Filecoin Filecoin Storage Providers, 17 Aug 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: HyperDrive Upgrade - Optimization Analysis of Gas by Athena Pool

Description

Peng Wang from Athena Pool updates the Filecoin Community at the first Storage Providers Show and Tell. With a focus on the HyperDrive upgrade, Peng updates us on the optimization analysis of gas.

A

Hello, everyone., I'm sorry. I can only speak chinese today.. My spoken english is really poor..

A

Here,, I am very happy to review with you some of our solutions.

A

To improve the gas cost of athena, pool.

A

This is what it was like before. Hyperdrive was introduced.. Let me first review some of the improvements.

A

We have made in the overall gas fees, some of the results, achieved, some official changes, and some of the benefits it has brought to us.

A

I won't go into more detail about the problem of gas fees,, which everyone has been talking about just now..

A

Let me focus on what we might...

A

We've been involved with the testnet since the testnet,, even before testnet network,, and we're pretty good at storage power, now.

A

therefore,. We always want to save the gas fee,.

A

So as to get some revenue.

A

Because,, even when gas costs were high and we were under pressure,, we were trying to sealing.

A

We did it from three aspects: mainly., I mean, before hyperdrive, we optimized gas costs.

A

From the following three aspects.

A

for example,, we established.

A

A node inspection mechanism by optimizing some systems and some workflow.

A

we regularly clean up our disk fragment,, because the gasused.

A

In the precommitsector area is closely related to the size and growth of the node.

A

Therefore, we periodically.

A

Defragment and remove some discontinuous sectors.

A

From the disk during node sealing,, which greatly reduces our gasused.

A

Second, we developed a system of our own,, which we called staggered travel..

A

Its main goal is to block the provecommitsector message.

A

We set a base fee, threshold, or a baseline,.

A

And when the value falls below that threshold, or when it is approaching the time window, we submit. by adopting this method,, we have indeed gained huge benefits.

A

About gasfee, it has peaks and troughs.

A

When it is in a downward trend or phase,.

A

We're going to get huge, benefits. thirdly, on the basis of this staggered peak system,.

A

We used machine learning to make a rough prediction.

A

Of the future trend of gasfee.

A

Based on this estimate,, we can know the base fee threshold and the time node for each of our messages to be submitted. through data analysis,.

A

We found that by using this set of technology and the staggered peak travel system, more than 90% of our messages.

A

Realized huge savings in gas, fee.

A

Of course, all three steps.

A

Are part of our efforts.

A

To reduce costs as much as possible before the release of hyperdrive., they don't solve the fundamental problems.

A

The fundamental move is still to improve it through protocol labs.

A

Let me just briefly review what protocol labs has done, and we're very pleased withsome of the things they did..

A

We are pleased to see some of the changes.

A

Let's have a review. around march,.

A

A scheme,, a temporary scheme, to be exact,, was officially introduced to refund the cost of burning windowedpost messages. of course,. After better measures came out,.

A

The proposal was cancelled.

A

The second big improvement is the off-chain validation mechanism for windowpost messages,, which should be a first in the entire blockchain.

A

When it was first launched,.

A

We found that the gasused for all messages was drastically reduced,.

A

And in particular, it brought a very significant cost reduction to windowpost.

A

Later, as the growth rate of.

A

The whole network expanded again,. This effect was reduced.

A

finally,. We got to this version of.

A

Protocol labs 'hyperdrive., we described it as an out-of-the-box version.

A

In fact, you're all familiar with how to use this version.

A

And how it works for us., let me highlight the comparisons of our sealing data here.

A

As soon as the official version was released,, we immediately started working on our updates.

A

And basically, about the provecommitaggregated way,.

A

We adopt the official recommendations,, which is to use batchings when it is above 0.15.

A

This scheme has now been adopted by most of our nodes.

A

and when we looked at the data,, both vertically with our own previous data, and horizontally with the whole network,, we found that we did get. We have indeed gained huge benefits..

A

This is our cost per t.

A

In other words,, since july,.

A

Our gas fees have decreased by at least 70 percent.

A

Let's pick another node horizontally., we found that there were some nodes.

A

That did not yet use the bulk packaging pattern.

A

for example,. Let's look at this 64p node.

A

Its daily packaging volume is about 800t.

A

Now, let's look at another node of similar size.

A

Its single tgas is around 0.17.

A

That's much higher than the average for the network and much higher than our level.

A

I don't know exactly why they haven't adopted it. Yet.

A

In fact,, we have strongly recommended using this version of bulk packaging..

A

It is true that there are many nodes that have not yet enabled this feature.

A

I'll end with a simple summary.

A

As one of the largest miners of the internet,.

A

We will closely follow protocol lab's, 'lead.

A

And continue to contribute to the filecoin network.

A

As you can see,, we have made some improvements and changes.

A

To the official code.

A

We have every reason to believe that filecoin's network will continue to grow steadily through the combined efforts of miners and protocol labs.

A

Eventually,, it will reach its final maturity after ecological landing.

A

Forgive me for speaking chinese.

A

That's all. thank you!.