Impact of Inter-Update Time Distributions on Wireless Data Access

Abstract

Modern mobile networks support wireless data applications. An application running on the wireless handheld device may repeatedly ac- cess a data entry received from the applica- tion server. If the data entry is not sensitive to time, then the customer may access the data stored in the cache of the wireless handheld de- vice instead of querying the application server, and the expensive wireless transmission over- head is reduced. If the data entry is sensitive to time, then the current data entry should be provided from the application server. Some time-sensitive wireless applications can toler- ate certain degree of inaccuracy. For this type of applications, we can set an expiration pe- riod t to predict when the data entry will be updated. During period t, the data entry in the handheld device is used. When t expires, the next data access results in a query to the mobile network. In this case, the application is weakly consistent where the wireless hand- held device may occasionally access stale data. A mechanism is required to predict when a data entry expires. In Apache and Squid, a time-to-live (TTL) interval t is dened for the data entry stored in the wireless handheld de- vice. We propose an analytic model to provide lower bound performance for the TTL predic- tion mechanism. Based on our model, we show how the mean, the variance, and the skewness of the inter-update time distribution aect the accuracy of TTL interval prediction.