### Table 2: Overloaded Distributed Matrix Operations.

"... In PAGE 4: ... It is within this le that calls to ppclient are made in order to perform the appropriate computation. Table2 lists functions... In PAGE 4: ...dense matrices. For dsparse matrices, individual elements can be retrieved and set. 3.3 p: Towards transparent constructors One problem with the constructors described in Table2 is that they do not directly corre- spond to Matlab apos;s matrix constructors such as zeros or rand. Users must learn a new set of functions for creating distributed objects.... ..."

### Table 2: Overloaded Distributed Matrix Operations.

"... In PAGE 4: ... It is within this le that calls to ppclient are made in order to perform the appropriate computation. Table2 lists functions... In PAGE 4: ...dense matrices. For dsparse matrices, individual elements can be retrieved and set. 3.3 p: Towards transparent constructors One problem with the constructors described in Table2 is that they do not directly corre- spond to Matlab apos;s matrix constructors such as zeros or rand. Users must learn a new set of functions for creating distributed objects.... ..."

### Table 2: Overloaded Distributed Matrix Operations.

"... In PAGE 4: ... It is within this le that calls to ppclient are made in order to perform the appropriate computation. Table2 lists functions... In PAGE 4: ...dense matrices. For dsparse matrices, individual elements can be retrieved and set. 3.3 p: Towards transparent constructors One problem with the constructors described in Table2 is that they do not directly corre- spond to Matlab apos;s matrix constructors such as zeros or rand. Users must learn a new set of functions for creating distributed objects.... ..."

### Table 5.5 Performance on small test problems.

1998

Cited by 44

### Table 1. Overload due to failures.

2001

"... In PAGE 7: ... Under high utilization, this is unrealistic. Table1 generalizes this analysis to replica groups with n nodes. Losing two of five nodes in a replica group, for example, implies a redirected load of 2/3 extra load (two loads spread over three remaining nodes) and an overload factor for those nodes of 5/3 or 166 percent of normal load.... In PAGE 8: ... Under the giant-scale services model, the basic question is how many locations to establish and how many replicas to put at each. To examine the load redirection problem, I return to Table1 . With two replicas at each of three locations, for exam- ple, we expect to lose 2/6 of the replicas during a natural disaster, which implies that each of the remaining sites must handle 50 percent more traf- fic.... ..."

Cited by 112

### Table 1. Overload due to failures.

2001

"... In PAGE 7: ... Under high utilization, this is unrealistic. Table1 generalizes this analysis to replica groups with n nodes. Losing two of five nodes in a replica group, for example, implies a redirected load of 2/3 extra load (two loads spread over three remaining nodes) and an overload factor for those nodes of 5/3 or 166 percent of normal load.... In PAGE 8: ... Under the giant-scale services model, the basic question is how many locations to establish and how many replicas to put at each. To examine the load redirection problem, I return to Table1 . With two replicas at each of three locations, for exam- ple, we expect to lose 2/6 of the replicas during a natural disaster, which implies that each of the remaining sites must handle 50 percent more traf- fic.... ..."

Cited by 112

### Table 5.1 Description of large test problem set.

1998

Cited by 44

### Table 1: Creating an Abstraction Hierarchy

"... In PAGE 2: ... An abstract space does not simply involve dropping preconditions or goals;; instead an abstract space is an abstract model of the original problem space, where both the operators and states are simpli ed. Table1 de nes the algorithm for creating a problem- speci c abstraction hierarchy. The rst step in the al- gorithm produces a set of constraints on the order of the literals in an abstraction hierarchy.... ..."

### Table 1. Correspondences created in i* to model a problem frame

"... In PAGE 7: ... This criterion makes it possible to follow a simple way for representing all frames. Table1 describes the elements created in the i* model, in order to represent each of the concepts presented in a problem frame, that is, in a class of problem. We have maintained the Problem Frames marks for phenomena (C-causal, Y-lexical, E-event), as element names in the i* models, (see Figure 2, 6, 10, 14 and18).... ..."

### TABLE IV UTILITY IN OVERLOAD.

2006