SunisDown

1 HTree 是基于 Dynamo + Bitcask 两篇论文来了，同时需要作为存储引擎的索引，以及方便数据同步。 2 同上 3 Key 数据量比较多的时候可以设置 256 个，通常情况下 16 应该够了。

This can be divided into two parts 1. form an indexer cluster This can be easily obtained based on gossip. However, it is important to distinguish between indexer and searcher....

@fmassot glad to know it. thanks for your reply. > On the distribution of indexing Are you going to store the routing information in meta store and let the indexing...

It would be great if we have a control plane to schedule the indexing task. just share my thoughts about the config 1. for large cluster, we would like to...

> which is configured with a timeout Agree with the configured timeout, If we configure the client side with larger timeout, by keeping the connection alive , this endpoint can...

@PSeitz just curious, May I know why we choose to store the data in Rocksdb instead of creating an index and uploading it directly to S3? I have read several...

@guilload Thanks for your reply. I have worked storage engines and distributed storage system and I totally agree with the WAL technique for database. > When indexing errors happen, we...

> committing often hurts the indexing throughput. agree, Frequent requests for small data can affect throughput. But throughput can also be increased by resizing the request data, and if it...

> Yes that may end up with too many small splits but the concept for split merging already exists so I would think the answer to that lies in enhancements...

> I don't think a rest PushAPI that does not queue documents, produces a split right away and blocks until this is done is something someone with a lot of...