HPE GreenLake Administration
- Community Home
- >
- Servers and Operating Systems
- >
- Operating Systems
- >
- Operating System - HP-UX
- >
- LVM raw devices vs real raw devices
Operating System - HP-UX
1837655
Members
2647
Online
110117
Solutions
Forums
Categories
Company
Local Language
back
Forums
Discussions
Forums
- Data Protection and Retention
- Entry Storage Systems
- Legacy
- Midrange and Enterprise Storage
- Storage Networking
- HPE Nimble Storage
Discussions
Forums
Discussions
Discussions
Discussions
Forums
Discussions
back
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
- BladeSystem Infrastructure and Application Solutions
- Appliance Servers
- Alpha Servers
- BackOffice Products
- Internet Products
- HPE 9000 and HPE e3000 Servers
- Networking
- Netservers
- Secure OS Software for Linux
- Server Management (Insight Manager 7)
- Windows Server 2003
- Operating System - Tru64 Unix
- ProLiant Deployment and Provisioning
- Linux-Based Community / Regional
- Microsoft System Center Integration
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Blogs
Information
Community
Resources
Community Language
Language
Forums
Blogs
Go to solution
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
09-21-2001 03:53 PM
09-21-2001 03:53 PM
I noted in IBM/Informix release notes that LVM raw devices are significantly slower than REAL raw devices.
Ok, how do I determine which is which? The db was set up before I arrived with 8 raw files/dev. Since then I have created two additinal raw dev. Once using the command line and once using sam.
tia
Doug
Ok, how do I determine which is which? The db was set up before I arrived with 8 raw files/dev. Since then I have created two additinal raw dev. Once using the command line and once using sam.
tia
Doug
Solved! Go to Solution.
3 REPLIES 3
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
09-21-2001 03:57 PM
09-21-2001 03:57 PM
Solution
I wouldn't think that there should be much difference between LVM and non-LVM, but that's just me.
Anyway, to answer your question: If the device file you are using looks something like /dev/vg02/rlvol1 or /dev/vg02/lvol1 then you are using LVM devices. If you were going directly to a disk drive the device file would be /dev/dsk/c1t2d0 or /dev/rdsk/c1t2d0.
LVM will give you a lot more flexibility if you have small disks and need a large amount of space for the DB.
Anyway, to answer your question: If the device file you are using looks something like /dev/vg02/rlvol1 or /dev/vg02/lvol1 then you are using LVM devices. If you were going directly to a disk drive the device file would be /dev/dsk/c1t2d0 or /dev/rdsk/c1t2d0.
LVM will give you a lot more flexibility if you have small disks and need a large amount of space for the DB.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
09-21-2001 05:41 PM
09-21-2001 05:41 PM
Re: LVM raw devices vs real raw devices
Hi Doug:
The actual overhead in the LVM is quite small. A write to say block 5609 of Character Major device 64 Minor device 0x020001(LVM /dev/vg02/rlvol1) simply does a table lookup to determine that that is really block 346 of Character Major 188 Minor 0x02a000. It is merely one lookup which is many, many times faster than the disk i/o itself. Moreover, because LVM raw i/o allows you to stripe multiblock i/o is typically faster. I'd stick with LVM it is generally faster and much more flexible.
Now having said that, I would also look at two other ways. 1) Try conventional cooked i/o; I've found that especially in 11x the performance is essentially identical to raw and surprisingly sometimes better. 2) Use the OnlineJFS mount options convosync=direct,mincache=direct,nodatainlog,datainlog. The mincache and convosync options bypass the buffer cache and you thus get all the benefits of raw LVM i/o with the convenience of files. It's also very easy to test: simply mount without those =direct options and restart the database.
The main thing I want to convey is to believe nothing and measure for yourself.
Regards, Clay
The actual overhead in the LVM is quite small. A write to say block 5609 of Character Major device 64 Minor device 0x020001(LVM /dev/vg02/rlvol1) simply does a table lookup to determine that that is really block 346 of Character Major 188 Minor 0x02a000. It is merely one lookup which is many, many times faster than the disk i/o itself. Moreover, because LVM raw i/o allows you to stripe multiblock i/o is typically faster. I'd stick with LVM it is generally faster and much more flexible.
Now having said that, I would also look at two other ways. 1) Try conventional cooked i/o; I've found that especially in 11x the performance is essentially identical to raw and surprisingly sometimes better. 2) Use the OnlineJFS mount options convosync=direct,mincache=direct,nodatainlog,datainlog. The mincache and convosync options bypass the buffer cache and you thus get all the benefits of raw LVM i/o with the convenience of files. It's also very easy to test: simply mount without those =direct options and restart the database.
The main thing I want to convey is to believe nothing and measure for yourself.
Regards, Clay
If it ain't broke, I can fix that.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
09-21-2001 06:01 PM
09-21-2001 06:01 PM
Re: LVM raw devices vs real raw devices
I think the mount options Clay means are: mincache=direct,convosync=direct,nodatainlog,delaylog
Clay had 'nodatainlog,datainlog' which you really don't want to do.
Clay had 'nodatainlog,datainlog' which you really don't want to do.
The opinions expressed above are the personal opinions of the authors, not of Hewlett Packard Enterprise. By using this site, you accept the Terms of Use and Rules of Participation.
Company
Events and news
Customer resources
© Copyright 2025 Hewlett Packard Enterprise Development LP