Hi
IHAC who has deployed Navision for Retail stores. The deployment architecture is described below:
Intel Blade with 1 Xeon E5320 Quad Core 1.86 GHz
16 GB memory
4x 146GB SAS Drives
The App sits on the C Drive of the Blade
The DB sits on an external storage connected via FC HBA
The external storage has a 4Gbps backplane. Uses 6 x 500 GB SATA Disks @ 7200 RPM in RAID 1+0 Configuration
There are 2 RAID controllers both being used for the volume
The deployment is facing huge performance issue. We observe the i/o wait queue is very very high (100% consistently)
Any suggestions on this?
Rgds
Mitesh: HiIHAC who has deployed Navision for Retail stores. The deployment architecture is described below:Intel Blade with 1 Xeon E5320 Quad Core 1.86 GHz16 GB memory4x 146GB SAS Drives The App sits on the C Drive of the BladeThe DB sits on an external storage connected via FC HBA The external storage has a 4Gbps backplane. Uses 6 x 500 GB SATA Disks @ 7200 RPM in RAID 1+0 ConfigurationThere are 2 RAID controllers both being used for the volume The deployment is facing huge performance issue. We observe the i/o wait queue is very very high (100% consistently)Any suggestions on this? Rgds
Hi Mitesh,
first welcome to The Dynamics User Group, please if you get time go to the introductions section, and tell us more about you.
Anyway, I know that at first glance, it looks like a lot of information in your post, but to answer you we really need to know more. Firstly quad core makes no sense, since Navision uses only one thread, NAV also uses a max of 1 gig, so the extra ram is wasted. Next the SAN is not clear, since its configured wrong, AND you havent said if its dedicated to NAV or shared.
Before starting, we need to know how many concurrent users, how big is the database, how many trnasactions per day, how many store locations do you have, and more. Please give tis information so we cna help.
Many thanks for the response. Here is some data that you asked...if u need further clarifications, pls let me know
Hi.
90 users and 160GB DB on that server can cause nothing but trouble.
Based on the information You gave You should go for something like this
You'll find the above recommendations and som more if you read the "Dynamics HW Sizing guide v3.pdf"
Minor mistake
It's Nav 4.0 executables with platform update 6.
Also why RAID 0+1 instead of RAID 1+0?
I think this is SQL DB, not Navision DB. Correct?
yes sql. in one of his post he has the .ndf files listed.
Rashed:yes sql. in one of his post he has the .ndf files listed.
Sorry I should have been more clear, I meant the message to Mitesh, to confirm before I moved this tot he SQL forums and changed the subject from "Navision DB" to "SQL DB".
But yes its pretty obvious, so I will just move it.
Moved to SQL forum.
Rashed:... Also why RAID 0+1 instead of RAID 1+0?
I agree.
Lars any reason you prefer 0+1 ?
It's You own choice to run RAID 10/1+0 or 0+1. Microsoft suggests 0+1 in their sizing guide. RAID 10 is more fault tolerant but RAID 0+1 is faster.
The key difference from RAID 1+0 is that RAID 0+1 creates a second striped set to mirror a primary striped set. You could say it's a RAID0 with a mirror of the complete RAID0 set. The array continues to operate with one or more drives failed in the same mirror set, but if two or more drives fail on different sides of the mirroring, the data on the RAID system is lost.
RAID 1+0 is mirrored sets in a striped. The key difference from RAID 0+1 is that RAID 1+0 creates a striped set from a series of mirrored drives. The array can sustain multiple drive losses as long as no two drives lost comprise a single pair of one mirror.
So if You wan't performance but still some redundancy choose 0+1. But if You loose disks on both sides of the mirror You are lost. If You choose 1+0 You can loose more disks and still be up and running. RAID 10 is more redundant but You pay with a bit lower performance.
You can read more here http://www.acnc.com/04_00.html (nice graphics) and here http://en.wikipedia.org/wiki/RAID
Lars
lars westman:It's You own choice to run RAID 10/1+0 or 0+1. Microsoft suggests 0+1 in their sizing guide. RAID 10 is more fault tolerant but RAID 0+1 is faster. The key difference from RAID 1+0 is that RAID 0+1 creates a second striped set to mirror a primary striped set. You could say it's a RAID0 with a mirror of the complete RAID0 set. The array continues to operate with one or more drives failed in the same mirror set, but if two or more drives fail on different sides of the mirroring, the data on the RAID system is lost. RAID 1+0 is mirrored sets in a striped. The key difference from RAID 0+1 is that RAID 1+0 creates a striped set from a series of mirrored drives. The array can sustain multiple drive losses as long as no two drives lost comprise a single pair of one mirror. So if You wan't performance but still some redundancy choose 0+1. But if You loose disks on both sides of the mirror You are lost. If You choose 1+0 You can loose more disks and still be up and running. RAID 10 is more redundant but You pay with a bit lower performance. You can read more here http://www.acnc.com/04_00.html (nice graphics) and here http://en.wikipedia.org/wiki/RAID Rgds Lars
Thanks Lars, that makes very good sense. i have always erred on the side of caution, and thus suggested RAID 10, but of course the likely hood of two drives failing at the same time on separate arrays is so slim that it does make sense to go RAID 0.
0+1: striping mirrored sets
1+0: mirroring striped arrays
The way it was explained to me: 0+1 is faster as well as more fault tolerant. You can theoretically lose one disk in each mirrored set and still run, losing one disk only affects the disk itself, since there is still a mirrored disk available. With 1+0, if you lose one disk, the whole striped array that this disk belongs to becomes useless.
DenSter:0+1: striping mirrored sets 1+0: mirroring striped arrays The way it was explained to me: 0+1 is faster as well as more fault tolerant. You can theoretically lose one disk in each mirrored set and still run, losing one disk only affects the disk itself, since there is still a mirrored disk available. With 1+0, if you lose one disk, the whole striped array that this disk belongs to becomes useless.
actually its the other way around.
If you have 10 drives, then in 0+1 you take a RAID zero of 5 drives , i.e a stripe, and then you take two of them and create a raid one from those two stripes. 1+0 you take 2 drives and RAID 1 them, and then combine the 5 RAID 1s as a stripe.
Really it all comes down to statistics, and its interesting to hear other peoples experience of what is better. In my opinion, the performance advantage of a RAID 0+1 is only in read, since statistically a read only requires a response from the fastest of the stripes, where as a in RAID1+0 you need to wait for the slowest response of each of the pairs. But in write, you still need to write to 10 drives, so theoretically there is no write difference. And Write is what generally kills us, so I go for RAID 10.
As to reliability, both RAID 10 and RAID zero can continue to operate with 5 failed drives out of 10, but in raid 0 this only applies if all the drives are in the same stripe, in RAID 0 it works only if it is one of each pair or RAID 1's fails. Statistically the chance or a second drive fail killing RAID 01 is definitely higher than the chance of a RAID 10 fail. So I would always go for a RAID 10, but thinking more of what Lars says, the odds are so slim, that really its probably better to get that slightly better read performance. Do the math, and work out the odds, but either way the MTBF in both cases is more than our life times, so....
Now if we look at the log file, then our concern is Write, so logically there is no advantage to RAID 01, so the Log we should put on RAID 10.
At this point I need to go an find some of my old Statistics books (which I have no idea if I even have anymore) to try and work out which is best. Ah Mr. Pascal, all your time spent in Casinos was not wasted after all. Just think, without Pascal where we would all be.
I really think it comes down to practical experience and what works best, which is why its great that we can have discussions like this.