在组内做了 zipkin 的分享,下面是分享摘要
Zipkin
Menghan
@menghan412
Outline
-
Google’s Dapper
-
Twitter’s Zipkin
-
Douban’s ???
Dapper
Why Google create launch Dapper?
What is Dapper like?
black-box V.S. annotation-based
Dapper
-
Low overhead
-
Application-level transparency
-
Scalability
-
Analyse quickly
Dapper design
-
data structure: trace, span, annotation, key-value annotation
-
instrumentation points
-
sample
-
collection: out of band
Dapper deploy
-
Dapper daemon
-
overhead calculation
-
adaptive sampling
-
adaptive collection
-
D-API
Dapper learned
-
A sample of just one out of thousands of requests provides sufficient information for many common uses of tracing data
-
integration with exception monitoring
-
coalescing effects
-
offline workload
-
find root cause
-
record kernel level infomation
Zipkin
http://www.infoq.com/presentations/Zipkin
Zipkin V.S. Dapper
-
scribe vs daemon
-
initiative vs passive
-
cassandra vs BigTable
Zipkin
usage
Douban
???