Performance optimisation of webmail Katerina Performance optimisation of webmail Mparm- popoulou, Periklis Stefopoulos Katerina Mparmpopoulou Periklis Stefopoulos Introduction Research question Supervised by: Michiel Leenaars Approach Experimental Results University of Amsterdam User Experience System & Network Engineering System Performance Conclusions July 3, 2013 Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 1 / 21
Performance optimisation Outline of webmail Katerina Mparm- popoulou, Periklis Stefopoulos 1 Introduction Introduction Research 2 Research question question Approach 3 Approach Experimental Results User Experience System Performance 4 Experimental Results Conclusions User Experience System Performance 5 Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 2 / 21
Performance optimisation Webmail System Components of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 3 / 21
Performance optimisation Research purpose of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction • Hundreds of millions of end users depend on Webmail Research question technologies Approach • Open source web frontends to mail servers are an often Experimental neglected area of development Results User Experience • Better understanding the performance of web mail System Performance applications is a prerequisite to better tuning these Conclusions applications Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 4 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research Question Research question What are the bottlenecks, in terms of performance, of Approach current Webmail implementations and which could be the Experimental most optimal solution? Results User Experience System Performance Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 5 / 21
Performance optimisation Experimental Environment of webmail Katerina Mparm- popoulou, Periklis Webmail Frontends Stefopoulos • Squirrelmail Introduction Research • Roundcube question • Horde IMP Approach Experimental • Afterlogic Webmail Lite Results User Experience System Performance Webmail Backends Conclusions • Courier - Postfix - Amavis - ClamAv - SpamAssassin • Dovecot - Postfix - Amavis - ClamAv - SpamAssassin • Cyrus - Postfix - Amavis - ClamAv - SpamAssassin Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 6 / 21
Performance optimisation Experiments of webmail Katerina Mparm- popoulou, 5 users with different mailbox size Periklis Stefopoulos • 1500 messages with only text Introduction • 1500 messages with attachments plus text Research question • 3000 messages with only text Approach • 4500 messages with only text Experimental Results • 6000 messages with only text User Experience System Performance Conclusions 3 different user actions • Log in to Webmail • Searching a keyword from the ”Subject” field • Searching a keyword from the entire message content Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 7 / 21
Performance optimisation Metrics of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Metric Extraction method Research question Latency tcpdump/Wireshark Approach CPU time 1 systat/sar Experimental Unique Set Size (USS) smem Results User Experience Proportional Set Size (PSS) smem System Performance Conclusions Table: Benchmark metrics and their extraction method 1 CPU time = CPU utilization ∗ elapsed time ∗ number of CPUs Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 8 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Figure: average fetching time during login Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 9 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 10 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 11 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Figure: Comparison of the most effective Solutions regarding searching from ”Subject” Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 12 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 13 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 14 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Figure: Comparison of the most effective Solutions regarding searching from Entire Message content Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 15 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 16 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Figure: users search a keyword from the Entire Message Content: CPU time consumption for Dovecot and Courier IMAP backends Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 17 / 21
Performance optimisation of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Figure: average memory consumption Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 18 / 21
Performance optimisation of webmail Katerina Mparm- Conclusions popoulou, Periklis Stefopoulos • Afterlogic achieves shortest latencies in searching Introduction • Horde IMP has the longest response times Research • Dovecot and Cyrus carry out the search from ”Subject” question request efficiently and with relatively the same latency Approach Experimental • Cyrus is the IMAP backend that performs the best during Results User Experience search from the entire message System Performance • Cyrus has the lowest CPU utilization and the highest Conclusions average memory consumption for all IMAP functions, followed by Dovecot Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 19 / 21
Performance optimisation Answering the Research Question of webmail Katerina Mparm- popoulou, Periklis Stefopoulos What are the bottlenecks, in terms of performance, of current Introduction Webmail implementations? Research question • The major bottleneck in an integrated webmail system is Approach the IMAP backend Experimental Results User Experience System Performance which could be the most optimal solution? Conclusions • the solution of using Cyrus IMAP combined with Afterlogic Webmail Lite performs better in terms of both user experience and system overall performance Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 20 / 21
Performance optimisation Questions of webmail Katerina Mparm- popoulou, Periklis Stefopoulos Introduction Research question Approach Experimental Results User Experience System Performance Conclusions Katerina Mparmpopoulou, Periklis Stefopoulos Performance optimisation of webmail July 3, 2013 21 / 21
Recommend
More recommend
