Hi FLEDGE participants, this is @peiwenhu who is working on the FLEDGE trusted server system which will eventually replace the current BYOS model. We would like to understand more about the trusted server here to make sure the product considers your needs. If you have input outside these starter questions listed below we are also interested.

I: Dataset

1. [If you run the BYOS mode server] How often do you update the data served by your key value server in the current BYOS mode? (Every minute, hour, day, etc?)
2. How often would you prefer to update your data served by your key value server eventually?
3. [If you run the BYOS mode server] How do you perform data updates to your key value server in the current BYOS mode? (batching, streaming, etc?)
4. How would you prefer to perform data updates to your key value server eventually?
5. For data updates, roughly how much percentage of the dataset is involved each time? (average, 90 percentile, 100 percentile)
6. How much data staleness can you tolerate (from when you start to update the key/value dataset to when they are available to user devices)? What is the impact if you must accommodate data staleness > 5min? 10min?
7. How large do you expect your dataset to be?
8. Would you be able to break your data down into different shards served by different trustedBiddingSignalsUrl/trustedScoringSignalsUrl completely separately?

II: Performance

1. How much request-per-second do you expect your system to perform at?
2. Have you run into any performance issues in the current BYOS model or would you like to call out any particular areas worth paying attention to?

III: Interface

1. Chrome intends to provide the specific server that key/value data is queried from. We’re also designing peripherals that will allow other operations such as updates, versioning of the data and techniques to support high workload. Would you prefer to build the infrastructure separately or use the Chrome reference implementation of that?
2. [If you run the BYOS mode server] What debugging scenarios have you run into in the current BYOS mode?
3. What debugging functionalities would you like to have?
4. For the “data version header”: What thoughts would you have if the server decides which version of data to return, and the data of a specific version can be looked up internally but the client software cannot specify it in requests. The server may also decide to not return the version.
5. [If you run the BYOS mode server] Do you run your server with any known difference from the server outlined in the FLEDGE proposal?