- 论坛徽章:
- 0
|
Sun4500 2G 内存 2个CPU
我主要配置了以下设置:
max memory:786432
default data cache:1000M IO 2k:910M 16K:90M
procedure cache size:125000(2K)
max online engines 和 number of engines at startup均配为1
number of locks配了10000.
number of user connections配了250
但应用速度比较慢,connection多了尤其慢.
用sp_sysmon看了一下.(于另外一个使用正常的Server进行比较.)
发现
1.Transaction Detail 中
Updates 80%为APL Direct In-place,而正常的Server 100%为APL Direct Cheap
Delete 100%为 APL Direct, 而正常的Server 100%为APL Deferred
Transaction Detail per sec per xact count % of total
------------------------- ------------ ------------ ---------- ----------
Inserts
APL Heap Table 374.5 177.6 449420 99.9 %
APL Clustered Table 0.2 0.1 293 0.1 %
Data Only Lock Table 0.0 0.0 0 0.0 %
------------------------- ------------ ------------ ---------- ----------
Total Rows Inserted 374.8 177.8 449713 99.2 %
Updates
APL Deferred 0.0 0.0 0 0.0 %
APL Direct In-place 2.3 1.1 2779 79.0 %
APL Direct Cheap 0.6 0.3 733 20.8 %
APL Direct Expensive 0.0 0.0 6 0.2 %
DOL Deferred 0.0 0.0 0 0.0 %
DOL Direct 0.0 0.0 0 0.0 %
------------------------- ------------ ------------ ---------- ----------
Total Rows Updated 2.9 1.4 3518 0.8 %
Data Only Locked Updates
Total Rows Updated 0.0 0.0 0 n/a
------------------------- ------------ ------------ ---------- ----------
Total DOL Rows Updated 0.0 0.0 0 0.0 %
Deletes
APL Deferred 0.0 0.0 0 0.0 %
APL Direct 0.0 0.0 29 100.0 %
DOL 0.0 0.0 0 0.0 %
------------------------- ------------ ------------ ---------- ----------
Total Rows Deleted 0.0 0.0 29 0.0 %
========================= ============ ============ ==========
Total Rows Affected 377.7 179.2 453260
2.
Lock Management中 Lock Hashtable Lookups非常大
Lock Summary per sec per xact count % of total
------------------------- ------------ ------------ ---------- ----------
Total Lock Requests 23973.4 11370.8 28768023 n/a
Avg Lock Contention 0.0 0.0 41 0.0 %
Deadlock Percentage 0.0 0.0 2 0.0 %
Lock Hashtable Lookups 24116.4 11438.6 28939686 n/a
Avg Hash Chain Length n/a n/a 0.00035 n/a
Lock Detail per sec per xact count % of total
------------------------- ------------ ------------ ---------- ----------
不知道是什么原因
3.
cache turnover比较多
Cache Turnover
Buffers Grabbed 18.2 8.6 21790 n/a
Buffers Grabbed Dirty 0.0 0.0 0 0.0 %
请问该如何改进? |
|