Ch. x web performance
0 likes1,489 views
The document discusses web performance and identifies several factors that can cause delays, including issues with client hardware/software, server hardware/software, and network connectivity. It notes that performance issues cost companies money by hurting the user experience, conversion rates, and other metrics. The document then outlines several laws of web performance and provides tips for optimizing performance on both the server-side and client-side of web applications and websites.
![Load Testing
ab - Apache HTTP server benchmarking tool
$ ab -n 1000 -c 5 http://www.inf.uth.gr
[...]
Requests per second: 0.59 [#/sec] (mean)
Time per request: 1688.597 [ms] (mean)
[...]
... or Apache JMeter
... or httperf (HP labs)
... or OpenSTA (heavyweight tool)
... or even cloud services like Blitz.io (paid)
Τετάρτη, 11 Ιανουαρίου 2012](https://image.slidesharecdn.com/ch-141204163553-conversion-gate02/85/Ch-x-web-performance-21-320.jpg)
Ad
More Related Content
What's hot (20)
Viewers also liked (11)
Ad
Similar to Ch. x web performance (20)
Ad
Recently uploaded (20)
Ch. x web performance
- 1. Performance of Web Technologies Thanos Theodoridis @attheodo - http://attheo.do Τετάρτη, 11 Ιανουαρίου 2012
- 2. What is web performance? Τετάρτη, 11 Ιανουαρίου 2012
- 3. What is web performance? Τετάρτη, 11 Ιανουαρίου 2012
- 4. “The delay perceived by the user between an action and a meaningful response” What is web performance? Τετάρτη, 11 Ιανουαρίου 2012
- 5. What causes the delay? Web technologies are mostly client-server AND/OR server-server architectures over the network Client • bad hardware • front-end (app/website/browser) sucks Server • bad hardware • back-end (operating system, web server, database software, your code) sucks Network • bad hardware (routers, switches, WiFi antennas) • limited availability & bandwidth (ever heard of 3G?) • dodgy ISPs & datacenters Τετάρτη, 11 Ιανουαρίου 2012
- 6. Why delay matters? Τετάρτη, 11 Ιανουαρίου 2012
- 7. Why delay matters? Τετάρτη, 11 Ιανουαρίου 2012
- 8. Why delay matters? It costs money to EVERYONE! Τετάρτη, 11 Ιανουαρίου 2012
- 9. Why delay matters? Τετάρτη, 11 Ιανουαρίου 2012
- 10. Why delay matters? Amazon Offline: $1M/hour Τετάρτη, 11 Ιανουαρίου 2012
- 11. Why delay matters? Amazon Offline: $1M/hour eBay Offline: $90K/hour Τετάρτη, 11 Ιανουαρίου 2012
- 12. How performance becomes a money issue • Bad UX (you are losing fans) • Bounce Rate sky-rockets (you are losing visitors) • Page Views go down (you are losing Sales) • Conversion Rate goes down (you are losing revenue) • Infrastructure expenses go high (you bleed money) Τετάρτη, 11 Ιανουαρίου 2012
- 13. 4 Laws of web performance Τετάρτη, 11 Ιανουαρίου 2012
- 14. The Law of Stickiness If your website performs, users stick with it. If your website disappoints users, they move to another and stick with that. Τετάρτη, 11 Ιανουαρίου 2012
- 15. The Law of User Perspective Measure performance from the end-userʼs point of view Τετάρτη, 11 Ιανουαρίου 2012
- 16. The Law of Responsibility Whatever happens, itʼs your fault. Even if hell breaks loose. Τετάρτη, 11 Ιανουαρίου 2012
- 17. The Law of Expectations Users expect your website to perform at least as well or better that a similar website theyʼve used Τετάρτη, 11 Ιανουαρίου 2012
- 18. The law of the laws “Premature optimization is the root of all evil.” Τετάρτη, 11 Ιανουαρίου 2012
- 19. Server-side performance Τετάρτη, 11 Ιανουαρίου 2012
- 20. Server-side performance • How good you stack components interoperate together • HTTP server • database • frameworks • operating system • How efficiently your code runs on your stack • query tuning / indexes (for the database) • runtime optimizations (compiler flags, configuration/tuning) • resource utilization (multicore CPU) requests / second avg response time / request Most important numbers Τετάρτη, 11 Ιανουαρίου 2012
- 21. Load Testing ab - Apache HTTP server benchmarking tool $ ab -n 1000 -c 5 http://www.inf.uth.gr [...] Requests per second: 0.59 [#/sec] (mean) Time per request: 1688.597 [ms] (mean) [...] ... or Apache JMeter ... or httperf (HP labs) ... or OpenSTA (heavyweight tool) ... or even cloud services like Blitz.io (paid) Τετάρτη, 11 Ιανουαρίου 2012
- 22. Application Level (code profiling) - PHP - XDebug, PECL APD Extension - Python - cProfile, timeit - Java Servlets - JProfiler - ....Good old gettime(); Database Level (optimize queries) - MySQL mysql> SET profiling=1; SELECT * FROM `table_name`; mysql> SHOW PROFILES; - MongoDB mongostat Operating System Level (resource monitoring) ./top (general) ./vmstat (Virtual Memory Statistics, paging, block I/O) ./iostat (disk I/O) Networking Level - tcpdump - netstat Τετάρτη, 11 Ιανουαρίου 2012
- 23. Client-side performance Τετάρτη, 11 Ιανουαρίου 2012
- 24. Client-side performance • How good is your front-end • Browser (you canʼt really do anything about it) • HTML syntax (standards compliant?) • CSS (standards compliant?) • Javascript code • How well all the above play together Most important numbers Initial Render Time Page (completely) Load time ~ 3-4 seconds < 700-1000ms < 3s Τετάρτη, 11 Ιανουαρίου 2012
- 25. Client-side performance • 0.1 seconds gives the feeling of instantaneous respone • 1 second keeps the userʼs flow of thought constant • 10 seconds keeps the userʼs attention • Most of the page load time is spent outside your hosting datacenter • Optimize your front-end • Optimize the interaction bertween the end-userʼs browser and your server • Fine tune the way the browser processes client-side elements HTML CSS Javascript Images Browser HTTP TCP/IP Τετάρτη, 11 Ιανουαρίου 2012
- 26. How a website loads Τετάρτη, 11 Ιανουαρίου 2012
- 27. Page Load Time Page Load Time = (RTT x Turns) + Server Processing Time + Client Processing Time + Page Size/Bandwidth • Page Size • HTML filesize + Stylesheets size + Javascripts size + images size + 3rd party banners + ... • RTT • latency between sending of a request to the Web server and the receipt of the first bytes • Turns • The number of TCP connections required to download a component • Base HTML contains instructions for locating and fetching additional website objects like images or scripts • Depends whether HTTP1.0/1.1 to determine how many per component • Processing time (Client + Server) • Non-deterministic, varies dramatically from application to application Τετάρτη, 11 Ιανουαρίου 2012
- 28. Web Performance Best Practices Every kid on the block has a set of rules and a tool •Google •PageSpeed Extension •http://code.google.com/speed/page-speed/docs/rules_intro.html •Yahoo •YSlow Extension •http://developer.yahoo.com/yslow/help/#guidelines 6 Rules to rule them all •Reduce page size to < 500kb •Enable gzip compression •Minify and merge CSS and javascript files •Reduce RTTs (<40 per page) •Structure properly: CSS first, javascript last (not always possible) •CACHE, CACHE, CACHE, OMG OMG CACHE!!! If you apply these properly, page download times will drop by 40-50% Τετάρτη, 11 Ιανουαρίου 2012
- 29. DEMO YSlow in action Τετάρτη, 11 Ιανουαρίου 2012
- 30. Questions? Τετάρτη, 11 Ιανουαρίου 2012