### Table 2. Estimated loss rates after synchronization, as a function of the number K of servers used, for the two load models considered

"... In PAGE 9: ...1. In Table2... In PAGE 11: ... 3. The PSQA function used as shown in Table2 . Figure 3 shows that the worst case is when MLBS = 1.... In PAGE 11: ... Figure 3 shows that the worst case is when MLBS = 1. Using this value and the loss rates given in the first column of Table2 , we obtain, as perceived quality estimates, values going from 10 (maximal quality) to approximately 9, which is almost maximal. For the geometric load (somehow an extreme situation), we observe loss ratios up to 0.... ..."

### Table 1. Mean Squared Error between theoretical and observed values of the used bandwidth, as a function of the total number K of servers. The error is computed summing on i, the server index, from 1 to K.

"... In PAGE 8: ...5. In Table1 we show the Mean Squared Error between the bandwidth measured during our experiments and the theo-... ..."

### Table 2 Details for H20855tpc-wH20856 for both servers.

"... In PAGE 11: ... In short, we feel that these results provide a conservative estimate of the energy efficiencies possible under varying workloads. Looking in greater detail into the performance outlined in Table2 , we see significantly greater NFS server activity for the SDS blades as opposed to the 1-U server. The image data set for our evaluation problem size, 100K, would fit into 4 GB of memory, the amount available on a fully configured 1-U server, which it could share with a second processor on a 1-U server.... ..."

### Table 8.1. Mirrored instances without place constraints. Instance Breaks TPA PGBA

2007

Cited by 5

### Table 8.2. Non-mirrored instances without place constraints. Instance Breaks TPA PGBA Instance Breaks TPA PGBA

2007

Cited by 5

### Table 2: Linear model estimation

2006

"... In PAGE 9: ...Computational experience (MIPLIB instances) 3 OUR METHOD Table2 compares the size of the measurement tree obtained by the linear model with the actual number of nodes in T. The last column shows the ratio between the two.... ..."

Cited by 2

### Table 1 does not show any update related communication because there is no write-sharing on 16 node system for the selected problem size. Di erence in the number of page requests and page replies is due to transfer of pages by the dynamic ownership protocol. Communication activity shown in last row of the table is because of copy set maintenance required to implement update based protocol. This application favors large data block size due to little sharing and little synchronization.

1997

"... In PAGE 7: ... Server 0.5K 1K 2K 4K Page Request Server 69250 34877 17459 8769 Page Reply Server 21393 10700 5351 2679 Update Copy Set Server 98881 49862 24996 12470 Table1 : MATMULT: the server calls for 16 node system Table 1 does not show any update related communication because there is no write-sharing on 16 node system for the selected problem size. Di erence in the number of page requests and page replies is due to transfer of pages by the dynamic ownership protocol.... ..."

Cited by 1

### Table 2: A sub-protocol for one-round (k; m)-DOT-?n1 .

"... In PAGE 13: ...ub-protocol. The general structure of the subprotocol is given in Table 1. In this section we describe one-round (k; m)-DOT-?n1 oblivious transfer protocols, which generalize the one-round (k; m)-DOT-?2 1 protocols proposed in [34]. In Table2 we describe 6Notice that, a (k;m)-DOT-?2 1 can be used as a black box to set up \more complex quot; oblivious transfer protocols in the same distributed model (see [23, 19, 11] for unconditionally secure reductions). In this case, any improvement in the design of the available (k;m)-DOT-?2 1 , implies directly an improvement of the performance of the more complex protocols.... In PAGE 14: ... Correctness. We show that the sub-protocol given in Table2 satis es De nition 4.1.... In PAGE 16: ...alues. Hence, the k ? 1 values do not give information about the secrets. Condition (4) is not satis ed. Indeed, it is possible to show that in the protocol given in Table2 the Receiver can learn a linear combination of the secrets. Indeed, if the Receiver does not follow the protocol and chooses certain values (Zy1(0); : : :; Zyn?1(0)), say for example (2; 3; : : :; 1), then she gets a linear combination of the secrets s0; : : :; sn?1.... In PAGE 16: ... Indeed, if the Receiver does not follow the protocol and chooses certain values (Zy1(0); : : :; Zyn?1(0)), say for example (2; 3; : : :; 1), then she gets a linear combination of the secrets s0; : : :; sn?1. Notice that, in [34], for the case of two secrets, a proof that the Receiver can get no more than a single linear combination by running the sub-protocol described in Table2 with k servers was given. It is not di cult to show that the proof easily generalises to the case of n secrets (we refer the interested reader to [34].... In PAGE 16: ...) Such a result can be used to construct a (k; m)-DOT-?n1 ; forcing the Receiver to get at most one of the secret held by the Sender and no joint information about the secrets, by using multiple instances of the sub-protocol. More precisely, the sub-protocol given in Table2 , can be used as a building block to set up a (k; m)-DOT-?n1 . The scheme is given in Tables 3 and 4.... In PAGE 17: ... Let X be a subset of k servers. The Receiver sends to each server in X the same queries described in Table2 , that is, R sends server Sj 2 X the values Zy1(j); : : :; Zyn?1(j), but re- ceives, from each of the k servers, n points, e.g.... In PAGE 17: ... The privacy property, stated by De nition 2.2, can be shown as follows: Condition (3) and Condition (5) follow exactly from the same argument we have applied discussing the protocol given in Table2 . Actually, notice that we are basically repeating n times the protocol of Table 2, with the constraint that the Receiver sends a single sequence of queries instead of n distinct sequences of queries.... In PAGE 17: ...rivacy. The privacy property, stated by De nition 2.2, can be shown as follows: Condition (3) and Condition (5) follow exactly from the same argument we have applied discussing the protocol given in Table 2. Actually, notice that we are basically repeating n times the protocol of Table2 , with the constraint that the Receiver sends a single sequence of queries instead of n distinct sequences of queries. Condition (4) can be shown by noticing that if the Receiver cheats by sending a disal- lowed sequence of queries, in order to learn information about more than one secret, she gets some linear combinations 1; : : :; n of s0; : : :; sn?1, from which she does not learn anything about the secrets.... In PAGE 19: ... The same strategy applied in Tables 3 and 4 can be used to set up a t-private one-round (k; m)-DOT-?n1 . Actually notice that, when the number of secrets n is equal to 2, the sub-protocols given in Table2 and Table 5 become exactly the sub-protocols given in [34], and the techniques we have used to develop the general case are an extension of the ones therein presented.... ..."

### Table 5.5 Performance on small test problems.

1998

Cited by 44