Lükkamis- ja tõmbamispõhine andmete edastus AJAX rakendustes.

Date

2011

Journal Title

Journal ISSN

Volume Title

Publisher

Tartu Ülikool

Abstract

Veebipõhistes rakenduses on populaarsust kogunud AJAX ja COMET tehnoloogiad, mis võimaldavad luua interaktiivseid veebirakendusi. COMET funktsionaalsuse saavutamiseks on mitu meetodit, nimelt pikk päring ning lükkamine. Käesoleva bakalaureusetöö eesmärgiks on võrrelda neid erinevaid tehnoloogiaid raamistikus, mille lõid Ali Mesbah et al., kes võrdlesid AJAX-it ning pikka pärimismeetodit oma artikklis „Performance Testing of Data Delivery Techniques for AJAX Applications“. Bakalaureusetöös loodi uus lükkamispõhine test ning võrreldi selle jõudlust teiste raamistikus olevate testidega. Tulemused näitasid, et lükkamispõhine rakendus kasutas väga vähe TCP pakette ning omas väikest latentsust, keskmiselt 1305ms. Samas oli CPU kasutus kõrge ning mõned teated ei jõudnud kasutajani. Tänu nendele tulemustele, arvestades ka teisi lükkamispõhise meetodi puudujääke, leidis autor, et lükkamispõhise meetodi kasutamine pika päringu asemel ei ole kirjutamise hetkel õigustatud.
The Internet contains an ever increasing number of interactive applications, powered by technologies like AJAX and COMET. The latter has multiple ways of achieving the desired functionality, namely long-polling and streaming. While there have been many comparisons between AJAX and the traditional model of the web and even AJAX and long-polling, little research has been done comparing streaming to long-polling. This thesis aims to compare the different technologies in a framework created by Ali Mesbah et al., who compared polling and long-polling solutions in their 2008 article „Performance Testing of Data Delivery Techniques for AJAX Applications“. A new streaming test was created and compared with the other tests included in the framework. The results showed that streaming used very little network resources, using only one packet per message after the initial connections were made. Latency was also not an issue as messages were received by the client in 1305ms on average. However CPU usage and data coherence were not as good. Streaming dropped 1-2 messages where long-polling dropped none and the CPU usage was high at unpredictable times. These results, along with other shortcomings of the streaming method, led the author to determine that streaming is not a viable alternative to long-polling or polling at this time.

Description

Keywords

Citation