S-CLONE - Data Storage for Social Networks: A Socially Aware Approach - page 39

b

M = 8 M = 16

a

FB (Random Partition)
FB (METIS)

FB (Random Partition)
FB (METIS)

1 3 5 7 1 3 5 7 9 11 13 15

K K

M = 8 M = 16

c

M = 32

FB (Random Partition)
FB (METIS)

1 5 9 1317212529

K

M = 32

Fig. 4.4 Write load balancing in terms of Gini coefficient. ( a ) M D 8 ( b ) M D 16 ( c ) M D 32

while replicating data, attempts to put those socially connected into the same server
as much as possible. Compared to random replication which is a de facto approach
for today's most OSNs, preliminary results have shown S-CLONE to be more
efficient by a substantial margin. To date, S-CLONE is the only socially aware
replication scheme applicable to OSNs that offers equal data availability for every
user and considers weighted links in the social graph and heterogeneous query rates
in the user activity. S-CLONE can also be modified easily to work with systems that
require low read load and balanced storage load instead of balanced write load.