SLslimsag·May 1, 2015
The awesome news is that all of the software I'm going to mention all uses almost identical terminology, which originates from Google's Dapper[1].
- Dapper itself is a closed-source and internal Google project, so you and I can't make use of it.
- There is something called AppNeta (previously named Tracelytics), a paid service, I haven't ever used it but it was on HN a while back. [2]
- Most notably is Twitter's Zipkin[3], which is what the rest of my comment will make comparison of (but take note I haven't used Zipkin extensively, I've only researched it).
Zipkin is written in Scala, and Appdash in Go. Depending on which you're familiar with might make your decision between the two easier. A language you know and love goes a long way for hacking on a project.
On a user-interface side; they're both pretty similar. I think Zipkin is gettin
Read the rest
g an aggregated data view sometime soon, but Appdash already has an extensive one with several different viewing modes (see screenshots in [4]).
I don't know Scala too well and couldn't determine during my research how extendable Zipkin is -- but Appdash was designed from the ground up with interfaces providing core logic: this means you can swap out where the traces get stored (in memory, evicted after a certain time, in a database, your own custom implementation, etc), define your own event types with their own meta-data associated with them (an HTTP event has headers, an SQL event has the literal query that was run, your own type could have anything you want, etc).
Appdash has a Python support library so you can generate traces from Python applications, too. This works because it uses Protobuf for communication. We hope support for more languages might be contributed. IIRC Zipkin uses Apache Thrift for communication, but doesn't officially support integration with other languages via libraries.
A really useful feature of Appdash that I think Zipkin doesn't have is the ability to import and export single or multiple traces as JSON from the UI. You can literally copy+paste traces to co-workers, etc, and they can load them up in their own Appdash instance to analyze them if they want.
Perhaps missed among all my discussion of HTTP-esque-tracing-systems is that Appdash is really suitable for analyzing the performance of any application at a high-level (i.e. not micro-benchmarks). If you can time something, want to log something, or have something important with a timestamp on it -- you can just give it to Appdash and it'll display it in a nice UI for you.
[1] http://research.google.com/pubs/pub36356.html
[2] https://news.ycombinator.com/item?id=3197490
[3] http://twitter.github.io/zipkin/
[4] https://github.com/sourcegraph/appdash/pull/58