Download to read offline
Uploaded byDaniel Khan
Hunting Performance Problems in Node.js and beyond
The document discusses performance issues in Node.js, highlighting the need for proactive defense against memory and CPU problems. It suggests using tools like n|solid for monitoring Node.js metrics and APM solutions such as Dynatrace for a comprehensive view of application performance. Key takeaways emphasize the importance of monitoring all transactions across the technology stack.
Hunting Performance Problems in Node.js and beyond
- 1. Hunting Performance Problemsin Node.js Daniel Khan | @dkhan
- 3. Node.js is what theCompany Outlaws use to finally introduce change
- 7. Node.js is what theCompany Hipsters threw in just to break everything
- 9.
- 10. Node.js is aC++ program controlled by V8 JavaScript
- 14.
- 16. process.memoryUsage() { rss: 4935680,heapTotal: 1826816, heapUsed: 650472 }
- 27. How to build aMemory Leak
- 29.
- 31. http://bit.ly/1PvijIy var snap =profiler.takeSnapshot(); snap.serialize(); v8-profiler
- 32.
- 33.
- 34.
- 35.
- 39.
- 41. get(‘/routeA’, function(){}) get(‘/routeB’, function(){}) get(‘/routeC’,function(){}) get(‘/routeD’, function(){}) get(‘/routeE’, function(){}) get(‘/routeA’, function(){}) get(‘/routeX’, function(){}) get(‘/^(route|router)/(.+)’, function(){}) HTTP GET /routeX get(‘/routeA’, function(){}) get(‘/routeX’, function(){}) O(n)
- 42.
- 53. 1980 20001995 20052010 2015 Stakeholders
- 57. “No man isan island” or … We need a holistic view.
- 58. Meet Application PerformanceMonitoring
- 69. Takeaways • Node.js introduceschange • Use dedicated tools like N|Solid to monitor Node.js specific metrics like CPU and Memory Usage • Protect your boundaries by actively monitoring incoming and outgoing transactions • Use APM Solutions like Dynatrace or Ruxit to get a holistic view for all transactions passing through all tiers of your stack
- 71.
Editor's Notes
- #11 But I really like this definition.
- #12 What V8 does: It translates JavaScript into machine code and runs it.
- #14 And every running program is represented by Work done on the CPU Data Stoed im memory And here we already have our two problem classes.
- #16 Let’s review the memory handling of Node.js It’s very common and quite like Java. RSS: Code and Stack where local variables are stored It contains the heap for where objects and closures – let’s say long living resources like objects or closures are stored
- #17 And it’s actually very easy to query the memory usage of a given Node process.
- #18 And this is what you get if you create a graph of a running Node application. As we already see the heapUsed graph looks funny. It somehow is very dynamic and seems to follow a pattern. Something seems to take care that memory is freed. And I have a real live example of how this works.
- #26 The console spits this out
- #29 And now that we have found our problem we can add this. And the weekend is saved.
- #33 My script will look at the memory usage and it constantly grows it will create a snapshot.
- #37 Again v8 profiler. Again a function already there in V8. This will give you this.
- #38 We can also look at some color coded chart that shows different function calls and the length. This pretty much resembles the flame charts we saw before. Some people like it – I don’t.
- #39 What I did is: I fed the json tree to d3 sunburst.
- #42 So this netflix latency problem.
- #73 Speaking about Node.js let’s outline what this talk is about. When ever there’s bad press about Node.js it’s about performance problems. Here are two examples.