Geography Reference
In-Depth Information
1600000
1400000
1200000
1000000
sin metatiling
metatile centrado
metatile min. corr.
without metatiling
centered metatile
arbitrary metatile
800000
600000
400000
200000
0
B=0
B=1
B=2
B=3
Buffer size
Figure 13.
Number of cached tiles for different buffer sizes and metatile configurations.
WMS server, a broaden population of the cache is achieved. These extra pre-generated map
image tiles stored in the cache will allow a faster delivery of future requests.
5.3. Cache replacement policies
When the tile cache runs out of space, it is necessary to determine which tiles should
be replaced by the new ones. Most important characteristics of Web objects, used to
determine candidate objects to evict in Web cache replacement strategies, are:
recency
(time since the last reference to the object),
frequency
(number of times the object has
been requested),
size
of the Web object and
cost
to fetch the object from its origin
server. These properties classifies replacement strategies as recency-based, frequency-based,
recency/frequency-based, function-based and randomized strategies [14]. Recency-based
strategies exploit the temporal locality of reference observed in Web requests, being usually
extensions of the well-known LRU strategy, which removes the least recently referenced
object. Another popular recency-based method is the Pyramidal Selection Scheme (PSS)
[15]. Frequency-based strategies rely on the fact that popularity of Web objects is related
to their frequency values, and are built around the LFU strategy, which removes the least
frequently requested object. Recency/frequency-based strategies combine both, recency and
frequency information, to take replacement decisions. Function-based strategies employ a
general function of several parameters to make decisions of which object to evict from the
cache. This is the case of GD-Size [16], GDSF [17] and Least-Unified Value (LUV) [18].
Randomized strategies use a non-deterministic approach to randomly select a candidate object
for replacement.
For a further background, a comprehensive survey of web cache replacement strategies is
presented in ([14]). According to that work, algorithms like GD-Size, GDSF, LUV and PSS
