[Disclaimer: CockroachDB engineer here, working on performance and benchmarking]
IIRC, in the case that aphyr refers to for these specific numbers, the reads are scans that span multiple shards[1], while the writes are writes to single shards.
[1] even though aphyr says it's just a hundred rows, the tables are split into multiple shards because aphyr in this case was specifically testing our correctness in multi-shard contention scenarios. In production you wouldn't have multi-shard reads crop up until you were doing scans for tables that were hundreds of thousands of rows in size[2]. It's easier to picture if you think of this performance speed difference in the scenario where you are doing full table scans spanning multiple shards on multiple computers while the underlying rows are being rapidly mutated by contending write transactions. The transactions get constantly aborted to avoid giving a non-serializable result, and performance is suffering. We agree that the numbers in this contention scenario are too low, and we are actively working on high-contention performance (and performance in general) leading up to our 1.0 release[3].
[2] Specifically, we break up into multiple shards when a single shard exceeds 64mb in size.
The benchmark isn't quite apples to apples. I'm used to people doing put/get benchmarks, not put/search workflows. Merging results from multiple machines is nothing to sneeze at, especially if you have replication between nodes.
I worked for a search engine dotbomb, and I had a constant worry that the core engineering team had not solved a very similar problem. But since the funding round failed I'll never know for sure what they had planned.
Are read-only transactions handled without contention? I understand that a read-write transaction needs to be aborted if a later conflicting write is committed. But read-only transactions can view a snapshot of what was committed before they begin - you trade a little bit of latency for being conflict free. It doesn't even have to be the default as long as you offer it as an option, you might want this if you know you are reading hot keys, or as a fallback if you've been aborted X times.
This was IMO the big innovation of true time in spanner - contention free reads at either now, or at some time in the past.
Yes, if your timestamp is far enough in the past (this is determined by the maximum clock offset configured for the cluster), and if you don't collide with writes of a long-running transaction (which may run at a timestamp that could influence what you read), you will not have to worry about contention. What you're suggesting are definitely options for reading consistent, if slightly out of date, information from a hot keyspace.
There's a blog post on time travel queries, which conveniently expose the required functionality as specified by the SQL standard.
IIRC, in the case that aphyr refers to for these specific numbers, the reads are scans that span multiple shards[1], while the writes are writes to single shards.
[1] even though aphyr says it's just a hundred rows, the tables are split into multiple shards because aphyr in this case was specifically testing our correctness in multi-shard contention scenarios. In production you wouldn't have multi-shard reads crop up until you were doing scans for tables that were hundreds of thousands of rows in size[2]. It's easier to picture if you think of this performance speed difference in the scenario where you are doing full table scans spanning multiple shards on multiple computers while the underlying rows are being rapidly mutated by contending write transactions. The transactions get constantly aborted to avoid giving a non-serializable result, and performance is suffering. We agree that the numbers in this contention scenario are too low, and we are actively working on high-contention performance (and performance in general) leading up to our 1.0 release[3].
[2] Specifically, we break up into multiple shards when a single shard exceeds 64mb in size.
[3] You can follow along on one of the PRs that address this specific performance issue here: https://github.com/cockroachdb/cockroach/pull/13501