Tuning the Cluster Interconnect - Expert Oracle RAC Performance Diagnostics and Tuning - page 481

The following output gives the list for a three-node RAC cluster:
Script: MVRACPDnTap_dlmmisc.sql
SELECT * FROM (SELECT INST_ID, NAME, VALUE FROM GV$DLM_MISC) PIVOT (SUM(value) for INST_ID IN
(1,2,3)) ORDER BY NAME;
NAME 1 2 3
-------------------------------------- ---------- ---------- ----------
messages flow controlled 55822 84833 96642
messages queue sent actual 5453201 7641557 6619674
messages queue sent logical 10441169 9808352 10101283
messages received actual 16978118 18765414 15525485
messages received logical 39326500 42175391 46091899
messages sent directly 6695061 6116041 8963878
messages sent indirectly 13049447 12369993 13557761
messages sent not implicit batched 1062 1099 1347
messages sent pbatched 24555803 26937853 36002459
msgs causing lmd to send msgs 4354994 6659117 3991411
msgs causing lms(s) to send msgs 2750828 2594875 1362431
msgs received queue time (ms) 12957956 13543472 9815095
msgs received queued 39326365 42175626 46091875
msgs sent queue time (ms) 858676 880076 1419012
msgs sent queue time on ksxp (ms) 53097467 58007840 77175847
msgs sent queued 14943725 16421747 15029984
msgs sent queued on ksxp 14755004 16571618 19910123
number of broadcasted resources 0 0 0
number of ges deadlock detected 0 0 0
process batch messages received 7265951 6886496 7716180
process batch messages sent 14623640 15792805 7610679
Enterprise Manager to Monitor Cluster Interconnect
The cluster database performance page in Figure 14-13 gives a glimpse of the overall performance of the database.
This EM page has four sections:
Cluster host load average: This shows maximum, average, and minimum values for
available hosts in the cluster for the previous hour.
Global cache block access latency: This chart gives the access latency for a block request
between all instances in the cluster.
Average active sessions: Probably one of the most important charts in the page shows
potential problems inside the cluster database. The chart displays wait classes showing how
much time the cluster database is waiting for a resource.
Through charts: This section of the page displays throughput by I/O parallel execution,
services, and instances.