Determining Server Locations in Server Migration Service to Minimize Monetary Penalty of Dynamic Server Migration

Minimizing the monetary penalty and energy cost of server migration service

“…Figure 2 depicts the NetApp model assumed in SMS. It is the same model assumed for a NetApp in our previous paper 6 . A NetApp consists of one or more NetApp servers $$$ {S}_i $$$ ($$$ i $$$ = 1, 2, $$$ \dots $$$, $$$ n $$$) and one or more NetApp clients $$$ {C}_i^j $$$ ($$$ i $$$ = 1, 2, $$$ \dots $$$, $$$ n $$$, $$$ j $$$ = 1, 2, $$$ \dots $$$, $$$ {m}_i $$$), where $$$ n $$$ denotes the total number of NetApp servers, and $$$ {m}_i $$$denotes the total number of NetApp clients that NetApp server $$$ {S}_i $$$ serves, and $$$ {C}_i^j $$$ denotes the $$$ j $$$th NetApp client that the $$$ i $$$th NetApp server serves.…”

“…In this article, we use the substrate network model described below. It is the same model assumed in our previous paper 6 …”

“…Our mixed‐integer programming model is based on the integer programming model developed without considering the total energy cost in our previous paper 5,6 . Using the newly developed mixed‐integer programming model, we determine when and to which WPs NetApp servers migrate in order to minimize the total operating cost of SMS (ie, a sum of the total monetary penalty and total energy cost of SMS), while keeping the monetary penalty of the non‐SMS service at an acceptable level.…”

“…Similarly to the integer programming model in our previous paper, 5,6 our mixed‐integer programming model produces a list of WPs showing at which WP each NetApp server is located in each slot in the time period of interest. Figure 5 shows an example of energy‐aware NetApp server migration problem.…”

“…Our model determines when and to which WPs NetApp servers migrate in order to minimize the total operating cost of SMS (ie, a sum of the total monetary penalty and total energy cost of SMS), while keeping the monetary penalty of the non‐SMS service at an acceptable level during a window of $$$ W $$$ successive slots. This window is referred to as the smoothing window and avoids the negative impact of the SMS traffic (ie, server‐client, server‐server, and server‐migration traffics) from concentrating on a single slot 6 . Figure 5 includes an example of the smoothing windows.…”