Web Performance & Scalability Tools
Download as PPT, PDF1 like876 views
This document discusses strategies for scaling web applications. It identifies key areas like the database, application, web server, scripting language, hardware, and hosting that impact scalability. Specific solutions proposed include adding servers horizontally, caching outputs, optimizing databases through indexing and load balancing, profiling applications to identify bottlenecks, tuning web and scripting language servers, using solid state drives, and leveraging the cloud for easy scaling. Resources are provided for further reading on high performance PHP and LAMP application scalability.
Web Performance & Scalability Tools
- 2. Scalability It is the ability of a system to handle a growing amount of traffic/data in a capable manner How much system can grow How easy it is to increase capacity
- 4. Solutions Add hardware. More CPU and RAM to machine Vertical Scalability Costly Higher TCO Add another server Horizontal Scalability More reliable Less costly as compared to other options Maintenance cost is higher
- 5. Opportunity Areas Database Application Web Server Scripting Language Hardware Hosting
- 6. Database Query Optimization Log slow queries Check query plan to make sure indexes are used Indexes A lot of how-to's available MySQL Query Cache Performance Schema
- 7. Database Review default values Index buffers Innodb Buffer Pool Number of open tables Innodb vs MyISAM Speed up file system Per-table file for Innodb Keep RDBMS updated A lot of diagnostic tools available
- 9. Database Load balancing Route read queries to slaves Write queries to Master Multi-master Architecture Offload Database server Sphinx, Solr Nosql (Radis, Mongo etc) MySQL Cluster Write intensive
- 10. Application Identify Bottleneck Newrelic xDebug Framework profiling tools
- 13. Application Use framework DB APIs Avoid custom queries It allows the use of plug-ins / extensions / modules Use Framework cache APIs Don’t by-pass framework or CMS Profile your code Time Memory Profile/Monitor third party services
- 14. Application Cache output Full page cache Can be done in: Application (Disk, memcache, apc etc) Proxy server (varnish, lightspeed) Very effective Wikipedia serves 85% cached content. Only 15% is fetched from DB Cache static files in memory Query Results Cache Reduces number of queries Reduce IO Avoid file_exists sort of checks Reduce file operations
- 15. Web Server Explore options Apache Nginx Litespeed Load balancing HAProxy Ldirector Varnish (define multiple backend) Nginx, Apache balancer modules Decide balancing strategy Round robin, Last used, Weighted Do heavy lifting in crons, message queues
- 17. Web Server Fine-tune web server Don’t rely on default configurations Make web server use multiple cores E.g worker_processes for nginx Check your specific server configuration KeepAlive sort of options A lot can be done
- 18. Scripting Language (php) Enable opcache APC Opcache in 5.5 eAccelerator Fine-tune settings Real path cache Buffers Max memory per script Look for best practices for that language
- 19. Hardware Disk is the slowest part of the system Go for SSD Reduce IO Offload Extensive downloads Cloud files CDN File system optimizations Number of files in folder RAID
- 20. Hosting Cloud vs. Dedicated Cloud makes scaling easy Easy to add hardware Easy to build Cloud is cheaper as compared to dedicated hardware Easy to separate DB & Application server Avoid having control panels on servers Cloud has slight performance disadvantage Does not work well when it is raining.
- 21. Resources http://www.slideshare.net/mkherlakian/high-performance-php-scaling-and- getting-the-most-out-of-your-infrastructure http://www.slideshare.net/mohamed_almasry_lamp/lamp-applications- scalability-lamp-web-applications http://20bits.com/article/10-tips-for-optimizing-mysql-queries-that-dont- suck