Editors Note: This article comes fromFirst class warehouse blockchain research institute (ID: first_vip1), reprinted by Odaily with authorization.
, reprinted by Odaily with authorization.
At present, there are two main long-term views on Filecoin, the king of storage (Ace Pigeon), in the market: bearish and bullish, but many of them are simply talking about long and short, and whats more, it is illogical to brag about FILs long-term Value—unfortunately, a complete logic has not been seen so far.
Its time to pour cold water on the storage king Filecoin with reason and evidence.
secondary title
Preface: Exploration of Decentralized Cloud Storage
In the decentralized cloud storage track, there are currently only four first-class warehouses that can store things: Filecoin, Storj, Arweave, and SiaCoin.
SiaCoin is over-centralized (there are only two mining pools in total, and one of them is closely related to the project party, see the first-class warehouse wealth code No. 13 for details), and there are serious problems in the incentive mechanism, which is no longer worthy of attention. For the remaining three, Arweave actually set up a separate portal to make permanent storage of the Alexander Library. This road and ceiling must be considered separately, and the valuation must also be calculated separately. Strictly speaking, Arweave and Filecoin are not suitable for Horizontal comparison.
And the heavyweight players who can really be regarded as head-to-head PK on the main battlefield of the decentralized cloud storage track are actually only Filecoin and Storj.
There is no doubt that they have chosen two completely different directions in the matter of decentralized cloud storage:
At all costs, Filecoin must first realize the perfect decentralization in the minds of blockchain geeks, regardless of storage performance and experience; while Storj focuses on providing high SLA service standards, high Stable performance to meet the high requirements of enterprise-level storage, and reduce high costs to meet the discerning practical needs of the long-tail enterprise market.
Or to put it simply, one of them pursues better decentralization, and the other pursues better cloud storage, so have they really achieved it?
Both of these two projects have been seriously working for several years. First Class Warehouse has conducted many internal brainstorming sessions. Next, we will start with an all-round analysis of Filecoin and Storj from the perspective of decentralization and core storage business. For comparison, the core of this article is divided into two parts. The first half makes a comprehensive comparison as much as possible through more than ten aspects such as the degree of decentralization, token model, incentive mechanism, mining machine threshold, product positioning, etc. The second half is to make a horizontal comparison of the capability level of the storage business.
(During several days of actual testing, the first-class warehouse found that due to the major flaws in the incentive mechanism of Filecoin, miners have a relatively high risk of centralization, which is explained in detail in points 2 and 16.)
secondary title
1. In-depth comparison between Storj and Filecoin in terms of degree of decentralization, incentive mechanism, token model, product positioning, etc.
1. Product positioning: Filecoin has a grand vision, but there is a big gap in reality; Storj has a small vision, but it is currently well implemented
The main positioning of Filecoin is to serve as the incentive layer of IPFS to solve the current efficiency problems of IPFS. The core point is decentralization. The main target group should be those decentralized geeks and DAPPs who pursue the freedom of data ownership. We dont think that traditional storage needs will be transferred to the Filecoin network at this stage, because the results of the actual measurement of Filecoins storage capacity are really surprising.
The main positioning of Storj is to benchmark Amazon S3s decentralized cloud storage platform, which is similar to Ubers role. It integrates and utilizes idle resources to output a standard and unified product. Its main user group is all users who use Amazons S3 storage service. Storj Provide the same SLA service standard, faster upload and download speed, better privacy, decentralization, and the price is only half of Amazons similar products.
In general, Filecoin has a grander vision, but it has just been launched, and its storage service capabilities are relatively poor. At the beginning of Storjs planning, Amazons S3 service standard was regarded as a milestone, and it was not officially launched until the milestone was reached. Therefore, Storj is obviously doing better in commercial cloud storage, and the entire access experience is closer to the storage of real users. Demand, storage and deletion, fast speed, flexible supply, and how much you pay for the service you use.
2. The degree of network decentralization: Although each has its own merits, in general, Filecoin is better
The relationship between the three roles of Storj is as follows: the client issues a request and uploads and downloads data, the node is responsible for storing and downloading data to the client, and the satellite assumes a series of important roles: when uploading and downloading, it selects the best A fast node records expenditure and income for both the user terminal and the node, records the location of the users stored files, and stores the reputation data of the node-you can think that the satellite has established a set of credit points similar to Alipay for the node. The credit system is used to manage storage nodes. It is undeniable that the existence of satellites makes Storj relatively centralized in the network structure. At the same time, satellites also have the risk of single point failure. That is to say, if a satellite is attacked or explodes, all the content stored through the satellite will be lost. However, Storjs official explanation for this is: the satellite is not a single server, but a server cluster, which will not easily fail, and there will be plans to open the satellite in the future, so that anyone can build a satellite.
image description
In terms of storage, Filecoins roles include storage users, storage miners, and retrieval miners. The role of retrieval miners is a bit like a satellite, but the difference is that retrieval miners are completely decentralized, and anyone can become a Retrieve the miner. Retrieval miners do not audit storage nodes like satellites, and retrieval miners do not audit Filecoin storage nodes. From the perspective of this network structure, we think Filecoin will be more decentralized.
Figure 2: The architecture diagram in the Filecoin white paper, showing the miner market and retrieval market structure
image description
Figure 3: Simplified schematic diagram of Figure 2
The storage process of Filecoin is as follows: users can choose the miners they want from a group of storage miners to place an order, and pay the order fee at the same time, and then the miners receive the order to store and seal the sector; when retrieving, the retrieval miners are automatically assigned, and the retrieval miners retrieve After the data is received, it is sent to the user, and the user pays the retrieval fee while sending the data.
Although we think that from the perspective of architecture design, Filecoin will be more decentralized, but in terms of the degree of decentralization, has Filecoin properly won?
In addition, from the point of view of node distribution, as shown in the figure below, Filecoin nodes are distributed all over the world, which is relatively balanced, and there are many in Asia; while Storj is spread all over Europe, North America, South America, and Australia, but there are relatively few in Asia, especially in China. There are less than 100 nodes, which directly leads to the fact that the upload and download speed in China is significantly lower than that in Europe and the United States. From this point of view, in terms of the distribution of nodes, we believe that Filecoin is better.
Figure 4: Storjs node distribution map, you can see the sparse node distribution in Asia, the number of nodes is 8000+
image description
Figure 5: Filecoins miner node distribution map, as of 2020/10/27, the number of miners is 685
Therefore, whether the network architecture is decentralized or not is a round, and each has its own winners and losers. From the perspective of network structure, Filecoin obviously won, but from the perspective of the number of nodes, Storj won, and from the perspective of node distribution, Filecoin has won again.
But here we need to add a major hidden danger in the design of the Filecoin mechanism that we think:
Since encapsulating sectors requires large calculations, the threshold for Filecoins mining machines is extremely high, which determines that the number of Filecoins miner nodes must be limited, and most of them are concentrated in the data center computer room. Suppose a user has stored 100 movies, each movie is 10GB, and a total of 1TB of data is stored in a miner’s hard drive. When 1,000 users want to watch these movies at the same time, the traffic generated may reach 1,000MB/S. The bandwidth of the miners will become a bottleneck, because high bandwidth means high cost at the same time, and the download link, as a storage miner, provides download bandwidth and traffic with zero income, and only retrieval miners will charge a sporadic retrieval fee According to the data, the retrieval fee is far lower than (one in thousands) bandwidth expenditure. Although Filecoin has a design to provide a part of the block rewards to the retrieval miners in the future, it is still just talk on paper. We It is believed that the bandwidth bottleneck may become the Achilles heel of Filecoin storage. In the future, due to the low packaging capacity of small miners (see points 4-7 for details), large miners will package files faster, so most of the user file storage will flow to large miners, which will also aggravate the outbreak of this bottleneck problem. We will continue to focus on observation.
3. Token model: Filecoin wins without any suspense
Filecoin’s token model, combined with the storage business, we consider a business subsidy model, where the system issues tokens to subsidize miners’ high mining machine costs, allowing miners to issue tokens through the system (commonly known as mining rewards) To quickly recover the cost of hardware. In the end, investors in the secondary market will take over and maintain a high currency price to pay for the cost. This is a typical strategy where the wool is on the pig. When miners recover hardware costs through subsidized (mining) token income, they no longer need to worry about hardware costs, so miners can provide external cloud storage services regardless of cost. This is the essence of the Filecoin economic model that first-class warehouses think about . This kind of economic model is very advantageous in terms of storage price. It’s like the owner of an online celebrity beef miscellaneous store in Guangzhou. He charged you five yuan for a large bowl of beef miscellaneous, because there are 10 buildings in his house to be demolished. In rent collection, people dont rely on grocery stores to make money. We believe that such an economic model has great advantages in providing cloud storage products with prices far below the market price. In addition, it has achieved good results in motivating miners and expanding the scale of the network. However, in terms of mining mechanism design, it is obviously too ruthless for miners. This is another story (details in point 4). Overall, we believe that Filecoins token mechanism design should be better than Storj.
Storjs token mechanism design does not have much design, it simply uses Storj as the companys securities + currency to circulate. Storjs economic model design is still based on the traditional concept of doing business. That is, a user deposits 1TB of data, and after passing through erasure code redundancy, it will be distributed to 80 miners for storage, occupying a total of 2.7 miners. TB of space, then charge users $10, give miners $1.5/TB*2.7 TB=$4.05, and then give $1 of the remaining $4 to partners, and keep $3 for the team. The remaining $2 is for some marketing or miscellaneous expenses. This is a very traditional economic model design. The relationship between the growth of the business and the increase in the price of the token is relatively weak. It can be said that Storjs token model design is not closely integrated with business.
As a project established at the same time as Ethereum, the Storj project’s token model and economic model were designed in the early days of the blockchain 2.0 development. Now it seems that it is relatively Old School (entering middle age). Although it is sincere, it is not sexy. . The first-class warehouse has also made suggestions to the team, but the team obviously has no plans to change the model.
In this round, no matter how many problems the Filecoin token model has, Filecoin wins.
4. Miner incentive mechanism: Filecoin is ruthless to miners, but miners can make money; Storj is gentle to miners, but miners cannot make money
Storjs miners (nodes) correspond to storage miners in Filecoin.
Regarding the miner incentive mechanism, we understand its goal as: to encourage miners to stay online in a long-term and stable manner to provide stable storage services. Based on this goal, Filecoin and Storj also adopt the pledge method to guarantee, but Filecoin uses a pre-pledge, that is, before the miners work, they need to buy coins to pledge first, and then pay the wages after the work is completed. However, Storj uses post-staking, that is, a part of the wage income of the miners is pledged. The former is more ruthless to the miners, and the latter is more gentle to the miners. The following are the details.
Storjs incentives for miners include several parts:
A set of node reputation system, allowing miners to prove that they can be online for a long time and provide high-performance services at any time (more than 95% of the time is online) + miners must prove that they have stored data + at any time to prove that they have saved data intact (not lost or damaged) , so that you can or have the opportunity to earn income. This process lasts for several months. Miners with poor performance will be eliminated, and qualified miners will receive income. The income includes content storage fees + upload and download traffic fees;
Nodes with higher configuration and faster response speed will be given priority, and can obtain more data and higher income;
In the first 1-9 months after joining the mining, part of the miners income will be pledged in the system. Miners should try their best to keep operating for at least 15 months in order to get full income. If the miners quit after less than 15 months of operation, or casually disconnected after 15 months or were fired by the satellite for poor performance, then some The income will be deducted, which is equivalent to using the previous income as a pledge. Having said so much, in fact, after the authors actual measurement, basically as long as it is not the kind that drops out every day, it is not a big problem to simply be a miner, and you dont need to worry about the equipment being turned on.
Through such a set of mechanisms, Storj is obviously encouraging miners to try their best to operate for at least 15 months, which greatly reduces the loss rate of miners, and the lower the loss rate of nodes, the higher the durability of customer files (the less likely they are lost). Of course, it doesn’t matter if some nodes suddenly go down. Each file has 80 fragments distributed all over the world. Even if a war tsunami occurs in a certain place and a batch of nodes are lost, the remaining nodes will automatically repair other fragments to maintain The files are available. Currently, the Storj network stores 170 million files, and the number of lost files is 0.
Filecoins storage miner incentive mechanism is not particularly different from Storj, although the technical path used is different:
Through the system, miners also have to prove that they have stored data (replication proof PpRep) + every 30 seconds to prove that the data they have stored is still and complete, not lost or damaged (time and space proof Post), and wait for users to pick up files at any time, but Miners dont need to accumulate credits for many months like Storj miners.
If the miner is offline for a short time or the storage sector is faulty, he will be fined on a daily basis, and he will also be fined if he forgets to declare his fault; if the fault lasts for more than 14 days, the sector will be stopped and he will be fined 90 days Block Rewards - Miners can only get back all the FIL pre-staked if they complete the promised order cycle.
Filecoin motivates miners to stay online in this way, which is the same as Storjs original intention, and the punishment is not light. Can motivate miners online to a certain extent.
The different parts of miner incentives mainly lie in pledge, income rules and mining machine subsidies: currently the effective computing power of Filecoins entire network is 660PiB, and the amount of FIL pledged is 13.6 million. On average, for every 1TiB of computing power, an average of 20.1 FILs are mortgaged. According to the latest data statistics, at present, every increase of 1TiB of computing power actually requires an average of 11.18 FIL to be mortgaged, and the reward distribution mechanism is block reward + storage fee, and the block reward is simple casting 30% + benchmark casting 70% , the block reward will increase with the increase of the effective adoption rate of the network. The current block reward is 11.37FIL/block, but it needs to be distributed linearly to the miners within 180 days (every day’s income is distributed to the next 180 days on average), so that Incentivize miners to operate for a long time. Currently, according to the update of the FIP-004 proposal, in order to reduce the pressure on miners to pledge, mining rewards are changed to release 25% of block rewards first, and the remaining 75% still need to be distributed linearly to miners within 180 days.
At the same time, because the system temporarily cannot distinguish between the data generated by the miners themselves and the real data of the actual users, Filecoin has designed several rules such as verified users and committed capacity to encourage miners to store real data as much as possible.
Verified users refer to users who have been verified to have real data needs. Storing such user data will bring 10 times the income of other data to storage miners. The relevant details of verifying users are described in detail in the chapter Point 9: Authenticity of stored data volume in this article.
In terms of incentives for miners, Storjs incentive mechanism can be described as simple and clear, and this set of rules can well serve the goal of high performance and high reliability of the system; while Filecoins rules have made great efforts to serve decentralization. , the architecture has also achieved decentralization, but there are no rules that can prevent miners from swiping data, refusing to store orders, and other negative or malicious behaviors, nor can they motivate miners to provide good enough cloud storage services. Miners are all profit-seeking Yes, the design of the mechanism makes miners prefer to provide storage space (sector encapsulation) instead of providing storage services (miners can reject orders, miners do not need to guarantee bandwidth, and miners are not responsible for transmission speed). From this point of view, although both Both of them have achieved the purpose of mechanism design to a certain extent, but the storage service provided by Storj will be more stable. The upload and download test of point 16 at the end of the article also verified this problem. However, although Filecoin is more ruthless to miners and has a strict punishment mechanism, it cannot support FIL to make more money. The scale effect allows miners to invest a lot and generate a lot of output. Storj pursues the decentralization of the network, so the government limits the development of miners. Scale, to meet the growth rate of storage demand, so Storj’s miners can’t invest in a large scale, they can only invest in a scattered manner. A NAS will be recognized as the same node in the Storj network, so it will only allocate data to you according to one node, which means that miners cannot invest in a large scale, and at the same time, they cannot achieve large-scale benefits. Therefore, from the perspective of miners’ income, although Filecoin is ruthless, it can make money; although Storj is gentle, it can only earn a small amount of money and cannot make a fortune.
5. Mining machine threshold: Storj is much lower than Filecoin
Storj’s mining machine has a relatively low threshold. The minimum requirements are: one processor core dedicated to one node, one 500G hard drive, 2TB of broadband traffic per month, at least 5MB/S uplink and 25MB/S downlink. Such a threshold is so low that almost every household can install their computer as a Storj node. The minimum cost can be 800 tuples for a NAS. Currently, I have installed several nodes for testing.
Filecoins mining machine threshold is much higher in comparison: the official basic equipment requirements for miner nodes: AMDs 8+ core CPU, at least 128GB of RAM (memory), and a powerful GPU (taking the average number of cores as 3000 , currently generally use 2080Ti/3080/3090 graphics cards), solid-state drives larger than 512GB (choose 1TB SSD NMVe interface). At present, the cost of a single Filecoin mining machine in Chengdu is 100,000+.
Obviously, the threshold for Filecoin mining machines is much higher than that of Storj. From an economic point of view, this will obviously lead to fewer people who can buy mining machines as miners. The direct result is that the number of nodes will inevitably be small. We expect Filecoin nodes to The quantity will be 1-2 orders of magnitude less than Storj. The current data is that the number of Filecoin nodes is 600+, and that of Storj is 8000+. Fewer nodes will make Filecoin far weaker than Storj in terms of traffic distribution and CDN service capabilities, and once Filecoin storage applications are rolled out, Filecoin large miners will also have greater traffic pressure.
6. File retrieval and deletion: Storj wins
Storj will generate a metadata for files uploaded to the network, which is to find the address of the storage location of the file. This address is stored in the satellite. Depending on this address, the satellite can retrieve the file. This process is very fast and simple. At the same time, operations such as deletion, transfer, and copying can also be performed. According to the actual measurement of the first-class warehouse, it only takes 3 seconds to delete a file from Storj. Storjs retrieval fee is a very small number ($0.0000022/time), which can be ignored unless it is retrieved on a daily basis.
Filecoins file retrieval mainly relies on the retrieval miners in the network. The retrieval miners issue retrieval requests on gossipsub according to the file information provided by users, and then obtain the files from the storage miners and send some files and pricing requirements to the user. Finally, the user confirms the order and pays It is relatively convenient for users.
However, because Filecoin packs and encapsulates files, each sector is folded layer by layer into a huge data crystal. Deleting data from it is equivalent to dismantling the deeply folded warehouse layer by layer, finding an object and throwing it out. Difficult - Sector packing takes 40-50 minutes. Therefore, the files in Filecoin cannot be deleted by the user at present. If the file needs to be deleted, it can only be completed through the negotiation between the user and the storage miner. First Class Warehouse consulted the official. Regarding the file deletion, the official answer is that it has to wait until the order for the file arrives. After the period, the file can be deleted. At the same time, the storage time of Filecoin starts from 180 days, because the minimum life cycle of a sector is 180 days, that is, as long as you store data in Filecoin, it must be stored for at least 180 days, and the storage fee for 180 days must be paid by the user in one lump sum.
In addition, after the file is uploaded, Filecoin needs to wait for the miner to successfully encapsulate the file into the sector before retrieving and downloading it. The encapsulation time mainly depends on the hardware capabilities of the miners. If the small miners are slow, it will take a few days, and if the big miners are fast, it will take an hour. At present, the head position has been tested, and the fastest is to upload at night and encapsulate into the sector the next morning. Moreover, when Filecoin retrieves (downloads) files, the delay is relatively severe, and the network is also unstable. Usually, it takes at least a dozen seconds, or at least three minutes, before the download starts. However, when Storj downloads files, it starts downloading directly, and has no sense of delay.
In addition, because Filecoin is a public chain, an order must be submitted to the miners every time it is uploaded and downloaded, and the fee will be paid at the same time when the order is submitted. Therefore, to pay successfully, it is necessary to ensure that the block is synchronized to the latest block, so the current Under normal circumstances, to upload and download things to the Filecoin network, you must first synchronize dozens of GB of block data, and the block data is still growing at a rate of 400M per day. This is a fatal injury that currently affects the popularity of Filecoin applications, so Filecoin officials also have plans to develop light nodes that do not need to synchronize the complete blockchain. I believe this problem should be solved. Storj, which is an ERC20 token, has no similar problems.
Judging from the above rules and actual test results, it is obvious that Storj will be closer to the actual storage user needs than Filecoin in object storage services, store and retrieve at any time, fast retrieval, flexible fee settlement, and deduct as much space and traffic as you use cost. Relatively speaking, whether Filecoin can be accessed or withdrawn depends on whether the block synchronization is completed, and the fee settlement overlord clause, apart from anything else, a 180-day fee will be charged first, and the stored things cannot be deleted. On the whole, we think that Storj wins in this link.
7. File encryption and privacy: default encryption VS default non-encryption, Storj wins
Storj uses client-side encryption. All data that needs to be stored in the network is encrypted by the client and then uploaded. The encrypted data needs to be decrypted with a key before viewing, and the key is stored in the local computer where the client is located. , cannot be obtained from the network. That is to say, even if the users data can be found through satellites, there is no way to know what is inside, and the files are still divided into 80 copies, distributed to the hosts of 80 miners, and the data of at least 29 miners must be obtained at the same time. After getting the complete file, you have to use the key to decrypt it, which basically eliminates the possibility of data leakage.
The data that Filecoin needs to upload is not encrypted by default, that is, whether it is encrypted or not depends on the user. Users have the possibility to upload unencrypted data. In addition, Filecoin is a single miner to a single user, or multiple copies of multiple miners to a single user. The users data is directly exposed to the miners, and as long as the CID of the file is known, the file can be obtained arbitrarily. Relatively, it is much worse. It can be said that the mechanism design of Filecoin does not consider too many privacy requirements, and only allows users to encrypt files by themselves.
Therefore, we believe that in terms of privacy, Storj wins without a doubt.
8. Data maintenance: Storj is even better
The highlight of Storjs data maintenance is the use of erasure code technology. When each file is uploaded, the data will be redundant by 2.7 times, and 80 fragments will be generated and stored in 80 nodes. Any 29 ore fragments can generate a complete file. The satellite will check from time to time to determine whether the 80 fragments are still there. And as time goes by, the number of their damage and loss, once the number of existing fragments drops below 50, the satellite will start data repair, get 29 fragments from the existing nodes to regenerate the file, and then slice to form the lost fragments and store them in other nodes to ensure that the network has a sufficient number of shards. The user is insensitive to the whole process.
The data of Filecoin is maintained by miners, and the role of maintenance miners is also played by other storage miners. Users store files on multiple miners. If the data of one miner is damaged, other miners with the same data will perform file transfer repair. (If the user only saves one copy, then there will be no maintenance miners, and if it is lost, it will be lost)
In terms of data maintenance, there is no doubt that Storj wins with its thoughtful and perfect design. The mechanism of Filecoin is completely dependent on the user to choose the data redundancy multiple and punish the miners under extremely strict conditions to ensure the durability of the file. Due to the strict punishment mechanism, miners will indeed actively maintain the security of files, but natural disasters and man-made disasters are inevitable. Alipay can be dug by forklifts, not to mention these miners who are concentrated in the data center? At present, the number of Storj files lost is 0, and Filecoin has just been launched, and the actual effect will take time to test, but in terms of user experience and mechanism design, we think Storj will be even better.
9. The authenticity of stored data volume: Storj is slightly better
Looking at this issue from the results, it is somewhat unfair, because Storj was launched half a year earlier than Filecoin, and there are already 8PB of real business data on it. Filecoin has only been online for half a month, and it is not yet full. So we discuss this issue, not from the results, but mainly from the mechanism.
In terms of mechanism, Storj is a decentralized storage. When users upload files, they don’t know which nodes the files will be stored on (and they will be distributed to you all over the world with a high probability), so there is no way to store them for themselves, and the uploading process There is a huge gap between direct cost and income. Users pay $10/month for uploading 1TB, while miners only earn $1.5/month for storing 1TB. Therefore, miners have no incentive to upload false files.
But this is not the case with Filecoin. Storage users can directly designate miners to store their own files, and the cost of storing 1TB files is less than $3 per month. If the miners encapsulate the 1TB files into sectors, then 6.6 FIL can be mined in a month, and the current price is $21.7. In addition, in order to encourage miners to store valid data, the official proposed the concept of verified user. Encapsulating these effective data can get a 10-fold computing power bonus. In this way, encapsulating 1TB of data can get 210 dollars per month. Such a mechanism will definitely make the miners work hard to pretend to be a verified user who needs a lot of storage. As long as they pass the community certification, they can get a large amount of credit, and then designate the miners to save it for themselves. This is perfect.
The following is a brief introduction to the community certification consecration process of the verified user of community governance:
The entire governance is divided into three roles: root key holder, notary, and customer group. These roles can be anyone. The root key holder is elected by the community, and its main task is to hold the root key and unconditionally implement community resolutions. The notary is also elected by the community, and its main responsibility is to allocate capacity quotas to customers who apply for quotas, and to apply for more quotas from the community. After the community votes, the quota will be opened by the root key holder. After the quota is opened, it can be allocated to customers for effective storage.
10. The theoretical vulnerability of the network (whether it is easy to be attacked): Storj is better
Filecoin miners are vulnerable to attacks. Miners are vulnerable to DDOS attacks and the bandwidth is full. If the miners are attacked and cannot submit storage proofs to the network in time, the miners will be fined and deducted 90 days of block rewards after more than 14 days , and the pledged tokens are deducted at the same time, the storage sector is regarded as an invalid sector, and the allocated computing power of the sector is cleared. At this time, the miner can choose to discard the file, which means a great risk of data loss for the user . In the testnet stage, there were cases where miners’ computing power was cleared to zero due to attacks.
This kind of attack is not a problem for Storjs miners. Any operation of the Storj client on the node needs to be carried out through the satellite, and even if individual nodes are attacked, it will not affect the retrieval of user data unless the 80% of the user data is stored. Among the nodes, more than 51 nodes were attacked at the same time, resulting in data that could not be retrieved or repaired. This difficulty is obviously much more difficult than attacking a Filecoin miner alone. The biggest risk of Storj is not the node, but the satellite. However, the satellite is not a single server, but a server cluster. The anti-attack capability of the cluster is obviously stronger than that of a single server.
And because Filecoins user storage data is stored by a single miner, once the miners network is delayed or is attacked by traffic, it is difficult to retrieve user data. Therefore, for useful data, users still have to choose multiple miners for storage.
However, Storjs data is stored in fragments. As long as 29 of the 80 nodes are online normally, the files can be retrieved completely. Relatively speaking, the reliability will be much higher than Filecoin.
11. Degree of community support: Juan, you should hire more community managers
This is a point I really dont want to mention, but Filecoin is really bad at this aspect, so I want to take this opportunity to attract official attention.
The main community of Filecoin is Slack. In Slack, if you consult information about current mechanism issues, system bugs, update issues, or miners’ mining questions, the response speed is relatively slow, basically concentrated at one point in time, and in many cases there is even no According to the response situation, it seems that there is only one community leader who mainly answers the questions in the community. Most of the time, colleagues in the chat room come out to answer questions, but some questions have not been officially confirmed, so we dare not believe them casually. ! I want to say that Filecoin has also raised a lot of money. I heard that there are hundreds of people in the development team alone. Why dont you recruit a few more people to properly answer the communitys questions? This virtually dissuades many people who want to know about Filecoin. I cant understand it myself, and I cant ask in the community.
In addition to Storj’s regular quarterly municipal meetings, there are Twitter, forums, and support platforms to submit questions. During the process of testing the Storj network, First Class Warehouse encountered some problems. Consult the official on these platforms, and almost all questions will be answered in the Someone replied the next day. Judging from the situation of the replies, several people were replying to questions in various fields. part of data.
12. Financial transparency: the two sides are almost the same
Storj will announce the token expenditure and lock-up status on a quarterly basis, and its token lock-up status can also be queried on the chain, but it has not announced the whereabouts of its legal currency fundraising funds. It is a relatively transparent project. According to the current official quarterly financial report, the coins in the hands of the team are enough for him to spend more than ten years.
The Filecoin blog did not disclose the use of funds, and the ICO totaled $257 million. We dont know how much money the team currently has. If anyone has relevant data, please provide it.
13. Market value price expectations: this is not important
In fact, we don’t want to do too much interpretation on this part, but simply talking about the results would be suspected of shouting orders, so let’s explain a little bit.
In the case of Filecoin, it can only fall all the way for the time being. At present, FIL can maintain a high price because of the design of the pledge mechanism, which forces miners to go to the market to increase their computing power. But this is definitely unhealthy, there is no sufficient, free market game, and there is no sufficient exchange of chips. It is unscientific to rely solely on pledges and lock-ups to lock in liquidity and lock in high prices and high market value. Such a mechanism , If FIL is allowed to maintain an overall market value of tens of billions of dollars all the way, then there will be no investment opportunities, just watch the show with the leeks. It’s just that as the mined coins continue to be released and the entire circulation continues to increase, we believe that the price will still go down slowly. It will eventually reach a balance, that is, the willingness of miners to buy coins for pledge, and the willingness of coin holders to sell To balance the cash out, focus on the comparison between token release speed and pledge speed. After the lending market goes online, you need to pay attention to the interest rate and loan amount in the lending market. In the long run, we are still optimistic that Filecoin will have the opportunity to solve some problems in the existing storage field, but if the market value of Filecoin today is 60 billion dollars, and the market value of Filecoin 5 years later is 120 billion dollars, then for you and me, there is another Whats the point. Isnt it nice to buy pancakes?
secondary title
2. PK of core business: who can provide better cloud storage
To answer the question of who can provide better cloud storage, we first need to understand the service standard SLA of cloud storage, which is currently the gold standard for global commercial cloud storage. We use the current SLA service standard of Amazon S3 cloud storage as the standard line to measure the performance of both parties in the core business of cloud storage (however, we will not list the specific content of Amazon SLA. If you are interested, you can go to the official website of Amazon. Check). Amazon S3 SLA mainly includes several dimensions:
14. Storage durability and availability: Storj wins
Storj mainly uses erasure code technology to solve the problems of durability and usability. The erasure code technology mainly divides files into 80 equal parts with 2.67 times redundancy, and distributes them to 80 nodes on the network. Only 29 of the nodes need to be online to restore complete files. Through such a mechanism, Storj has achieved 99.99999% file durability. Currently, 170 million files are stored, and the number of lost files is 0. Because the Storj V3 network design was designed to match the SLA service standard of Amazon S3 cloud storage at the beginning, Storj will be relatively superior in this part of its capabilities.
Filecoin currently relies on a severe punishment mechanism for miners in terms of file durability and usability. Every time a miner encapsulates a sector, a large amount of FIL tokens must be pledged on the sector. If the sector is damaged or lost, the miner will The pledged tokens will be confiscated to motivate miners to provide high-durability and high-availability storage services. However, at present, sector damage and data loss cannot be avoided. In addition, in the mechanism designed by Filecoin, as long as the user divides the file into multiple miners and stores one copy separately, as long as the data is lost, these miners can repair the file with each other, but the premise still requires the user to be more redundant. Therefore, we believe that Filecoin is currently not guaranteed in terms of file durability and availability. To store high-value data, users need to perform multiple redundancy by themselves, which directly causes the storage cost to increase exponentially. Moreover, the effect of multiple redundancy is not as secure as that of Storj, which disperses 80 fragments and stores them separately.
15. Storage pricing and mechanism: Filecoin is definitely cheaper
Storjs pricing is based on the service price of Amazons S3 storage, combined with the main components of its network to form idle hardware resources, and due to erasure codes, Storj will automatically have 2.7 times the redundancy of stored content. In summary , Storj’s storage service price is half of Amazon’s, the price of storing 1TB of data is $10/month, and the data traffic fee is $45/TB, while the price Storj pays to miners is $1.5/month for 1TB of data, traffic fee It is $20/TB. Since Storjs main target is To B users, it adopts a centralized pricing method, pricing in legal currency, avoiding token price fluctuations, and providing a unified market price.
The pricing of Filecoin is determined by market factors. The miners and the storage service provide quotations respectively, and finally the gossip will match the transactions between the two parties. After matching, the storage users and miners will automatically connect. Users upload data and miners store it. The pricing method uses FIL to set the price. The advantage of this method is that the price of the final storage service is determined by market supply and demand. From the current point of view, the price is $0.67 per month for storing 1TB of data, which is one-thirtieth of Amazon’s price. Prices are very competitive. But the disadvantage is that due to the use of FIL pricing, and the pricing rules are determined by the relationship between market supply and demand, therefore, in the current situation that Filecoin has not been online for long enough, the price of tokens fluctuates greatly, and the market game between storage supply and demand is not sufficient, so the storage service The price fluctuates greatly, and it is only suitable for some storage tests for the time being, and it is not suitable for B-end users to carry out actual large-scale storage. Moreover, according to the author’s actual measurement, the pricing of each miner is very different, and the difference is even a hundred times higher. Such a mechanism is only theoretically feasible, and the actual situation can’t help but make people raise a row of question marks!
In addition, from the comparison of the two models, Storj focuses on using idle resources, turning waste into treasure, and providing cloud storage products with high SLA standards through erasure codes and storage node management rules. Storj still uses a comparison The traditional business logic is to provide cloud storage services. The business logic is: Suppose Storj purchases 2,700TB of capacity per month from storage miners at a cost of $1.5/TB, and provides it to users with storage needs at a cost of $10/TB. Sell 1000 TB at a price (the erasure code generates 2.7 times redundancy), and the spread in the middle is the money earned by the team. We believe that the method of wholesale at a low price, selling at a higher price, and earning a spread makes Storj look very traditional.
However, Filecoin issues tokens through the system, and the tokens are used to pay the fees to the miners. In the end, the token trading market pays for the price of the tokens, so that the wool is on the pigs, and investors in the secondary market spend money to buy tokens. To pay the high cost of mining equipment for miners, so miners can provide cloud storage service quotations regardless of equipment costs, which we call subsidies, and the system issues tokens to subsidize miners to provide low-cost storage, and ultimately market investors pay for it . Such a token economic model is designed ingeniously, which effectively transfers costs and improves the price competitiveness of storage service products. However, we believe that FIL is still risky. First, the current Filecoin storage node hardware threshold is high, resulting in extremely high hardware costs. The hardware cost of the same unit of storage capacity is more than ten times that of Amazon. The main cost is for mining It is obviously unreasonable to use mines instead of providing better storage. This part is ultimately paid by investors, but we think it is unsustainable. Protocol Labs should eventually find a way to balance hardware costs and mining security, otherwise Filecoins price advantage will be difficult to sustain.
16. Storage upload and download speed: I have to say, Storj still wins
In terms of mechanism, Storj uses erasure code technology. The file download process is to select 35 fastest nodes from 80 storage nodes for competitive transmission, and finally the client retrieves data from the 29 fastest nodes. debris. Storj is fast. First of all, multi-point transmission is faster than single-point transmission. One problem with centralized cloud storage is to transfer files from a single point to the client when downloading. The transmission speed is very limited by the bandwidth. Storj uses multi-point transmission to disperse The way of bandwidth demand solves the problem of limited bandwidth at a single point—the bandwidth of 29 nodes is used to share the transmission traffic. The Filecoin mechanism allows users to choose a single or multiple nodes for storage, but the actual download is still downloaded from a single node, so its speed will be limited by the bandwidth of a single node.
The second is node competition. Due to the existence of node competition, Storjs nodes are forced to maintain high bandwidth to ensure their high upload and download speeds. Every time a user uploads data, there are 130 nodes competing to store data fragments, among which 80 nodes that complete storage faster can get this part of data and subsequent income, and the other 50 get nothing. When users download data, some slower nodes will also get nothing. Therefore, the faster the upload and download speed of the node, the more it can guarantee the opportunity to store data fragments/transmit data fragments and earn income. This ensures the overall upload and download speed of the network.
The following is the actual uploading and downloading speed of Filecoin and Storj in order to obtain the actual uploading and downloading speed of First Class Warehouse. We must first emphasize that we only need to understand the approximate order of magnitude of the speed of both parties. We do not require the accuracy of the experimental data. The test process is divided into two parts. One part is the domestic test based on home bandwidth. The second part is to rent a foreign server and test based on foreign network conditions. In order to let everyone better experience the two networks There is a difference in storage capacity, so we restore every detail of the test process as much as possible, and also leave information such as the CID of the test file. You can check it in the block browser. The entire test process is as follows:
1) Local testing
The authors coordinates are in Xiamen, China (because the first-class warehouse is in Xiamen), and the test file size is 1006MB using a home 200Mb telecom broadband.
Filecoin test:
File fingerprint CID generated after importing into Lotus: bafykbzaceb7jcjsnozpv45wpqyqrzilmehqgzin5ozhfwfvfbu7rn7oon3iu2
Randomly selected several suitable miners f08157, f03488, f09675 from Slate.host
Select miner f09675 for the first time to submit storage order (Deal)
Generate order CID after 30 seconds: bafyreiheedhw7veoryrux6pznhrpagl32v5sb26gelhcpozwqwgits3dwq
But about 6 minutes later, it was found that the Deal disappeared. The possible argument is that the miner f09675 refused the transaction.
So the first commit fails.
Select miner f08157 for the second time to submit storage order (Deal)
It took a total of 17 minutes and 19 seconds from the submission of the storage order (Deal) to the final StorageDealTransferring (the file transfer was completed to the miners). The specific steps are as follows:
Step 1: Submit the storage order (Deal), and it takes 1 minute and 14 seconds to generate the Deal CID. The generated order CID (DEAL CID): bafyreidk5lzgtagrromd3qs4rk7akd57keaz7jlj5bibu7jhiy3wguf6sm
Step 2: The status of the order changes to StorageDealClientFunding, which means that the miners are checking the balance, checking the order and so on. 4 minutes and 48 seconds
Step 3: File transfer, it took 11 minutes and 14 seconds
Step 4: The Deal id is generated: 961927, and the Deal status is displayed as StorageDealSealing (under packaging)
If only the third step file transfer time is used to calculate the file upload speed, the average speed should be 1.49MB/S
If the time from Deal submission to file transfer completion is taken, the average speed is 0.97MB/S
It is up to you to choose which speed is more representative of the upload speed of this test. It has to be reminded that the mechanism of Filecoin single-point (user) to single-point (miner) transmission determines the speed, not whether the Filecoin network is awesome or not, but whether the network connection between the storage user and the miner it chooses is smooth and high-speed. , so this test can only represent the authors own network situation and the network connectivity of the miners I randomly selected. It cannot represent whether the Filecoin network is awesome or not. Everyones test speed may be different. I learned from the official Slack forum Some peoples speed is as fast as ten MB, while others speed is as slow as 250KB. Everyone treats it rationally.
The rest is to wait for the miners to pack into the sector. The expected time is several hours to a few days, and this period of time will not be counted.
(After waiting for two days, it still has not been encapsulated into the sector.)
After waiting for two days, miner f08157 still has not successfully encapsulated the file, so he cannot directly test the download speed of the miner’s node, so he has to download the same file from other miners who have successfully encapsulated it for testing. At present, the selected miner f01782 has completed the packaging, and the following test retrieves (downloads) files from the miners node.
The first time I ran the retrieval command, I only successfully retrieved 1.05MB of data, and then got stuck for 10 minutes, and the retrieval failed
Run the retrieval command for the second time, showing Recv: 0 B, stuck for more than ten minutes, retrieval failed
Run the retrieve command for the third time, the result is the same as above, it got stuck, and the retrieval failed. I found out the reason. It turned out that the block synchronization of my local Lotus was stuck, and the latest block was not synchronized.
After a whole night of tossing, the block height has not been synchronized, and the block synchronization speed is obviously lower than the generation speed of new blocks. From this point of view, the block synchronization problem of the test network has not been solved well. Obviously the synchronization was good yesterday, and there are obviously many domestic nodes, so why is the synchronization speed so slow today, no matter what you do. And as long as the blocks are not synchronized, the file cannot be downloaded. I went to the official forum to find a solution, and the result was to change the computer and use AMDs CPU+NVMe solid-state drive (I am so poor, there is no way). Finally, as a last resort, I reinstalled everything again. After a few hours, I finally successfully synchronized to the latest block, and successfully carried out the fourth download test.
The fourth time I ran the retrieval command, I was stuck in the DealStatusCreatingChannel link for 3 minutes and 13 seconds. I thought it might be blocked by a wall, so I tried to turn on the VPN for 20 seconds, then turned off the VPN, and then went to the next step smoothly to start the transmission The file and file transfer process took a total of 3 minutes and 28 seconds, with an average speed of 4.83MB/S. In order to make myself look more rigorous, I downloaded and tested it again. The result was that the download process took 9 minutes and 11 seconds, with an average speed of 4.83MB/S. The speed is 1.82MB/S. In order to save some face for Filecoin, I still take 4.83MB/S as the result of this download test. So far, after three failures and the fourth success, this download test is finally completed.
Storj speed test: (under the same hardware and network conditions)
Storjs upload test is relatively simple, I chose https://asia-east-1.tardigrade.io/ Asian satellite. After registering above, get the API, and then download Filezilla. Using Filezilla, you can directly connect to Storjs Tardigrade network, and the operation is relatively simple.
The upload process is to configure Filezilla, select the file to be uploaded, and click upload directly. The uploaded file is the same 1006MB file as the Filecoin test. The upload time is 14 minutes and 27 seconds, and the average speed is 1.16MB/S. However, during the upload process, Every upload of 64MB files will stop for about 1 minute. The main reason is that Storj needs to cut the file into 64MB segments, and then perform erasure code redundancy and divide it into 80 small parts. These times slow down Storjs upload speed. So it took more than 14 minutes.
In addition, due to the fact that there are too few nodes in Asia, when uploading large files in China, Storj’s upload is not stable. I uploaded it 5 times and only succeeded once. Most of them are stuck when uploading to 60%-80%, and then the feedback timeout fails.
The download process is relatively simple, just select the file directly, click download, the download takes 2 minutes and 21 seconds, the average speed is 7.13MB/S, and the download is relatively stable, and the direct download is successful after several attempts.
2) Foreign server test
The foreign server uses Vultrs highest configuration of Bare Metal, located in Singapore, with 4 cores and 8 threads of CPU, and 32GB of memory. The file is still the same file, the size is 1006MB, and the file CID remains unchanged.
Filecoin test:
For the first time, a suitable miner was randomly selected from Slate.host f020904 to submit a storage order (Deal)
After about 30 seconds, the Deal CID is generated: bafyreicnzkb7csxxqwaqsbzxieoktlsk7ro42ad6slgncfoqmkmsmwg2ui
About 4 minutes and 30 seconds later, after the order status changed to StorageDealFundsEnsured, the order disappeared inexplicably. It turns out that the order submission failed!
Randomly select the miner f09569 for the second time
Step 1: Deal CID is generated in about 28 seconds
Step 2: It’s just a flash, it’s too fast to see clearly, let’s think it’s only one second
Step 3: Deal enters the StorageDealTransferring state and starts to transfer data, but this link took 32 minutes and 15 seconds (shocking)
Step 4: After the file transfer is completed, a Deal id is generated: 962551 and the Deal status is displayed as StorageDealSealing (under packaging)
Calculated according to the 32-minute test file transfer time, the file upload speed is 0.52MB/S. I feel that I may have encountered a hot chicken miner, or I am not good enough with this miner. In order not to wrong Filecoin, I decided to re- Choose a miner and test again.
The third time I couldn’t take it anymore, I decided to choose a big miner with a reputation, I used the query-ask command to inquire about the top five miners in terms of computing power, only f01782 could get the price, the other miners no matter whether they used Domestic clients, or foreign clients, did not respond to inquiries. Could it be that everyone does not open the door to receive user data? As a Filecoin miner, if your hard drive is only used for mining, what is the difference between you and BHD? Despise it first!
Step 1: Submit a storage order (Deal)
Generate order CID after 27 seconds: bafyreiczimrrwlhv2v2csn7bqmnq7yivtq4st4i3eyybkpiun3esxejnhy
Step 2: It’s just a flash, it’s too fast to see clearly, let’s think it’s only one second
Step 3: Deal enters the StorageDealTransferring (data transmission) state and starts to transfer data. The whole process took 2 minutes and 31 seconds
Step 4: After the file transfer is completed, a Deal id is generated: 962675 and the order status is displayed as StorageDealSealing (in packaging)
(After getting up the next day, the Deal status is already StorageDealActive, and the package is successful)
Calculated from the third test result, if the upload time is calculated as 2 minutes and 31 seconds, the average upload speed is 6.66MB/S. If the upload time is calculated from the time of submitting the Deal, the duration is 2 minutes and 59 seconds, and the average upload speed It is 5.62MB/S.
To download the file is to run the retrieve command to retrieve the file from the miner f01782
After the first file retrieval, only 10MB of files were retrieved, an error was reported and timed out, and the retrieval failed
The second file retrieval took 3 minutes and 53 seconds. Average speed 4.32MB/S
The author downloaded another 1.03GB file from the node of miner f020904, which was previously uploaded for testing. It took 1 minute and 34 seconds to download the file, with an average speed of 11.21MB/S. I think this should also be one of the problems of Filecoin. That is, the storage performance and bandwidth provided by different miners are inconsistent. There is a big difference.
Storj speed test: (under the same hardware and network conditions)
Using a foreign server to upload files to Storj, the upload speed is obviously much faster. I still choose https://asia-east-1.tardigrade.io/ Asian satellite, and directly configure the uplink to the server. Start speed test:
Upload test: select the same file, the size is 1006MB, the upload time is 58 seconds, and the average speed is 17.34MB/S
Download test: Download the same file to the server, the size is 1006MB, the download time is 1 minute and 45 seconds, and the average speed is 9.58MB/S
The final summary test results are as follows:
Domestic environment: Filecoin upload speed: 1.49MB/S, download speed: 4.83MB/S
Storj upload speed: 1.16MB/S, download speed: 7.13MB/S
Foreign environment: Filecoin upload speed: 6.66MB/S, download speed: 4.32MB/S
Judging from the above test process, the performance of Filecoin is obviously unstable, with good times and bad times. The upload and download speed depends on the miner you choose. Storj is relatively stable overall. However, the stability of Storj in China is also relatively poor, especially the domestic upload of large files is prone to failure, but except for the slow domestic upload speed, the other three items are better than Filecoin. From the experience of the entire operation, Filecoin makes me feel that its target should not be Amazon Cloud Storage. It is a storage layer lower than Amazon Cloud Storage. It focuses on decentralization, because Amazon Cloud Storage Storage focuses not only on storage, but also on user experience. However, Filecoin focuses on better decentralization and does not care about storage experience. In our understanding, Filecoin should hope that ecological participants can build a more user-friendly storage service experience platform through off-chain methods , such as the official launch of Slate, Fleek Space and other applications, but how effective they are and whether developers pay for them depends on the subsequent business development. With the current storage characteristics of Filecoin, we cant imagine how developers can use the Filecoin network as a storage layer. For example, file storage must choose miners; submit storage orders to limit storage for at least 180 days; content cannot be deleted after uploading; anyone can obtain files ;etc.
secondary title
After the full in-depth PK is completed, lets go back and think about the topic, who is the better decentralized cloud storage platform?
We think it is more appropriate to divide this question into two parts. The first part is who can be more decentralized, and the second part is who is a better cloud storage platform.
Regarding the question of who is more decentralized, the first-class warehouse’s point of view is that in terms of mechanism design and network architecture, Filecoin is more decentralized, and in terms of the number of nodes, file fragmentation and erasure of decentralized storage In terms of code rule design, Storj is more decentralized. If you have to choose one of the two, then we think Filecoin will be better.
As for who is the better cloud storage platform, we think that Storj has done far more than the current Filecoin in cloud storage. Regarding the provision of high-speed and stable storage services, I think the advantages and disadvantages of Filecoin and Storj are obvious. Due to Storjs high SLA service standards, the unified performance of the entire network guaranteed by the erasure code mechanism and High durability and high availability, the upload and download speed is only affected by the users local network conditions, which is more conducive to storing useful data (I dare to save it myself). However, Filecoin is currently unifying the performance and service standards of miners. This matter is flawed. For B-end users, if there is important data that needs to be stored on Filecoin, then it is really necessary to choose a good miner from China. One is not enough , possibly three.
