Avg. Disk sec/Read,Avg. Disk sec/Write,Current Disk Queue Length都没打到阈值，但是%Disk Time超过100%了，请问这样算是正常的吗？

Question

Avg. Disk sec/Read,Avg. Disk sec/Write,Current Disk Queue Length都没打到阈值，但是%Disk Time超过100%了，请问这样算是正常的吗？

前面两项都是小于0.03 第三项小于2.但是第四项就大于100%了。

这个算是正常吗？

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 1

你要看100% - idle time，因为多个磁盘/磁头的关系，disk time不准确

---

想不想时已是想，不如不想都不想。

Answer 2

你要看100% - idle time，因为多个磁盘/磁头的关系，disk time不准确

---

想不想时已是想，不如不想都不想。

你的意思是如果% Idel time很大的话，那么其实磁盘就不是那么的繁忙，即使%Disk Time看起来很高，也不会存在问题吗？

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 3

You answered your own question.

Answer 4

% Disk Time	% Idle Time	Avg. Disk sec/Read	Avg. Disk sec/Write
25.40701219	93.82069864	0.0137424	0.000362291
108.089258	92.90446623	0.034604193	0.002022498
48.34773839	93.0939626	0.015679886	0.000420457
6750.234895	78.60590743	0.364771902	0.527571379
164.1065992	90.41576402	0.039507703	0.000593254
8868.277979	68.39152022	0.326508629	0.323287097
2081.94811	84.3157978	0.115787018	0.005072511
46.47627432	93.14087836	0.022609781	0.000899997
8777.304476	64.87275444	0.256961684	0.09892702
11776.81527	27.34259265	0.126571454	0.057373537
3406.219068	78.77413759	0.163221179	0.116621014
223.5995661	89.67014944	0.033308632	0.003031479
93.30450784	91.72572796	0.043553681	0.000656108
7.305146322	95.25264487	0.006265659	0.001546347
19.25167382	96.32911731	0.013076641	0.002168026
3272.229148	80.46644006	0.138721187	0.056768851
3304.109998	85.94729975	0.311473417	0.007488657
8.492146917	94.8478346	0.006475717	0.000341576
307.4702632	95.43633429	0.142497026	0.000501073
12.22826522	95.09592416	0.009145181	0.000688272
50.08420505	96.11398794	0.032174933	0.000585187
45.40287826	97.0852193	0.022641457	0.005795795
17.08933855	97.94370281	0.028382133	0.000392501
6.455209393	96.99705432	0.008316468	0.0017575
503.2155883	88.09405859	0.038956742	0.01323393
6.309712187	97.7745394	0.006363191	0.001192374
13.81523475	98.01478478	0.017389018	0.002527277
15.9791932	91.95348449	0.003994904	0.000497916
13.49282427	97.44237911	0.017040741	0.000598372
7.104196933	97.81770523	0.010700757	0.0009875

 

看起来% Idel Time都蛮大的，也就是说不恨频繁的被使用。但是这个时候%Disk Time这么大说明了什么问题呢？

而且

Avg. Disk sec/Read

Avg. Disk sec/Write

这两个值大部分时间都不太正常。。。

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 5

Here are definitions:

% disk time - the percentage of elapsed time that the selected disk drive was busy servicing read or write requests.

% idle time - the percentage of time during the sample interval that the disk was idle.

Did you look those numbers for total of all disks or individual disk? I'll not say server works abnormal just because one perfmon count has unusual number occasionally.

Answer 6

Here are definitions:

% disk time - the percentage of elapsed time that the selected disk drive was busy servicing read or write requests.

% idle time - the percentage of time during the sample interval that the disk was idle.

Did you look those numbers for total of all disks or individual disk? I'll not say server works abnormal just because one perfmon count has unusual number occasionally.

这是单个磁盘的监控结果。

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 7

今天看监控结果感觉出现瓶颈了。是针对单个磁盘的

首先：

 %DiskTime的平均值是6220.746938，里面80%以上都是在1000以上。

Avg. Disk sec/Read和Avg. Disk sec/Write差不多为0.238253和0.140705。

再看了下，Read Queue Length大部分时间都在400多，Write Queue Length在5以上。

Page Life expectancy只有100-200了。但是Free pages至少在1000以上。看了下可用的内存是700多M。

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 8

磁盘是有瓶颈。不过你首先应该关注sql的waits。

---

想不想时已是想，不如不想都不想。

Answer 9

今天看监控结果感觉出现瓶颈了。是针对单个磁盘的

首先：

 %DiskTime的平均值是6220.746938，里面80%以上都是在1000以上。

Avg. Disk sec/Read和Avg. Disk sec/Write差不多为0.238253和0.140705。

再看了下，Read Queue Length大部分时间都在400多，Write Queue Length在5以上。

Page Life expectancy只有100-200了。但是Free pages至少在1000以上。看了下可用的内存是700多M。

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

What do you have on that disk?

Answer 10

今天看监控结果感觉出现瓶颈了。是针对单个磁盘的

首先：

 %DiskTime的平均值是6220.746938，里面80%以上都是在1000以上。

Avg. Disk sec/Read和Avg. Disk sec/Write差不多为0.238253和0.140705。

再看了下，Read Queue Length大部分时间都在400多，Write Queue Length在5以上。

Page Life expectancy只有100-200了。但是Free pages至少在1000以上。看了下可用的内存是700多M。

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

What do you have on that disk?

在这个磁盘上仅仅包含数据库的数据文件。

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 11

磁盘是有瓶颈。不过你首先应该关注sql的waits。

---

想不想时已是想，不如不想都不想。

你的意思是？

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 12

Use sys.dm_io_virtual_file_stats (if your sql has it) to find out which db file has high activity then ask db user why, may need move that db to another disk if it causes server performance problem. 

Answer 13

Use sys.dm_io_virtual_file_stats (if your sql has it) to find out which db file has high activity then ask db user why, may need move that db to another disk if it causes server performance problem. 

现在仅仅一个用户数据库的数据文件放置在那个磁盘上的。

我以前就是通过sys.dm_io_virtual_file_stats来观察IO的，后来将日志文件和tempdb的文件都移动到另外个磁盘了，现在这个磁盘上仅仅剩余这个用户数据库的数据文件。

 

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 14

Does user complain performance? Did you check with user for db activities? You have to work with user to find and solve the issue.

Answer 15

Does user complain performance? Did you check with user for db activities? You have to work with user to find and solve the issue.

现在是在说性能有点问题，于是我观察性能指标，发现这部分IO比较大，以前也是比较大的。

我看了下，有些时候，相同的sid，但是存在30-40个线程。。。这一部分sql已经抓取出来了。

---

If you haven't all the things you want,be grateful for the things you don't have that you didn't want.

Answer 16

Find out what did the spid do and why. I had similar problem last week and found one app caused heavy disk i/o, so I contacted app owner for reason. They looked at app config, found one option checked by mistake that logs huge amount of data to sql table. The point is that you can't fix performance issue yourself along.