How to join the Filecoin test? The latest official release of must-read information!

Original link: https://filecoin.io/blog/filecoin-testnet-mining/

Becoming a Filecoin miner provides community members with an opportunity to help store and preserve access to humanity's most important information. By providing network security and infrastructure, Filecoin miners can help build a more open, resilient, and trustworthy Internet.

Filecoin mining is a novel opportunity to earn revenue by providing data storage and retrieval services to clients across a vast network. Miners earn fees by storing and retrieving client information on the Filecoin network. The more client data a Filecoin miner stores, the greater the storage power of the miner. More storage power on the network means a particular miner is more likely to generate blocks and win block rewards. Miners can choose whether to participate in storage power consensus, storage mining, and/or retrieval mining. We expect many miners to provide targeted, specialized services in these categories.

The mining hardware you place on the Filecoin network will provide useful data storage services to clients in addition to providing network security and batch production. This distinguishes mining on the Filecoin network from traditional proof-of-work networks.

Reallocating excess storage capacity to run a Filecoin node transforms your once latent storage assets into a secure storage service. The block production, storage, and retrieval services provided by miners are invaluable to Filecoin’s ecosystem. By bringing together many miners and storage providers in a decentralized network, Filecoin unites storage providers in the first-ever algorithmic marketplace for renting data storage.

Ready to hear more? Keep reading to learn how to get started.

To join the upcoming Filecoin testnet, you’ll need suitable hardware and a Lotus implementation installed and setup. For more information on installation and setup, see the Lotus docs.

Next, you will need to stake Filecoin tokens as collateral to set up a miner. “Pledged Collateral” is the initial amount of Filecoins required to join the network as a storage miner. Over time, miners must also deposit more “deal collateral” for each additional storage deal they enter with a client. Collateral is deployed on the network to ensure miners behave properly and is denominated in FIL tokens. Details on collateral will be finalized in the Filecoin spec in the coming weeks. During the testnet, you can withdraw test Filecoins from our network faucet to use as collateral to start mining. Test Filecoins have no value - the official Filecoin token will not be released until after the mainnet launch.

(Please read carefully)

Note: This section describes the standard test configuration that the Protocol Labs team uses to test Filecoin storage mining on the Lotus implementation. There is no guarantee that this test configuration will work for Filecoin storage mining at mainnet launch. If you need to purchase new hardware to join the Filecoin testnet, it is recommended that you purchase no more hardware than you need for testing.

Our standard test configuration is one of many possibilities for Filecoin testnet mining. We hope that testnet miners will find other hardware combinations, some of which may be more efficient than the test configuration described below. If you use a different hardware configuration to run Lotus, please run the Lotus benchmark tool and post the results in this thread.

The sector size and minimum guaranteed storage required to mine a block are two very important Filecoin testnet parameters that influence hardware decisions. We will continue to refine these parameters (and other proof parameters) during the testnet. Therefore, our standard testing configuration for Filecoin mainnet may change.

Sector size: If you choose to purchase hardware to test on the Filecoin testnet, we recommend that you purchase computing hardware that supports sealing 32GiB sectors. The Testnet will temporarily support 1GiB "test sectors" so that miners can easily join the network. Such a small sector size is not realistic for the mainnet. After the testnet is launched, we will perform a fork upgrade to stop supporting 1GiB "test sectors" and will announce the timing and mechanism of the upgrade on our blog.

Miner size: To be eligible to mine blocks during the testnet (not just accept storage deals), miners must pledge at least 512GiB of disk storage. This requirement may increase for the Filecoin mainnet. If you only want to participate in storage deals, you can join the network with less storage. However, if your current pledge to the network is less than 512GiB, your verified storage will not count toward the network's total power and will not entitle you to significant rewards.

Below is a configuration we have tested for sealing 32GiB sectors on Lotus.

This configuration is based on the steps required in the Filecoin mining process. We made specific tradeoffs related to computing power in the standard test configuration (i.e., we allocated resources to GPUs instead of more CPU power). It is possible to mine on the testnet using other configurations, but since we have not tested all of them, we will not discuss their feasibility or performance in this article.

We are testing powerful CPUs (with at least 8 cores) to replicate client data during PoRep. Powerful CPUs allow miners to process data faster and generate PoRep faster, gaining power on the network.

The replication process utilizes fast disks to offload data from memory to disk. For this purpose, we have tested an SSD (512GB) that can be used as a large cache to store unsealed in-flight sectors (during the pre-commit and commit phases of PoRep) and cached during the sealing process.

Our standard configuration also runs on a lot of RAM (128GB). RAM is needed to compute in-memory data replication and currently has a 2-3x relationship with sector size. The circuits we design to run during SNARK generation for PoRep and PoSt are very large and consume a lot of RAM. These large circuits minimize the footprint of the SNARK proof output for efficient on-chain submission.

We also use powerful GPU power to generate SNARKs in a more time-efficient and cost-effective manner. SNARKs are generated on the PoRep proof of each sealed replica, and the SNARK output is submitted to the chain. We specifically note that the Lotus implementation is designed for GPU chips manufactured by NVIDIA, however, we expect support for cards from other manufacturers in the future. As we continue to tune the proof parameters, we expect the mainnet circuit to require more computing power than the testnet circuit.

We have observed in testing that having multiple “sealers” working on a storage range can be more efficient than running the entire Filecoin mining process (sealing and PoSt) on a single computer. This allows miners to efficiently run multiple processes at once.

For more information on how the Proof of Replication process relates to mining hardware, see the diagram below:

Our test configuration takes storage overhead into account – we currently estimate the overhead to be 2x the committed storage, but this may change for mainnet. We are currently running ~1.5TB of storage in disk storage in sealed storage, but we believe there are more efficient ways to configure sealed storage to reduce overhead. We are also testing fast parallel disk access to sealed sectors to generate tickets to potentially win election tickets for block rewards in ElectionPost.

When running post processing, our current test configuration stores sealed sectors on hard disk, operating constantly (24/7) due to the frequent read submissions.

A SNARK is run in Election Spacetime (“ElectionPoSt”) only when a miner has won an election ticket due to winning a block reward in a given epoch. Our configuration uses the power of the GPU to compute the output of the PoSt SNARK whenever a miner wins an election ticket, accurately computing the PoSt SNARK within each epoch period (~20 seconds). The GPU power required for PoSt depends largely on how many winning election tickets a miner has in any given epoch, which is related to their storage ratio in the network.

We also recommend using a lower latency network connection to submit PoSt on time. If a miner fails to submit PoSt on time multiple times, it may result in winning less block rewards and increase the risk of being slashed.

For more information on the ElectionPoSt process in relation to mining hardware, see the diagram below:

Remember that Filecoin mining is based on storage power consensus: having more power (proven storage) on the network directly correlates to winning more block rewards. The amount of power a miner puts on the network scales linearly with the amount of storage the miner has on the network, not with the number of GPUs the miner is running. In contrast, in Proof-of-Work GPU mining, miners must compete on GPU power to win block rewards. In Filecoin mining, miners compete on storage power consensus, not GPU power. Miners only use GPUs in ElectionPoSt if they win election tickets. The best and cheapest way to win power is to add more useful storage to the Filecoin network.

If you are a smaller miner concerned about the cost of GPUs, rest assured that we are working to enable miners to securely outsource SNARK computations to third-party service providers.

Now that you’ve got the gears covered, you’re ready to put them to work! You’ll have the opportunity to join the Lotus testnet soon.

If you want to learn more details about the components of the Filecoin network, check out the Filecoin specification. Have questions about the topics discussed in this article? Please ask them in the appropriate category in our forum. For real-time troubleshooting on setting up a node, connecting to the Filecoin testnet, and more, use the help channel in Filecoin chat.

Thank you for supporting our team!