neo.CLI/config.mainnet.json

shargon · 2024-12-12T08:56:49Z

src/Plugins/DBFTPlugin/Consensus/ConsensusService.OnMessage.cs

-            // around 2*15/M=30.0/5 ~ 40% block time (for M=5)
-            ExtendTimerByFactor(2);
+            // around 4*3/M=12.0/5 ~ 80% block time (for M=5)
+            ExtendTimerByFactor(4);


This should be computed according the configured time per block

vncoelho added 2 commits December 5, 2024 15:51

First commit and no format

b5fd4ec

Start adjusting param but devcontainer not working

91c4bd2

vncoelho requested review from shargon, cschuchardt88, Jim8y and roman-khimov December 5, 2024 19:01

roman-khimov reviewed Dec 5, 2024

View reviewed changes

vncoelho and others added 2 commits December 6, 2024 00:03

Merge branch 'master' into decreaseblocktime-3s

9ba85d8

Merge branch 'master' into decreaseblocktime-3s

a095496

cschuchardt88 reviewed Dec 7, 2024

View reviewed changes

docs/config.json.md Outdated Show resolved Hide resolved

Update docs/config.json.md

3329c54

Co-authored-by: Christopher Schuchardt <[email protected]>

cschuchardt88 added the Waiting for Review label Dec 7, 2024

Merge branch 'master' into decreaseblocktime-3s

774471a

vncoelho commented Dec 9, 2024

View reviewed changes

vncoelho added 2 commits December 9, 2024 18:59

format

739beaf

Merge branch 'master' into decreaseblocktime-3s

af42801

Jim8y mentioned this pull request Dec 11, 2024

Add hardfork HF_Echidna #3454

Draft

10 tasks

AnnaShaleva mentioned this pull request Dec 11, 2024

Decrease block time to 3 seconds on N3 networks nspcc-dev/neo-go#3745

Open

Merge branch 'master' into decreaseblocktime-3s

9d677c5

shargon reviewed Dec 12, 2024

View reviewed changes

src/Neo.CLI/config.fs.mainnet.json Outdated Show resolved Hide resolved

Update src/Neo.CLI/config.fs.mainnet.json

cd91440

shargon reviewed Dec 12, 2024

View reviewed changes

src/Neo.CLI/config.json Outdated Show resolved Hide resolved

Update src/Neo.CLI/config.json

26b1e58

shargon reviewed Dec 12, 2024

View reviewed changes

src/Neo.CLI/config.mainnet.json Outdated Show resolved Hide resolved

Update src/Neo.CLI/config.mainnet.json

3b52a3c

shargon reviewed Dec 12, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Decrease block time to 3s #3612

Decrease block time to 3s #3612

vncoelho commented Dec 5, 2024 •

edited

Loading

vncoelho commented Dec 5, 2024

roman-khimov Dec 5, 2024

vncoelho Dec 6, 2024

AnnaShaleva Dec 6, 2024

Jim8y Dec 7, 2024

roman-khimov Dec 7, 2024

vncoelho Dec 9, 2024

roman-khimov Dec 5, 2024

vncoelho Dec 6, 2024

shargon Dec 12, 2024

roman-khimov Dec 5, 2024

vncoelho Dec 6, 2024

roman-khimov Dec 5, 2024

vncoelho Dec 6, 2024

roman-khimov Dec 9, 2024

roman-khimov Dec 5, 2024

vncoelho Dec 6, 2024

roman-khimov Dec 5, 2024

vncoelho Dec 6, 2024 •

edited

Loading

vncoelho Dec 6, 2024 •

edited

Loading

roman-khimov Dec 5, 2024

shargon Dec 5, 2024

vncoelho Dec 6, 2024

vncoelho Dec 9, 2024

vncoelho Dec 9, 2024

shargon Dec 12, 2024

roman-khimov Dec 12, 2024

Jim8y commented Dec 11, 2024

Jim8y commented Dec 12, 2024

shargon Dec 12, 2024

Decrease block time to 3s #3612

Are you sure you want to change the base?

Decrease block time to 3s #3612

Conversation

vncoelho commented Dec 5, 2024 • edited Loading

vncoelho commented Dec 5, 2024

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

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

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

vncoelho Dec 6, 2024 • edited Loading

Choose a reason for hiding this comment

vncoelho Dec 6, 2024 • 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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Jim8y commented Dec 11, 2024

Jim8y commented Dec 12, 2024

Choose a reason for hiding this comment

vncoelho commented Dec 5, 2024 •

edited

Loading

vncoelho Dec 6, 2024 •

edited

Loading

vncoelho Dec 6, 2024 •

edited

Loading