Common Package Specification (CPS) in practice: A full round trip implementation in Conan C++ package manager CppCon24Outline - Introduction to Common Package Specification (CPS) - Creation of CPS

Experimental dataset NYC Taxi and Uber Trips › Where to download: https://www1.nyc.gov/site/tlc/about/tlc-trip-record-data.page › How to import data into ClickHouse: https://clickhouse.yandex/docs/en/gettin data WHERE rand() % N < M SELECT trip_id FROM trips WHERE (rand() % 1000) = 0 LIMIT 1 SETTINGS max_threads = 1 ┌───trip_id─┐ │ 960186089 │ └───────────┘ SELECT trip_id FROM trips WHERE (rand() % 1000) SETTINGS max_threads = 1 ┌───trip_id─┐ │ 335608036 │ └───────────┘ 9 / 62 How to sample data WHERE hash(...) % N < M SELECT trip_id FROM trips WHERE (sipHash64(trip_id) % 1000) = 0 LIMIT 1 SETTINGS

Experimental dataset NYC Taxi and Uber Trips › Where to download: https://www1.nyc.gov/site/tlc/about/tlc-trip-record-data.page › How to import data into ClickHouse: https://clickhouse.yandex/docs/en/gettin data WHERE rand() % N < M SELECT trip_id FROM trips WHERE (rand() % 1000) = 0 LIMIT 1 SETTINGS max_threads = 1 ┌───trip_id─┐ │ 960186089 │ └───────────┘ SELECT trip_id FROM trips WHERE (rand() % 1000) SETTINGS max_threads = 1 ┌───trip_id─┐ │ 335608036 │ └───────────┘ 9 / 62 How to sample data WHERE hash(...) % N < M SELECT trip_id FROM trips WHERE (sipHash64(trip_id) % 1000) = 0 LIMIT 1 SETTINGS

⼀一天你会笑着说出来 --晏凌⽺羊 Post-Trip On-Trip 2014年年架构 Conn ection Node
 (CN) Dispatch Req Trip History & User data Dispatch
 (Monolith) Record trip complete & fares API
 (Monolith) Postgres
 流量量⼀一年年涨5倍 • 不不变的话, ⼏几个⽉月后会被流量量压垮 2016年年架构 Demand Supply Optic Disco Geobase RTAPI Eyeball requests Trip requests • RTAPI: Uber’s edge gateway • Optic: Rider state machine • Demand: Job state machine

approach you have to wait for a response for each request you make so are limited by the network round trip time, or latency of your network. With an asynchronous system you can pipeline flows of messages in response is received. Blocking on each messages sent is costly since it requires a network round trip for each message sent. By not blocking and receiving send acknowledgements asynchronously you can would then get sent to the client side, if one was available. This would involve a network round trip for every message and reduce performance. Therefore, by default, Apache ActiveMQ Artemis pre-fetches

approach you have to wait for a response for each request you make so are limited by the network round trip time, or latency of your network. With an asynchronous system you can pipeline flows of messages in response is received. Blocking on each messages sent is costly since it requires a network round trip for each message sent. By not blocking and receiving send acknowledgements asynchronously you can would then get sent to the client side, if one was available. This would involve a network round trip for every message and reduce performance. Therefore, by default, Apache ActiveMQ Artemis pre-fetches

approach you have to wait for a response for each request you make so are limited by the network round trip time, or latency of your network. With an asynchronous system you can pipeline flows of messages in response is received. Blocking on each messages sent is costly since it requires a network round trip for each message sent. By not blocking and receiving send acknowledgements asynchronously you can would then get sent to the client side, if one was available. This would involve a network round trip for every message and reduce performance. Therefore, by default, Apache ActiveMQ Artemis pre-fetches

approach you have to wait for a response for each request you make so are limited by the network round trip time, or latency of your network. With an asynchronous system you can pipeline flows of messages response is received. Blocking on each messages sent is costly since it requires a network round trip for each message sent. By not blocking and receiving send acknowledgements asynchronously you can would then get sent to the client side, if one was available. This would involve a network round trip for every message and reduce performance. Therefore, by default, Apache ActiveMQ Artemis pre-fetches

approach you have to wait for a response for each request you make so are limited by the network round trip time, or latency of your network. With an asynchronous system you can pipeline flows of messages response is received. Blocking on each messages sent is costly since it requires a network round trip for each message sent. By not blocking and receiving send acknowledgements asynchronously you can sent to the client side, if one was available. Examples 86 This would involve a network round trip for every message and reduce performance. Therefore, by default, Apache ActiveMQ Artemis pre-fetches

approach you have to wait for a response for each request you make so are limited by the network round trip time, or latency of your network. With an asynchronous system you can pipeline flows of messages response is received. Blocking on each messages sent is costly since it requires a network round trip for each message sent. By not blocking and receiving send acknowledgements asynchronously you can sent to the client side, if one was available. Examples 65 This would involve a network round trip for every message and reduce performance. Therefore, by default, Apache ActiveMQ Artemis pre-fetches