| OCR Text |
Show 55 fca.tmPs (would not fi t in the graph witb ESM and ObServer) 2 , so while comparing POSs f'or such k:at ures, PosLg res time nwasu rements a re not taken in to co usicle ra- L iun. 5.5 .1.1 Local vs. Remote This test was done to fin d out bow each POS diff'ers in its performance in local <LJJd n-·rnote sPLtings for diffe rent u1wratio ns. For t his test, the cn.c:ate , load, and l!pd a.Le t imes are compared for each POS. T he elapsed t imes for these operat ions <l J·e plotted again st the number of OlDs in an object graph and again st the object size or an object graph with only one object . In a cli st ribJJ ted sett ing, t here is less CPU contention but there is the added cost of network tran smiss ion. In cases where th ere is less network traffic, di st r ibuted execution works better. Fi gun~ .5 .2 shows the time taken to create object graphs 1viLb varyiug number of objects and object sizes in ObServer. Figure 5.3 shows the 1 inw 1 Cl.kcn to load and Fig11n': .5.11 shows the t ime taken to update an obj ect graph \\"illl V<-Jryillg IJIJJ!lil('J or ohj e<Ls Clllcl object sizes in ObSe rve r. For load (\. Jlcl upd ate, 1l }('rl' is little diff<JJence lw t w<-'PJJ t he lo cal a.J Jcl Llw remote t inws with a va.ry 111g number of objects. I{l the case of vary ing object size , t he remote case performs wo rse t han t he local case for very large objects. This can be attributed to t he i 1Jcr easc~ d network t ra nsmi ss ion cost. Figme .5.5 shows the time taken Lo create object gr a.pb witb va. ry1ng number or ohjt'dS and o hjecL sizes in GSl\1!. Figurf: 5.6 shows tbe time ta,ken to load and F'i g111(' :) .7 s l10ws t he time taken to update a.n object graph with a varying number of obj ects and object sizes in ESM . 2ThE' rnE'asu rern ents for Pos tgres should be f'va lua t,ecl carefull y. A I though ESM and () hSer ve r pe r foJTr r better than PosLgres on nrost. o f t he Les ts 1·epor tecl her e, these t es t s are not necessarily l" f' j)l'eSP Ji tat. i vf' of what r'os tgres was cl es if!; IWd to do. |
