2nd Community Review of Marshall Fil+ Allocator

[Marshall-btc]

https://compliance.allocator.tech/report/f03012911/1730766444/report.md

1. SPs have a good success rate for Spark retrieval, with high ones reaching 95% or more.
   
   

2. However, there is not enough variety of customers, 1 enterprise dataset and 2 public datasets. I plan to follow up by encouraging more enterprise datasets to apply.

I will be planning to attend the Bangkok November Web3 meetup, my goal is to find 1-2 enterprise customers with real stored data and get them to try to enter the Filecoin space, looking forward to our efforts and progress.

I would like to get more guidance on Filecoin Allocator's work, thank you very much.

[filecoin-watchdog]

@Marshall-btc
Allocator Application
Compliance Report
1st Review
1st Review score: 5PiB

5 PiB granted to existing clients:

Client Name	DC	status
USANASA	2PiB	Existing
LiZhu Pharmaceutical Group Co.	2PiB	Existing
NationalAstronomical Observatories, Chinese Academy of Sciences	1PiB	Existing

USANASA

• The client introduced 13 new SPs in this round(20 total), while only 6 were updated in the GitHub issue. Allocator did ask for the reason for doing so.
• 4SPs have bad retrieval(~0%), while the rest have mixed; oscillating between 30-95%.

LiZhu Pharmaceutical Group Co

• The client introduced 15 new SPs in this round(23 total) and updated some in the GitHub issue. Allocator did ask for the reason for doing so.
  The following SPs were not updated in the GitHub issue:
  f03178077
  f03179572
  f03178144
  f03214937
  f03229933
  f03173127
• 3SPs has bad retrieval(~0%), while the rest have mixed; oscillating between 30-98%.

NationalAstronomical Observatories, Chinese Academy of Sciences

• The client introduced 13 new SPs in this round(20 total) and updated 7 in the GitHub issue.
• The Client declared 6 replicas, while 8 of them already exist.
• 3SPs has bad retrieval(~0%), 3SPs uses DDO and the retrieval is unavailable, the rest have mixed; oscillating between 30-98%.

---

Overall, the allocator has implemented a new process of reviewing a random sample of SPs by asking their clients to prove the geographic location of selected SPs.
For all applications, the allocation schedule was compliant with the allocator application. The exception is the allocation at National Astronomical Observatories, where the allocator ran out of DC.

[Marshall-btc]

It's really amazing! Thank you @filecoin-watchdog for your objective and fair review.
Overall, our allocator operations have been positively acknowledged and praised by you. Based on your review, I believe there are mainly three areas that need improvement, which will also be the focus of my future review work.

1. Some of the newly introduced SPs were not updated on GitHub in a timely manner.
Solution: Although most of the newly added SPs were queried in github. However, since inadvertent oversight is inevitable, we will review each Datacap before sending it out to avoid omissions. In the meantime, we will ask Client to rectify the situation by e-mail, and to disclose the information in advance before starting each new package.

2.A small number of SPs had a low success rate of Spark retrieval;
Solution: Although some SPs have explained that they used the DDO order mode, to prevent SPs from cheating, we will limit and stop cooperation with SPs using DDO deals. Despite the fact that the Spark retrieval rate for well-performing SPs has exceeded 90%, for SPs with a Spark retrieval rate of 30%-50%, our team will assist in connecting and communicating with experienced successful SPs for guidance. If they still cannot improve the Spark retrieval rate, we will restrict and stop further approval of Datacap.

3. "The Client declared 6 replicas, while 8 of them already exist."
Solution: @filecoin-watchdog Shockingly diligent work! This is indeed an area that I overlooked in my previous review, and I will focus on it in the future. In the meantime, I will ask the other party to perform further explanation.

Additionally, regarding your mention of "For all applications, the allocation schedule was compliant with the allocator application. The exception is the allocation at National Astronomical Observatories, where the allocator ran out of DC."
Explanation: For all Datacap allocations, we are following the rules. However, for these two SPs f01518369 and f01889668, the Spark retrieval rate was only 0% without using the DDO order mode, so I asked the client for further explanation, and the client's explanation was that the data center malfunctioned, resulting in a broken network. Out of compliance caution, I limited the original 2 PiB Datacap approval to a 1 PiB Datacap approval.

Finally, as I mentioned earlier, our team traveled to Thailand for the Web 3 event in November. We met with a number of enterprise customers who have data storage needs and we plan to recommend them to apply for our Allocator application in the next 1-2 months, as well as introduce them to a number of reputable SPs.
Based on our past good performance in Allocator operation, we hope to get 10 - 15 PiB Datacap support. Thanks again to the official team for their hard work!

[Marshall-btc]

Marshall-btc/Marshall-Fil-Data-Pathway#5 (comment)

The previously mentioned issue with the number of copies was reasonably explained by the customer and I assessed it as understandable. I will also focus on the backup issue in the follow-up.

[Marshall-btc]

As I said, I have sent an email to the client about the important things about Data cap. Reminding them of the precautions again.