Networking and APIsIdentity headers
A caller sends X-User-Id: admin. Does your gateway believe it?
0%
Page 30
More questions.
Data and storageTwo philosophies
A cheque drop box never makes you queue. How does the bank stay correct?
0%
Running in productionDependencies
Why can adding pods during a slowdown knock over your database?
0%
Why is it slow? Performance anti-patternsServerless
Your Lambda takes 120 ms, its API 30. Where did the rest go?
0%
Key technologiesData model
JSON, maps and vectors. Do you really need three more databases?
0%
Core conceptsThe budget
99.9% sounds near perfect. How many minutes down does it allow?
0%
Networking and APIsThe classic bug
Why can a fresh key on every retry double-charge your customers?
0%
Data and storageLost update
Two payments hit a ₹500 wallet at once. Why does it show ₹300?
0%
Running in productionCanary analysis
Why does a 1% canary on a quiet service tell you nothing?
0%
Why is it slow? Performance anti-patternsShuffle sharding
Why does giving each tenant a random pair of workers shrink the blast radius?
0%
Key technologiesTime buckets
Why does a busy chat channel need a new partition every hour?
0%
Core conceptsCloud SLAs
EC2 goes dark for 43 minutes on sale day. What does AWS owe you?
0%
Networking and APIsWindow edges
Why does a 5-per-second window let 10 through in half a second?
0%
Data and storageOrder of magnitude
Why did Figma’s first sharded Postgres table take about nine months?
0%
Running in productionBackfills
One UPDATE fills a new column on 200 million rows. What could go wrong?
0%
Why is it slow? Performance anti-patternsGoodput
Why can a 200 OK in your logs mean nobody got the answer?
0%
Key technologiesGuarantees
Elasticsearch returned 200 OK. Why were some dishes missing?
0%
Core conceptsHeadroom
Suppose checkout needs 8 servers. Why do three zones beat two?
0%
Networking and APIsOutage
Why did Facebook vanish while its DNS servers kept running?
0%
Data and storageWrite amplification
Change 100 bytes in InnoDB. Why do 32 KB hit the disk?
0%
Running in productionArchitecture
How did Prime Video cut a monitoring service’s cost by over 90%?
0%
Key technologiesLIST
Why can finding one user’s photos in S3 take 83 minutes?
0%
Core conceptsGitHub, 2018
A 43-second network blip. Why did GitHub run degraded for a day?
0%
Networking and APIsNAT ports
CPU idle, partner healthy, yet calls time out at peak. What ran out?
0%
Data and storageFull-text search
Why does LIKE '%kurta%' read 10 GB to answer one search?
0%
Running in productionAnti-corruption layer
Legacy stores a waitlist as “WL/017”. Should your new code ever see it?
0%