cache document timeout

We now investigate the effect of the cache timeout on the size of the cache needed and on the bandwidth savings. Figure 7 shows as a function of the cache document timeout for zone 0 the byte volume requested from the central server and the maximum size of the cache (in MB) needed for any one minute of the two-day period. The figure shows the tradeoff that the timeout poses between the need for documents from the central server and the local cache disk space requirements. Selecting an appropriate timeout involves balancing cache management costs with the cost of using the central server. This latter cost may have several components, including delay, the cost of accessing servers that charge for usage, or the cost of using network service providers who charge for transmission of requested documents.

Figure 8 shows for all zones the maximum size of the cache (in MB) needed for any one minute of the two-day period. For this specific data set a cache of at most a few hundred megabytes would suffice for any of the zone caches, even with a timeout of several hours. In fact, increasing the timeout beyond an hour has little effect on the required cache size. In this example, the zone with the largest cache size requirement would need about 125MB of caching space. As figure 9 illustrates, these 125MB would have offered almost 4GB of savings in transmissions from the central server. Even small caches and small timeouts yield a significant benefit; the marginal benefit of larger timeouts is quite diminished.