Capacity Planning and Architecture - Expert Oracle RAC Performance Diagnostics and Tuning - page 44

9

13

10

SSKY2 SSKY3 SSKY4

SSKY1

R R7

R1 R5

R3

14

500:9996

R2

R4

500:9996
500:10010

500:9996

12 11

SCN 9996

Block 450
Block 490

Block 459 Block 550

Block 500

Block 600

SSKYDB

Figure 2-9. Read/write behavior

These are the steps shown in the figure:

12. Instance SSKY3 , based on the request received from the GCS, will

a.

send the block to instance SSKY1 along with an indicator that it is closing its own
resource and giving an exclusive resource for use to instance SSKY1 ;

b. close its own resource, marking the buffer holding the block image as copy for

consistent read (CR) and informing itself that the buffer area is available for reuse.

13. Instance SSKY1 converts its resource, makes the required updates to the block, and assigns

it a new SCN number. SSKY1 then sends a message to the GCS indicating/confirming that it
has an exclusive resource on the block. The message also piggybacks the message received
from instance SSKY3 indicating that it has closed its own resource on this block. GCS now
updates the GRD regarding the status of the block, and instance SSKY1 can now modify
the block.

Please note that at this stage, the copies of blocks on other instances would also be
removed from the GRD.
As illustrated in Figure 2-9 , instance SSKY1 has now modified the block and the new SCN is 10010.

The GCS confirms with instance SSKY3 that it has received notification regarding the status
of the block in its buffer.

14.