[email protected] © CHARACTERIZATION OF WORKLOAD AND RESOURCE CONSUMPTION FOR AN ONLINE TRAVEL AND BOOKING SITE! AA 2011/2012 [email protected] 16 Maggio 2012 Refs ¨ Characterization of Workload and Resource Consumption for an Online Travel and Booking Site. Nicolás Poggi, David Carrera, Ricard Gavaldà, Jordi Torres, Eduard Ayguadé. 2010 IEEE Intl. Symp. on Workload Characterization (IISWC) IEEE Press, 2010. ¤ http://www.lsi.upc.edu/%7Egavalda/iiscw10.pdf [email protected] © 16 Maggio 2012 Characterization of Workload and Resource Consumption for an Online Travel and Booking Site A mix of products ¨ A mix of technologies ¨ Great variability of access pattern ¨ A challenge for workload characterization [email protected] © 16 Maggio 2012 Characterization of Workload and Resource Consumption for an Online Travel and Booking Site ¨ Real production log from OTA ¤ client workload (1 week, milion of requests) ¤ load on 35+ nodes ¤ Ajax 3-tiers app on LAMP open source tech. ¨ Log includes ¤ HTTP data ¤ Web server load status ¤ CPU time (user/system) application server ¤ DB access time and num of requests ¤ B2B external services service time [email protected] © 16 Maggio 2012 Characterization of Workload and Resource Consumption for an Online Travel and Booking Site ¨ Workload characterization ¤ client behaviour ¤ server resource consumption [email protected] © 16 Maggio 2012 Scenario Application characteristics ¨ Computing Infrastructure ¨ Dataset properties ¨ [email protected] © 16 Maggio 2012 Workload characteristics ¨ No overload conditions ¨ 6.0 pages/session ¨ 6:48 minutes spent on the site ¨ ¨ ¨ ¨ 3.0s response time for dynamic page generation 8MB of RAM memory consumption highest traffic is on Mondays and decreases to the weekend. Average page views as well as the time spent on the site both increase during the week, peaking at the weekend [email protected] © 16 Maggio 2012 A model of traffic volume [email protected] © 16 Maggio 2012 Traffic composition [email protected] © 16 Maggio 2012 Per-session characteristics [email protected] © 16 Maggio 2012 Per-session model [email protected] © 16 Maggio 2012 Responce time analysis [email protected] © 16 Maggio 2012 Load average [email protected] © 16 Maggio 2012 R time vs Load avg [email protected] © 16 Maggio 2012 CPU % vs load avg [email protected] © 16 Maggio 2012 Memory % vs load avg [email protected] © 16 Maggio 2012 DB R time vs load avg [email protected] © 16 Maggio 2012 Ext R. time vs load avg [email protected] © 16 Maggio 2012 Num of click vs load avg [email protected] © 16 Maggio 2012 Resource usage [email protected] © 16 Maggio 2012 Resource usage [email protected] © 16 Maggio 2012 Progetti d’esame [email protected] © 16 Maggio 2012 Modellazione ¨ Realizzazione di un modello simulativo e studio estensivo del comportamento del sistema al variare dei parametri di ingresso ¤ Tipicamente si parte da un sistema reale descritto in un articolo scientifico e dalla relativa caratterizzazione del carico ¨ Esempi di sistemi: ¤ Datacenter ¤ Cloud federati ¤ Web clusters [email protected] © 16 Maggio 2012 Misure e caratterizzazione del carico ¨ Progettazione di un sistema di monitoraggio per applicazioni web ¤ realizzazione di un piccolo prototipo in laboratorio o misura di un server reale ¤ caratterizzazione del carico ¨ Caratterizzazione di tracce reali (e.g. wikipedia) [email protected] © 16 Maggio 2012 Caratterizzazione del traffico ¨ Analysis of internet traffic traces ¤ DITL ¤ Internet traffic [email protected] © 16 Maggio 2012