The Perfect Payment Experience

Infrastructure and how Square makes commerce easy.

Transcript

1. The Pe ect Payment Experience Infrastructure and how Square makes commerce easy
2. Abhay Kumar Staff Engineer · Square, Inc @abhay · abhay@squareup.com
3. ‣ Three countries: United States, Canada, Japan ‣ Millions of individuals and businesses ‣ Processing $15 billion annually, excluding Starbucks ‣ Card and Wallet transactions at 7,000 Starbucks stores Square Stats
4. ‣ 180+ million Americans carry a payment card. ‣ 600+ million cards in the United States. ‣ 75% of businesses only accept cash. Industry Stats
5. Payments are easy.
6. A Pe ect Payment Experience is hard.
7. ‣ Signup ‣ Activation ‣ Receiving a reader ‣ First swipe ‣ Subsequent swipes ‣ Settlement / Receiving funds ‣ Refunds ‣ Chargebacks Where do we sta ?
8. First define a metric.
9. ‣ Swipe card. ‣ Authorize transaction. ‣ Capture signature. ‣ Repeat a few times. ‣ Get paid. Swipe to Settle
10. Record all the things.
11. ‣ Total request duration including background calls. ‣ Time transitions between screens. ‣ How long did the server take? ‣ Did the merchant swipe once? twice? ‣ Did the authorization fail? Did the merchant retry? ‣ Did the client crash? ‣ Link all the interactions to make it easier. ‣ Use an identifier that spans the ‘session’. ‣ Forward everything to the server in the background. Client-side metrics
12. ‣ End-to-end ‣ User lookup ‣ Decryption / Tokenization ‣ Datastore ‣ Gateway / Card network ‣ Risk decision engine Server-side metrics Online processing
13. ‣ Time to cardholder statement ‣ Time to merchant payout ‣ Successful deposit into merchant accounts Server-side metrics Batch processing
14. Collect all the things.
15. ‣ Provenance is impo ant ‣ Make it easy to access this information ‣ Anonymize/sanitize as necessary ‣ Make it available in all data-centers ‣ Store a ‘raw’ copy of everything Invest in data collection
16. I think we’re gonna need a bigger boat.
17. ‣ Samsa: set of applications used for high-throughput messaging. ‣ Samsa producer clients: metrics, logging, rpc-ale ing, sawmill ‣ Kafka: message broker. messages are available for a configurable retention period ‣ Samsa consumer applications: graphite, splunk, distributed tracing, HDFS Tools we use Samsa + Kafka
18. ‣ Storing everything has benefits ‣ MapReduce to provide intermediate a ifacts ‣ Generate your metrics in batch or near-realtime Tools that we use Hadoop
19. ‣ We use it a lot. It’s not going to scale. ‣ Intermediate products and some source data ends up here. ‣ Sta sharding data. ‣ Consider alternatives for your analysts. Tools that we use MySQL
20. Be available.
21. ‣ Online card processing should be in multiple data centers. ‣ Pin transactions to your data centers. ‣ Degrade gracefully Availability
22. ‣ Invest in multiple processing networks. ‣ Invest in redundant connectivity to them. ‣ Get closer to the metal. Networks
23. Questions?