Unsolved

This post is more than 5 years old

165 Posts

1754

March 26th, 2009 15:00

Question about tool to send load to a disk

Hi

I installed a Solaris server 2 days ago, I assigned disk to it, I want to know the following:

is there a EMC or third part tool to send load to the disks, I mean, I need to deliver the host to the database team, but I want to check if the host does not have io problems, I have to run any tool to make sure.

Regards

1.3K Posts

March 26th, 2009 15:00

IORATE is an excellent IO generator for Unix.

You can download the source and binaries from the public FTP site (ftp.emc.com) in the /pub/elab/iorate/ directory

2 Intern

 • 

1.3K Posts

March 26th, 2009 17:00

can you check if there is 64bit version available for Linux? Or 32bit can be used for both?

1.3K Posts

March 26th, 2009 18:00

I'm not positive if the 32 bit Linux version we have posted will operate OK on 64 bit systems or not, but the source is there too, so you can compile on a 64 bit system.

1.3K Posts

March 27th, 2009 06:00

Sure, I'll be glad to help.

I rarely use files for the devices, I usually use raw devices.

If you want to use files, you need to create a file of the proper size. The touch command created a file, but is is of a zero size.

You could use dd to create your 1GB file.

dd if=/dev/zero of=/tascb_data1/test.txt bs=1024k count=1024

(1MB IOs and 1024 of them)

Also I'm not sure what patterns or tests are in the default files. You may want to at least look at them to see what sort of IO you are going to generate.

165 Posts

March 27th, 2009 06:00

Hi

I dowloaded the IORATE tool, I sent it by ftp to my server, I untar the file.

I modified the files
devices.ior


But it did not work well, could you help me please?

In My server I dont have database installed, I have only the file systems,

I want to run the IOrate for this file system /tascb_data1,

I modified the file devices.ior to

Device = "/tascb_data1/test.txt" capacity 1GB;

Let me tell you that the file test.txt was created for this test by using touch command.

the other two file patterns.ior tests.ior were not modified.


I'll appreciate your help on this

Regards

6 Operator

 • 

2.8K Posts

March 27th, 2009 06:00

Don't forget to mark question as answered when you solved your problems and to award points marking correct/helpful answers.

1.3K Posts

March 27th, 2009 06:00

I'm not in this for the points, just trying to be helpful :-)

6 Operator

 • 

2.8K Posts

March 27th, 2009 07:00

You are absolutely right .. That's why I said it's crucial :-)

6 Operator

 • 

2.8K Posts

March 27th, 2009 07:00

Nor do I .. but awarding points is crucial for our forum's existence. And someone usually doesn't award points .. ;-)

2 Intern

 • 

292 Posts

March 27th, 2009 07:00

Sometimes there are a lot of different opinions and options in a thread and by marking helpful/correct answers it is an easy way for others who may have the same question to know what worked best.

165 Posts

March 27th, 2009 08:00

Hi

I just ran the IORATE tool and got this values:

device name:/tascb_data21/hola.dat
total_sec:600
writes 464483
KB_written:928966
write_resp 1.209
total_I/Os 464483
total_KB 928966
total_resp 1.209
I/Os_per_sec:814


I think that those values are good, for the total_resp but does any tell me your comments?

Regards

1.3K Posts

March 27th, 2009 09:00

I need to know what the "test" was. Can you please post the tests.ior file and the patterns.ior file? Just the uncommented lines from the tests.ior file and the pattern types that were used in the test.

Also how many FAs were active, and what sort of BE configuration the file was placed on (RAID5/6/1, spindle count, speed of disks, etc)

1.3K Posts

March 27th, 2009 10:00

That seems low for a 2k random write across that many drives. I forgot to ask for your devices file too, because that will determine the thread count. With a single threaded 2k write test, you should be getting close to 4,000 IOs/sec (assuming a DMX3/4 running 72 or 73 code)

Maybe you should start with a front-end read hit test, just to make sure the front end ports/switches are performing up to par before we move to a backend test.

I would suggest first front end test with multiple threads, both large and small IOs. Then back end reads, large and small, then move onto backend writes.

165 Posts

March 27th, 2009 10:00

Hi

I have two FA dedicated to this server
It is RAID 5
146Gb 15RPM
8-way metas 1920 stipe size (oracle Recomendation)

pattterns.ior file

Pattern 1 = "2k Seq Read" io size 2KB sequential read;
Pattern 2 = "2k Random Read" io size 2KB random read;
Pattern 3 = "2k Seq Write" io size 2KB sequential write;
Pattern 4 = "2k Random Write" io size 2KB random write;
Pattern 5 = "8k Seq Read" io size 8KB sequential read;
Pattern 6 = "8k Random Read" io size 8KB random read;
Pattern 7 = "8k Seq Write" io size 8KB sequential write;
Pattern 8 = "8k Random Write" io size 8KB random write;
Pattern 9 = "64k Seq Read" io size 64KB sequential read;
Pattern 10 = "64k Random Read" io size 64KB random read;
Pattern 11 = "64k Seq Write" io size 64KB sequential write;
Pattern 12 = "64k Random Write" io size 64KB random write;
Pattern 13 = "512 Random Write" io size 512 random write;
Pattern 14 = "2k Random Read" io size 2KB random read;
Pattern 20 = "4k Mixed random 50% Hit" io size 4KB random read 50%
history 250 reuse 50%;
Pattern 21 = "4k random read 4 zones" io size 4KB random read
zone 5% count 4;

test.ior file

Test = " 2k Bus Test" for 600 sec ignore 30 sec size 1GB 100% pat 4;

Regards

165 Posts

March 27th, 2009 11:00

Hi

Device file is VXFS and FS

Regards

0 events found

No Events found!

Top