In this chapter, the experimental results are described. We have users join the auction with the time-optimization algorithm and the improved algorithm separately to observe the way users assign tasks to resource and the operating status in the different deadline and the performance of the P/C. We also observe if the exclusive situations caused by Egoism exist that arises when many users bid at the same time.
The users join the auction with two job scheduling strategies, the deadline for the bidding= [500, 1000, 2000, 3000, 4000, 10000], the number of the task= [10, 60, 100], and the length of each task is between 9000~11000 MI. The experimental results show that users join the auction with the improved algorithm are able to complete the job within the deadline regardless of the quantity of the task. The users join the auction with the time-optimization algorithm are able to complete the job only under the situation of the less number of tasks and the increasing deadline. In addition, the experimental results also show that the improved algorithm presents the higher P/C which demonstrates that the users are able to complete the job with less money.
We found out that when the users join the auction with the time-optimization algorithm, they may sometimes assign a large number of tasks to only one resource and cause efficient decreased. Table 5.1 shows that the value of P/C decreases when the deadline reduces and the task number increases. Once the deadline increases or the number of task decreases, the value of P/C increases.
Table 5.1 Two Users Join the Auction with the Time-optimization Scheduling
Two Users Join the Auction with the Time-optimization Scheduling
user1 user2 user1 user2 user1 user2
number of task per user 10 60 100
total length 101428 104167 599482 601644 998075 1003061
deadline 500
turnaround time 410 414 983 986 999 984
total cost 17577.1 18041.7 405856.1 407388.7 682389.2 679928.2 Finishing rate in deadline 100% 100% 62.17% 62.71% 63.32% 64.26%
P/C 5.770462704 5.773679864 1.477080177 1.476830359 1.4626184 1.475245474
deadline 1000
turnaround time 371 425 1966 1974 1989 1987
total cost 9434.9 8559.4 222066.8 222823.3 455827.3 458407.4 Finishing rate in deadline 100% 100% 59.72% 59.61% 74.82% 75.09%
P/C 10.75029942 12.16989509 2.69955707 2.700094649 2.189590224 2.188143123
deadline 2000
turnaround time 360 433 2406 2394 3992 3969
total cost 6558.1 6345 225268.3 226250.6 394180.8 394570.7 Finishing rate in deadline 100% 100% 82.85% 83.66% 50.10% 50.94%
P/C 15.46606487 16.41717888 2.661191122 2.659192948 2.532023376 2.542157844
deadline 3000
turnaround time 358 434 2387 2409 4001 4006
total cost 6033.4 5927.4 85312 84728.2 389399.8 391185.8 Finishing rate in deadline 100% 100% 100% 100% 74.94% 74.97%
P/C 16.81108496 17.57380976 7.026936422 7.100870784 2.563111229 2.564154936
deadline 4000
turnaround time 371 427 2336 2435 4003 4002
total cost 5794 5746.8 36930.5 36265.9 202969.2 203963.4 Finishing rate in deadline 100% 100% 100% 100% 99.93% 99.95%
P/C 17.50569555 18.12608756 16.23270738 16.58979923 4.9173717 4.917848006
Figure 5.1 and Figure 5.2 demonstrate the results that user1 and user2 bid in the different deadline when complete 10 or 100 tasks. Two users were able to complete the job on time with less volume of tasks. However, once the volume of tasks increases, both users were able to complete the job only if the deadline=4000.
0
500 1000 2000 3000 4000 Deadline(second)
Time(second)
user1 user2
Figure 5.1 Two Users Complete 10 Tasks with the Time-optimization Algorithm
0
500 1000 2000 3000 4000 Deadline(second)
Time(second)
user1 user2
Figure 5.2 Two Users Complete 100 Tasks with the Time-optimization Algorithm
The result is due to different task assigning strategies with two algorithms. Figure 5.3 shows the results that user1 assigned 100 tasks to resources with the time-optimization algorithm. User1 set the standard to assign task by the performance and the price of the resource then assigned tasks starting from the resources with slowest performance and lowest cost under that standard.
deadline=500 R1 91478 968 51.66% R1 193841.9 1989 50.29% R1 394180.8 3992 50.10%
R2 474459.2 999 50.13% R2 261985.4 1009 99.16% R2 0 0
R1 389399.8 4001 74.94% R1 202969.2 4003 99.93%
R2 0 0 R2 0 0
R3 0 0 R3 0 0
R4 0 0 R4 0 0
R5 0 0 R5 0 0
Figure 5.3 User1 Assigns 100 Tasks to Resources with the Time-optimization Algorithm when There are Two Users Join the Auction at the Same Time
Table 5.2 Two Users Join the Auction with the Improved Scheduling
Two Users Join the Auction with the Improved Scheduling
user1 user2 user1 user2 user1 user2
number of task per user 10 60 100
total length 101428 104167 599482 601644 998075 1003061
deadline 500
turnaround time 402 414 499 497 497 499
total cost 10834 10799.3 170186.2 172725 348450.9 352586.8 Finishing rate in deadline 100% 100% 100% 100% 100% 100%
P/C 9.362008492 9.645717778 3.522506525 3.483247937 2.864320339 2.844862598
deadline 1000
turnaround time 381 426 969 961 990 988
total cost 5452.8 5038.6 64459.9 62911.5 241810.2 246604.6 Finishing rate in deadline 100% 100% 100% 100% 100% 100%
P/C 18.60108568 20.67379828 9.300076482 9.563338976 4.127514059 4.067486981
deadline 2000
turnaround time 370 435 1969 1962 1962 1969
total cost 4082.5 3839.3 75054.5 75389.6 88698 90711.9 Finishing rate in deadline 100% 100% 100% 100% 100% 100%
P/C 24.84458053 27.13176881 7.987289236 7.980464149 11.25250851 11.05765616
deadline 3000
turnaround time 389 448 2421 2371 2969 2962
total cost 3009.1 3110.2 16511.3 17353.2 114936.4 113673.4 Finishing rate in deadline 100% 100% 100% 100% 100% 100%
P/C 33.7070885 33.49205839 36.30737737 34.67049305 8.683715516 8.824060862
deadline 4000
turnaround time 410 410 2353 2403 3980 3980
total cost 2483.5 2583.3 10589 9764.6 157162.5 158481.4 Finishing rate in deadline 100% 100% 100% 100% 100% 100%
P/C 40.84074894 40.32322998 56.61365568 61.61481269 6.35059254 6.329203301
In the case of deadline=500, R1 were able to complete at most 26 tasks, R2 were able to complete at most 50 tasks, as for the remaining tasks, they would be sent to R3 to complete.
When the deadline=2000, the user broker judged that R1 were able to finish the entire tasks before the deadline. Therefore, all tasks were assigned to R1. Besides, R2 also adopted the same approach sending all the tasks for R1. However, R1 would not have the sufficient power to finish the huge amount of tasks from two users to meet the deadline under this situation.
When user1 and user2 couldn’t acquire enough resource to complete jobs within the deadline in each auction, the hostile competition existed in the next auction by raising prices in order to obtain more resource share. Yet, two users are still unable to obtain the sufficient amount of resources and much money is wasted at the same time.
In this study, the job scheduling is modified and improved. Assigning tasks by the standard of the resource share acquired in the first auction, it could be ensured that the overtime situation would not exist causing by the inadequate resource in the coming auction.
Table 5.2 shows the results that two users have the higher value of P/C with the improved algorithm than the time-optimization algorithm when they complete 10, 60 and 100 tasks in the different deadline.
Figure 5.4 and Figure 5.5 demonstrate the results of two users auctioning with the improved algorithm in the different deadline. The tasks were able to be completed within the deadline no matter the more or the fewer task number is.
0
500 1000 2000 3000 4000 Deadline(second)
Time(second)
user1 user2
Figure 5.4 Two Users Complete 10 Tasks with the Improved Algorithm
0
500 1000 2000 3000 4000 Deadline(second)
Time(second)
user1 user2
Figure 5.5 Two Users Complete 100 Tasks with the Improved Algorithm
Figure 5.6 shows the results that user1 assigned 100 tasks to each resource with the improved algorithm. In the case of deadline=500, user1 and user2 join the auction for R1 first, then user1 forecasted that the proportion of the resources gained is able to complete 12 tasks within the deadline, therefore, user1 assigned 12 tasks to R1. To reason by analogy, User1 assigned 25 tasks to R2 and assigned 34 tasks to R3 with the same approach, the remaining tasks were assigned to R4 finally. User1 and user2 had already assigned tasks according to the resource share they gained in the first auction; all tasks were able to be done within the deadline accordingly only if user1 and user2 maintain the same proportion of resources. By doing this, the tasks could be completed before the deadline and two users would not have to raise price constantly to acquire more resources also. Therefore, the higher P/C would exist.
deadline=500 R1 101515.8 2969 100% R1 157162.5 3980 100%
R2 13420.6 485 100% R2 0 0
R3 0 0 R3 0 0
R4 0 0 R4 0 0
R5 0 0 R5 0 0
Figure 5.6 User1 Assigns 100 Tasks to Resources with the Improved Algorithm when There are Two Users Join the Auction at the Same Time
Table 5.3 Five Users Join the Auction with the Time-optimization Scheduling
Five Users Join the Auction with the Time-optimization Scheduling
user1 user2 user3 user4 user5
number of task per user 100
total length 998075 1003061 1007222 998249 1004192
deadline 1000
turnaround time 4957 4954 4900 4938 4960
total cost 1718200 1731400 1753500 1723200 1734100 Finishing rate in deadline 30.06% 29.91% 29.78% 30.05% 29.87%
P/C 0.580884065 0.57933522 0.574406615 0.579299559 0.579085405
deadline 2000
turnaround time 9981 9921 9964 9986 9934
total cost 993100 995350 999550 993300 996550 Finishing rate in deadline 20/04% 21.04% 20.94% 20.04% 21.01%
P/C 1.005009566 1.007747024 1.007675454 1.004982382 1.007668456
deadline 3000
turnaround time 9981 10019 10045 9984 10028
total cost 993100 996800 999300 993300 997500 Finishing rate in deadline 30.06% 29.91% 29.78% 30.05% 29.87%
P/C 1.005009566 1.0062811 1.007927549 1.004982382 1.006708772
deadline 4000
turnaround time 9981 10019 10045 9984 10028
total cost 993100 996800 999300 993300 997500 Finishing rate in deadline 40.08% 39.88% 39.71% 40.07% 39.83%
P/C 1.005009566 1.0062811 1.007927549 1.004982382 1.006708772
As following we have five users join the auction with two job scheduling strategies individually to observe the performance of numbers of users with two strategies under competitive circumstances. Table 5.3 shows the results that five users bided with the time-optimization algorithm and assigned 100 tasks to resources. All users forecasted that all
tasks could be done in R1 when the deadline=2000, 3000 4000, and then caused R1 overloading and all users must spend a huge amount of money to defeat other competitive to access more resources. However, it is still unable to complete jobs within the deadline.
Figure 5.7 and Figure 5.8 demonstrate the results that five users bid with the time-optimization algorithm to compute 10 and 100 tasks with the different deadlines. If each user has 10 tasks to complete as schedule, the deadline has to be bigger than 2000. If the deadline is less than 1000, the users are unable to complete their jobs on time. However, it is impossible to finish punctually when each user has 100 tasks to be completed whatever the number of the deadline is.
750
500 1000 2000 3000 4000 Deadline(second)
Time(second)
user1 user2 user3 user4 user5
Figure 5.7 Five Users Complete 10 Tasks with the Time-optimization Algorithm
0
1000 2000 3000 4000 Deadline(second)
Time(second)
user1 user2 user3 user4 user5
Figure 5.8 Five Users Complete 100 Tasks with the Time-optimization Algorithm
deadline=1000
R1 52%
R2 48%
deadline=2000
R1 100%
cost time Finishing rate in deadline cost time Finishing rate in deadline R1 993100 4957 20.12% R1 993100 9981 20.04%
R2 1227500 2505 39.92% R2 0 0
R3 0 0 R3 0 0
R4 0 0 R4 0 0
R5 0 0 R5 0 0
deadline=3000
R1 100%
deadline=4000
R1 100%
cost time Finishing rate in deadline cost time Finishing rate in deadline R1 993100 9981 30.06% R1 993100 9981 40.08%
R2 0 0 R2 0 0
R3 0 0 R3 0 0
R4 0 0 R4 0 0
R5 0 0 R5 0 0
Figure 5.9 User1 Assigns 100 Tasks to Resources with the Time-optimization Algorithm when There are Five Users Join the Auction at the Same Time
The main reason causing overtime situation is ignoring that other appearing competitors share resources which cause the expected job loads couldn’t be completed when assigning
tasks with the time-optimization algorithm. Figure 5.9 shows the phenomenon of tasks concentrated overly in R1. R1 is able to complete the task load for one user, but if many users adopt the same strategy that make the burden of R1, it may cause all users to have the overtime situation and pay unnecessary extra cost.
Table 5.4 Five Users Join the Auction with the Improved Scheduling
Five Users Join the Auction with the Improved Scheduling
user1 User2 user3 user4 user5
number of task per user 100
total length 998075 1003061 1007222 998249 1004192
deadline 1000
turnaround time 987 985 982 982 982
total cost 783366.7 787576.2 779371.7 779454.7 788112 Finishing rate in deadline 100% 100% 100% 100% 100%
P/C 1.274084027 1.273605018 1.292351262 1.280701752 1.274174229
deadline 2000
turnaround time 1979 1972 1971 1968 1976
total cost 558053.2 566231.7 568854.6 563021.2 559264.4 Finishing rate in deadline 100% 100% 100% 100% 100%
P/C 1.788494359 1.771467405 1.770614143 1.773022046 1.795558594
deadline 3000
turnaround time 2974 2971 2977 2974 2981
total cost 703657.8 708071.8 705180 704324.5 705487.2 Finishing rate in deadline 100% 100% 100% 100% 100%
P/C 1.41840963 1.416609163 1.428319011 1.417314036 1.423402154
deadline 4000
turnaround time 3933 3940 3948 3926 3929
total cost 231301.5 226681.5 231319.7 228030.5 226329.9 Finishing rate in deadline 100% 100% 100% 100% 100%
P/C 4.315039029 4.424979542 4.354242202 4.377699474 4.436850809
Table 5.4 shows the results that five users assign 100 tasks by using the improved algorithm for bid. When all five users use the improved algorithm as the scheduling strategy, not only all tasks were completed when the deadline= 1000，2000，3000 and 4000 but there were also the higher values of P/C.
0
500 1000 2000 3000 4000 Deadline(second)
Time(second)
user1 user2 user3 user4 user5
Figure 5.10 Five Users Complete 10 Tasks with the Improved Algorithm
It demonstrates in figure 5.10 and 5.11 that all users were able to finish the jobs within the deadline with the increasing task amount and the longer finishing time.
0
1000 2000 3000 4000 Deadline(second)
Time(second)
user1 user2 user3 user4 user5
Figure 5.11 Five Users Complete 100 Tasks with the Improved Algorithm
deadline=1000
cost time Finishing rate in deadline cost time Finishing rate in deadline R1 34797 939 100% R1 66527.1 1917 100%
R2 209165 987 100% R2 370251.9 1979 100%
R3 221584.2 976 100% R3 121274.2 1725 100%
R4 271538.4 985 100% R4 0 0
cost time Finishing rate in deadline cost time Finishing rate in deadline R1 115157.2 2950 100% R1 140810.2 3933 100%
R2 568379.1 2974 100% R2 90491.3 2974 100%
R3 20121.5 426 100% R3 0 0
R4 0 0 R4 0 0
R5 0 0 R5 0 0
Figure 5.12 User1 Assigns 100 Tasks to Resources with the Improved Algorithm when there are Five Users Join the Auction at the Same Time
Figure 5.12 shows that the tasks are dispersed properly under the improved algorithm.
Although assigning tasks to R2~R5 may spend the higher price, it is the primary objective to make all tasks be finished on time without the vicious competition. It makes the P/C of total
computing higher than assigning up all tasks to R1. The price of R1 is cheaper than others, but it is unable to afford the huge amount of computing power for all users at the same time.
Table 5.5 Ten Users Join the Auction with the Time-optimization Acheduling
Ten Users Join the Auction with the Time-optimization Acheduling
user1 user2 user3 user4 user5
number of task per user 100
total length 998075 1003061 1007222 998249 1004192
Deadline 4000 turnaround time 19962 20044 20101 19966 20062
total cost 1991200 1999300 2004900 1991500 2000900 Finishing rate in deadline 20.04% 19.93% 19.86% 20.04% 19.92%
P/C 0.501242969 0.501706097 0.502380169 0.501254833 0.501870158
user6 user7 user8 user9 user10
number of task 100
total length 999764 1000642 1001314 1002260 1002053
Deadline 4000 turnaround time 19998 20012 20022 20036 20033
total cost 1994600 1995900 1996800 1998100 1997800 Finishing rate in deadline 20.00% 19.99% 19.97% 19.95% 19.96%
P/C 0.501235335 0.501348765 0.501459335 0.501606526 0.501578236
Then we increase users to 10 and 50 with 100 tasks individually to join the auction in order to observe the performance of two job scheduling strategies. Table 5.5 and table 5.6 show the results of ten users completing 100 tasks. It is shown that the value of P/C decreases while the number of user increases which makes the user pays more money to gain more resource share. However, the P/C under the improved algorithm is slightly higher than the P/C under the time optimization algorithm. Besides, the job finishing rate is lower when the tasks
are concentrated overly in R1 under the time optimization algorithm; it is guarantee that the tasks would be completed before the deadline under the improved algorithm instead.
Table 5.6 Ten Users Join the Auction with the Improved Scheduling
Ten Users Join the Auction with the Improved Scheduling
user1 user2 user3 user4 user5
number of task per user 100
total length 998075 1003061 1007222 998249 1004192
Deadline 4000
turnaround time 3915 3920 3919 3933 3928
total cost 1991200 1999300 2004900 1991500 2000900 Finishing rate in deadline 100% 100% 100% 100% 100%
P/C 0.940200666 0.931649593 0.935057966 0.937703369 0.94493423
user6 user7 user8 user9 user10
number of task 100
total length 999764 1000642 1001314 1002260 1002053
Deadline 4000
turnaround time 3930 3939 3924 3936 3942
total cost 1994600 1995900 1996800 1998100 1997800 Finishing rate in deadline 100% 100% 100% 100% 100%
P/C 0.938643303 0.93864717 0.939976757 0.948421883 0.945109224
Table 5.7 and Table 5.8 show the results of fifty users completing 100 tasks with the deadline=10000. The lower value of P/C existed which is caused by the competition for five resources by numbers of users. However, the P/C under the improved algorithm is still slightly higher than the P/C under the time optimization algorithm. The point is that under the improved algorithm it could still be guaranteed that all users finish jobs within the deadline.
The finishing rate of users with the time optimization algorithm is only about 10% which is
much lower than under the situation of less user competition.
Table 5.7 Fifty Users Join the Auction with the Time-optimization Scheduling
Fifty Users Join the Auction with the Time-optimization Scheduling
number of task per user 100
deadline 10000 user1 user2 user3 user4 user5 user6 user7 user8 user9 user10
turnaround time 99691 99990 100112 99705 100029 99813 99868 99909 99955 99945 Finishing rate in deadline 10.02% 9.97% 9.93% 10.02% 9.96% 10.00% 10.00% 9.99% 9.98% 9.98%
P/C 0.100167 0.100367 0.100662 0.100171 0.100443 0.100216 0.10025 0.100277 0.100326 0.100316 user11 user12 user13 user14 user15 user16 user17 user18 user19 user20
turnaround time 98881 99596 99799 99609 99391 100009 100133 99899 98892 99856 Finishing rate in deadline 10.11% 10.03% 10.00% 10.03% 10.06% 9.97% 9.92% 9.99% 10.11% 10.00%
P/C 0.100052 0.100142 0.100208 0.100144 0.100098 0.100392 0.100742 0.100266 0.100052 0.100241 user21 user22 user23 user24 user25 user26 user27 user28 user29 user30
turnaround time 100039 99756 99801 99626 100201 100156 99307 100052 99944 99921 Finishing rate in deadline 9.96% 10.01% 10.01% 10.03% 9.88% 9.91% 10.07% 9.95% 9.98% 9.99%
P/C 0.100456 0.100191 0.100207 0.100148 0.101062 0.100839 0.100088 0.100482 0.100312 0.100287 user31 user32 user33 user34 user35 user36 user37 user38 user39 user40
turnaround time 99940 100193 99787 100142 100096 100101 99919 99383 99502 99880 Finishing rate in deadline 9.98% 9.89% 10.01% 9.92% 9.94% 9.94% 9.99% 10.06% 10.04% 9.99%
P/C 0.100304 0.100994 0.100201 0.10077 0.100579 0.100586 0.100283 0.100096 0.10012 0.100252 user41 user42 user43 user44 user45 user46 user47 user48 user49 user50
turnaround time 99222 99989 99388 100027 99784 98839 99891 98765 100107 100174 Finishing rate in deadline 10.08% 9.97% 10.06% 9.96% 10.01% 10.12% 9.99% 10.13% 9.94% 9.90%
P/C 0.100081 0.100351 0.100097 0.10042 0.1002 0.10005 0.100257 0.10005 0.100607 0.100885
Table 5.8 Fifty Users Join the Auction with the Improved Scheduling
Fifty Users Join the Auction with the Improved Scheduling
number of task per user 100
deadline 10000 user1 user2 user3 user4 user5 user6 user7 user8 user9 user10
turnaround time 9901 9844 9834 9895 9854 9864 9886 9831 9838 9914 Finishing rate in deadline 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
P/C 0.131992 0.133608 0.134991 0.154199 0.132931 0.132967 0.13369 0.134305 0.133425 0.134252 user11 user12 user13 user14 user15 user16 user17 user18 user19 user20
turnaround time 9835 9796 9860 9899 9814 9856 9805 9813 9889 9810 Finishing rate in deadline 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
P/C 0.131939 0.132648 0.134453 0.132883 0.133782 0.134072 0.134287 0.135437 0.132381 0.133277 user21 user22 user23 user24 user25 user26 user27 user28 user29 user30
turnaround time 9830 9789 9855 9841 9839 9823 9865 9846 9844 9867 Finishing rate in deadline 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
P/C 0.133711 0.132697 0.132346 0.132691 0.13423 0.133952 0.132373 0.134088 0.134301 0.133446 user31 user32 user33 user34 user35 user36 user37 user38 user39 user40
turnaround time 9877 9815 9857 9911 9851 9843 9880 9845 9892 9847 Finishing rate in deadline 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
P/C 0.134324 0.134433 0.133682 0.133885 0.134057 0.135267 0.133468 0.133342 0.131903 0.134801 user41 user42 user43 user44 user45 user46 user47 user48 user49 user50
turnaround time 9850 9832 9907 9839 9787 9849 9862 9883 9807 9818 Finishing rate in deadline 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
P/C 0.133545 0.135263 0.131975 0.134553 0.133228 0.132307 0.132552 0.131683 0.133729 0.135436
The strategy of assigning tasks is the main reason causing such a big gap between the time optimization algorithm and the improved algorithm. Figure 5.13 shows that the improved algorithm assigns 100 tasks to five resources according to the result of first auction and each resource is able to complete the job within the deadline.
cost time Finishing rate in deadline R1 348026.3 9383 100%
R2 1893694.9 9738 100%
R3 2176469.1 9747 100%
R4 2677413.8 9901 100%
R5 466026.1 6038 100%
Figure 5.13 User1 Assigns 100 Tasks to Resources with the Improved Algorithm when There are Fifty Users Join the Auction at the Same Time
The experiment shows that users could obtain better results by using the improved algorithm instead of the time-optimization algorithm under the condition that users bid at the same time. Users could assign tasks flexibly to resources regardless of the longer and the shorter the deadline is without the vicious competition that causes money waste.
Using cheaper resource in computing is the expectation under both the time-optimization algorithm and the improved algorithm. By acquiring cheaper resources, more money would be saved and the tasks would be finished within the deadline. However, it will be counter-productive if the influence of resources sharing by other competitors is ignored. The improved algorithm makes the result of first auction as the standard to assign tasks to resources which excludes the case of competition at the beginning. It makes all users adjust the bid price flexibly and finish all tasks within the deadline. Although part of the tasks would be assigned to the more expensive resource for the implementation, it is still guaranteed that all the jobs would not be finished overtime, which is our primary objective. In addition, the
implementation of decentralization would not lead to the vicious competition; instead, it is more efficient and economical than concentrated the tasks in the cheaper resource. The lower total cost and the higher value of P/C means that the improved algorithm meet users’
expectations better.