[DataCap Refresh] <3rd> Review of <StorSwift>

[spaceT9]

What type of Audit did you opt into? Choose from dropdown

Public Open Data

What is your JSON hash (starts with 'rec')

reclqkAojcbQlyMKJ

Link to your allocator operations description

filecoin-project/Allocator-Registry#981

Compliance Report

https://datacapstats.io/allocators/f03019866/reports/904eb983-3c7b-40db-95ca-2b38818cbac5

Previous Reviews

• [1st review] <link to the review> Outcome: <10> PiB
• [2nd review] <link to the review> Outcome: <10> PiB

Client Allocation Table

Client name	DC granted
Xinzhou	3PiB
Jingshengya	6 PiB

Client-specific Breakdown

Client 1

I. Xinzhou

• DC requested: 10 PiB
• DC granted so far: 3.72 PiB

II. Dataset Completion
(https://www.xinzhou-sg.com/category/xuanchuan
https://www.xinzhou-sg.com/category/pinpai
https://www.xinzhou-sg.com/category/guanggao)

III. SP Match
( Basically match the list in the CID report)

IV. Replica Count
Promised vs delivered: 10 VS 10

V. SP Retrieval Performance

---

Client 2

I. JIngshengya

• DC requested: 15 PiB
• DC granted so far: 9.52 PiB

II. Dataset Completion
(http://wx.jingshengya.net/)

III. SP Match
(Basically match the list in the CID report)

IV. Replica Count
Promised vs delivered: 10 Vs 9

V. SP Retrieval Performance

---

(Add more client blocks as needed)

Notes from the Allocator

Due to a recent malfunction in the Spark service, the audit did not focus on the Spark retrieval success rate, but rather on business license information, data size, data preparation details, index files, and the number of replicas. Future audits will gradually shift back to the Spark retrieval rate. Of course, during the audit process, we will also randomly check the manual retrieval situation of SPs, and generally speaking, most SPs support manual retrieval.

Issues or discrepancies reported?

During the allocator audit process, the governance team provided a Datacap recharge limit of 10 PiB, but only 5 PiB of Datacap was actually received. We can submit a ticket for processing (fidlabs/allocator-tooling#160). Currently 10 PiB has been recharged.

Steps to minimize unfair or risky practices

• The datacap-bot developed by the official team provides excellent guidance.
• Pay attention to the size of the datasets.
• Implement stricter KYC, focusing on business license information or personal identification documents for further KYC verification.
• Monitor the SP's sealing process at all times and ask questions to clients.
• Request the index files.
• Place greater emphasis on the data preparation process.
• Focus on the distribution of SPs.

Value to the Filecoin ecosystem

This round primarily serves two clients: an existing client, Jingshengya, and a new client, Xinzhou, representing the education and film/media industries, respectively. This enriches the diversity of Filecoin data storage.

Confirm you maintained your application standards for all client allocations

Yes

Do you understand that a diligence review will follow and require updates to this issue?

Yes

[filecoin-watchdog]

[DataCap Application] Xinzhou # 34

• http://154.201.1.144:7843/piece/baga6ea4seaqfcjrk4h2u3nhyvei7mkt43gms7el32lpfoniu4vjrrgram5mroei
  After 18% of the download, the server returned 404 Not Found, an issue repeated several times at different times of the day.
• The smart contract application should be started as soon as possible, as it is well documented.
• Without a real download, it is not possible to confirm whether the dataset is similar to those indicated by the similarity bot. Many applications were labeled as “media, promotional videos, etc.”
• Since the dataset (as declared) contains video materials, was the right to use this content and make it publicly retrievable confirmed in any way?
• No index or mapping file could be found to help identify what exactly is in each piece—both in terms of file names and broader context. However, even if such a file existed, without a full HTTP download and processing, it could not be properly compared with exact samples of the data.
• Why was the application migrated from [DataCap Application] <Xinzhou> StorSwift-team/fil-allocator-bookkeeping#32, and what will happen with the DataCap already assigned there? (A smart contract would allow a decrease if necessary.)
• Evidence of client verification was provided.
• The justification for dataset size is only in the form of screenshots.
• Spark retrievability and RPA values are very low.
• Some nodes hold significantly more data than others, with values ranging from 4% to 18%.
• As this application belongs to Round 6, no more than 8 replicas should be allowed, and the standard tranche schedule should apply, which does not allow more than 256 TiB of data distribution in the first round.

[DataCap Application] Jingshengya # 8

• http://120.48.133.192:58421/piece/baga6ea4seaqgatmruf6w7nlzc2n3jqcfq4qbkt4rdk7h3vdltfkklvefdf7s6ha
  After 18% of the download, the server returned 404 Not Found, an issue repeated several times during testing.
• Most nodes are underperforming in Spark retrievability, and real download attempts confirmed access problems.
• The user exceeded the declared unique dataset size, storing 1.65 PiB while declaring 1,500 TiB.
• No problems were observed with IPNI reporting.
• Evidence of client verification was provided.
• The justification for dataset size is only in the form of screenshots.

[Amin-Foundation]

Transparency & Community Engagement: The allocator’s team continues to demonstrate responsiveness and operational transparency, consistently addressing requests for clarification and complying with proposed improvements. Client diligence remains high, with KYC/KYB validation practices. While there is no significant step-change from the previous cycle in this area, the allocator’s continued proactive engagement reflects positively on their operational credibility.

Data Distribution: Geographical distribution remains strong, with Storage Providers (SPs) spanning multiple continents and avoiding concentration risk. DataCap allocation is evenly distributed among SPs, with no single provider receiving more than 4% of the total distribution.

Replication & Redundancy: Replication levels remain within acceptable compliance limits. All current clients report below the required maximum copies threshold of eight, though variations exist between instances - for round 6, the replicas counts will be expected to be reduced to respect no more than 8 so the team will not be penalized in this area. SPs carry more replicated data than others.

Retrievability & Quality: Retrievability remains critically low at 25% (showing a decline from the previous 29%) and falling far short of the 75% compliance benchmark. A limited number of SPs report compliant levels of retrievability, but these are offset by several non-reporting instances, pulling the average down.
Data quality is also of concern as no file indexes were provided for spot-auditing stored content, making it impossible to verify accessibility and audit-readiness. Immediate corrective action is required in this area. Sustained failure to address retrievability and indexing issues will jeopardize future allocation eligibility.

DataCap Usage Analysis: Tranche allocation velocity remains consistent with expectations, with no irregularities in schedule or distribution patterns. However, a notable concentration of allocations was observed: one client (f03327650) received three tranches of 2 PiB each during July. While overall balance is preserved at the SP level, this client-level concentration warrants monitoring to ensure distribution diversity.

Notable Flags

• Low retrievability performance (25%, down from 29% in previous cycle)
• Absence of file indexing, echoing past concerns over data quality and accessibility

There is a slight degradation in both replication and retrieval performance over this last refresh cycle. The lack of progress on data quality improvements, and the decline in retrieval success rate are significant concerns. Without corrective measures, the allocator’s eligibility for continued DataCap allocation is at risk.
We recommend throttling the allocation to 5 PiB for the upcoming cycle, with the expectation that the team addresses the flagged issues.

[spaceT9]

Thank governance team for the review. We have received all feedback and will work hard to improve the relevant statistics in the following round.

[Amin-Foundation]

Top-up confirmed for 5 PiB
Transaction ID - 0xfa8b1047f0095c2db6f82a19afe58faf574e20789a918b5215da909cb5730ad7