An application running on IRF is likely to require significant resources:
Here are some data on the resources used by IRF/SampleApp to index and retrieve from a collection of 25,000 text documents from the Foreign Broadcast Information Service that comprise approximately 90 megabytes of data.
The SampleApp was run on an otherwise very lightly loaded UltraSparc Model 60 (296mHz, 512mB of RAM, 1gB of swap) using Java virtual machine size settings of 400mB initial and max. Runs on PCs with about the same clock speeds and available memory took significantly longer.
Note that the raw document text is not requried after conversion and indexing - the indexes contain it. Two indexes are built for each document field - one by document and one by term. Positional information is kept in the index.
Disk
space File
(in bytes) name
----------- ----------------------
440 128 DB.FBIS
135 846 760 DB.HanD
87 346 635 DB.HTML
85 334 DB.IF
3 409 987 DB.IFWS
157 021 391 DB.IFWT
21 525 DB.Indx
1 DB.OID
3 625 815 DB.Str
166 388 586 DB.VecP
3 092 DBdocAbstractindexsBv0
46 642 DBdocAbstractindexvBs0
67 DBdocNumberindexsBv0
625 017 DBdocNumberindexvBs0
219 DBfileNames
2 385 092 DBtextindexsBv0
625 017 DBtextindexvBs0
285 242 DBtitleindexsBv0
624 167 DBtitleindexvBs0
-----------
558 780 717 = 6.2 times raw doc collection size
Elapsed seconds to convert and index = 21108 = 5.9 hrs
Elapsed seconds to update = 21549 = 6.0 hrs.
We tested retrieval by starting the sample application, choosing the existing FBIS collection, and performing 100 retrievals: 50 for which the queries were the 2- or 3- word title sections from the TREC-8 ad hoc topics (401-450) and 50 for which the queries were the sentence-length description sections from the same topics. Here are the total elapsed time in seconds for each retrieval and the number of documents retrieved:
TREC Topic | Title Description | -------- ----------- | Time Time | | Docs | Docs | | | | | 401 262 9073 572 17613 402 2 525 116 7779 403 0 0 144 10082 404 82 7846 208 12370 405 10 1832 221 10522 406 2 325 77 6440 407 4 925 77 6639 408 1 231 41 4816 409 13 2229 109 9496 410 38 4901 220 12472 411 1 122 74 6365 412 58 6238 252 11229 413 44 3769 193 10845 414 12 2209 341 12894 415 6 1057 93 7081 416 14 2358 9 1373 417 5 1096 118 8092 418 4 1022 170 10213 419 2 394 413 15039 420 1 222 7 999 421 31 4156 77 6070 422 3 574 9 1320 423 1 119 9 1076 424 0 115 35 4210 425 9 1894 21 2781 426 36 4111 179 11112 427 7 1370 111 8060 428 24 3264 228 11941 429 1 328 7 1134 430 13 2042 41 4785 431 10 1610 114 7652 432 13 2108 69 7229 433 3 533 4 647 434 29 3442 816 12191 435 18 3005 348 14510 436 3 657 13 2258 437 9 1530 87 7480 438 40 4533 239 12295 439 6 1135 163 9966 440 7 1464 309 13205 441 1 324 11 2429 442 17 2816 157 10396 443 21 3084 276 12290 444 1 49 65 6432 445 10 705 625 17903 446 6 1049 49 5965 447 5 882 242 12892 448 7 1552 81 7431 449 1 123 208 12143 450 40 4528 321 13921
Last updated: Tuesday, 01-Aug-2000 06:34:34 MDT