The coordinator runs on a separate PC with two 1
GHz Pentium III and 512 MB RAM. All nodes are interconnected by a switched
100 MBit Ethernet.
OLAP with a Database Cluster 24
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission
of Idea Group Inc. is prohibited.
Comparison.of.Query.Routing.Algorithms
First, we are interested in the overall performance of the different query routing
strategies as discussed earlier in this chapter for different cluster sizes, that is, their
scalability. Therefore, we measured the performance of three different variants of
a cache approximation router versus a conventional router for cluster sizes up to
24 nodes:
??? FCFFS:.Conventional first-come-first-free-server routing which is query independent
and not cache-aware. Its performance serves as an orientation point
for the other strategies.
??? CAF: Dynamic cache approximation routing based on the FROM clause of
queries.
??? CAS:.Dynamic cache approximation routing based on predicate signatures.
??? CASweighted: Like CAS, but with a refined benefit model using normalised
benefits.
The results are shown in Figure 4 (R?¶hm et al., 2001).
Figure 4. Query routing performance and scalability: (a) throughput, (b) throughput
scaled to FCFFS, (c) mean response time, and (d) MRT scaled to FCFFS
.
Pages:
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467