BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Why Scheduling - Minor Gordon\, Computer Laboratory\, University o
 f Cambridge
DTSTART:20070123T143000Z
DTEND:20070123T153000Z
UID:TALK6180@talks.cam.ac.uk
CONTACT:Minor Gordon
DESCRIPTION:Research designs for high performance web servers have long be
 en defined by the strategy they employ to handle many thousands of request
 s concurrently. A number of efficient designs have emerged in the last dec
 ade\, with the most prominent of them (Flash and SEDA) occupying the middl
 e ground between the extremes of purely thread-based (Apache) and purely e
 vent-based (Zeus) concurrency.\n\nWhat is not well understood is how the v
 arious concurrency strategies scale beyond uniprocessors. Multicore and mu
 ltiprocessor environments induce new sources of latency such as remote cac
 he misses\, with slower clock speeds making disk reads even more expensive
 . Fortunately\, with intelligent disk scheduling and a large RAM a web ser
 ver can significantly reduce the impact of disk I/O on server performance 
 under a typical static file workload. However\, once the server is working
  primarily from memory the main source of latency becomes the memory hiera
 rchy\, particularly L2 data cache misses.\n\nI am currently investigating 
 the effects of different concurrency strategies on server software efficie
 ncy. In this talk I'll present the application server I've been working on
  for the past year\, explain why I think SPECweb2005 is all but useless\, 
 and sketch my plans for a thesis evaluation.\n
LOCATION:Room FW11\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR