Advice about my configuration

[Addrelyn]

I would very much appreciate it if someone used to freenas could tell me if my configuration is compatible with freenas 8 and zfs, or if I'm missing something...;)

Motherboard : M4A78LT-M LE
CPU : AMD Sempron™ 140 1 core 2.70 GHz
RAM : 2x3 GB frequency 1600 MHz Latency: 7-10-10-24

2x1TB in mirror
5x2TB in raid z

Is there enough ram memory and is the cpu powerful enough?

Thx

 

[jfr2006]

More than enough. See my testing i posted on the 4Kb blocks issue thread!

 

[Addrelyn]

Thx
In fact I finally took the same configuration as you :


Motherboard : M4A78LT-M
CPU : AMD 245E
RAM : 2x4 GB frequency 1600 MHz (and I can add 8 more if needed)

2x1TB in mirror
6x2TB in raid z (Spinpoint F4 Ecogreen S-ATA - 2 To - 32 Mo)

 

[Tekkie]

Looks like I am not the only one charmed by the M4A78LT-M board. :)

Although I did go with a 415e as I couldn't find a 245e, never hurts to have a core more. ;)

The 2TB and 4KB discussion however has me worried, looks like FreeNAS8 doesn't have proper support for 4KB drives just yet, but I could be mistaken. http://forums.freenas.org/showthread.php?66-4kb-Disks-and-FreeNAS

 

[Thingie]

I would suggest to use disks from different vendors for the raid. If you order 6x2tb disks at once the chances are great they were produced around the same time and will fail around the same time. I have had it happen once were two disks I ordered in one order broke down within 3 days of each other. Just a suggestion.

Grtz

Thingie

 

[matthewowen01]

I would suggest to use disks from different vendors for the raid. If you order 6x2tb disks at once the chances are great they were produced around the same time and will fail around the same time. I have had it happen once were two disks I ordered in one order broke down within 3 days of each other. Just a suggestion.

Grtz

Thingie

I second this, kinda. i would buy the same make and model, just order them from different resellers. or order some, then wait a few days then order more.

 

[Thingie]

I second this, kinda. i would buy the same make and model, just order them from different resellers. or order some, then wait a few days then order more.

That's also a good way to do it, probably better for performance.

 

[ohnename]

In a RAID array the HDDs should be exactly the same else you will get get slow access times.
If you really want to use different drives then make 2 arrays with 3 drives until one array with 6 drives.

 

[Tekkie]

In a RAID array the HDDs should be exactly the same else you will get get slow access times.
If you really want to use different drives then make 2 arrays with 3 drives until one array with 6 drives.

I believe the array will be as fast as the slowest drive in the pack, having drives from 2+ vendors is not going to make the array any slower than it would be was it made with the slowest vendor solely.

 

[joeschmuck]

The 2TB and 4KB discussion however has me worried, looks like FreeNAS8 doesn't have proper support for 4KB drives just yet, but I could be mistaken. http://forums.freenas.org/showthread.php?66-4kb-Disks-and-FreeNAS

Looks like in your previous posting you said you were going for the F4 drives. These are emulation drives so unless Samsung comes out with a new firmware to disable 512 byte emulation, the 4K thing really isn't a factor and this has nothing to do with FreeNAS, it's the drive. Also just because I haven't seen it posted here, there is a firmware update for these drives. I forget the date on the drive you should look for but I think it's NOV 2010, if manufactured after that date you don't need the update, one or before that date you do. Checkout the Samsung site support for that drive.

 

[ohnename]

I believe the array will be as fast as the slowest drive in the pack, having drives from 2+ vendors is not going to make the array any slower than it would be was it made with the slowest vendor solely.

The access times (in ms) will be bad with different drives in one array not the performance (in MB/s). Thats the reason why its better to use the same drives in a array.

 

[Tekkie]

The access times (in ms) will be bad with different drives in one array not the performance (in MB/s). Thats the reason why its better to use the same drives in a array.

Why exactly will a mix of drives affect access times?

 

[Bohs Hansen]

i think 2 things got mixed up here. "2 different brands and models" and "2 different shipments of the same brand and type".

Its about getting same brand and model of drive, but from different production sets. This decreases the odds of all drives failing at the same time.

 

[btechnet]

Why exactly will a mix of drives affect access times?

Each drive / manufacturer is going to have differently rotational latency along with mechanical latency.

This causes each different drive to have to buffer more or less depending on that latency. (access time)

It rather minimal between drives that are 7200RPM's.

But drives that are a mix of 5400 and 7200 or 10000RPM's may have a very large difference.

SSD's are even faster that 10000RPM drives and can cause a very exaggerated difference in access time.

But for the most part if you stick with only 7200's or only 5400's or only 10000's then your pretty safe.

 

[esamett]

the new beta release 8.0.1 is said to have a 4k checkbox.

+1 on disk purchase advice. I bought same model drives from same vendor over about a month. one shipment of two drives was doa on both. all others, including the replacements for the doas are fine so far. it is intuitively reasonable to order 2 disks at a time for a zfs/z2 array. the system should have adequate redundancy should one batch go bad at once.

 

[joeschmuck]

I would hope the original poster would buy similar type drives (same rotational speed, capacity, buffer size) however when running a home system (poster didn't specify if this was for home or work), having these differences I would not expect to generate a problem for a home or small office. If this was for work I would expect enterprise drives to be purchased for a work environment not the F4's.

Of course I could be really wrong but that is my opinion, not fact.

 

[Tekkie]

Mixed vendor array results

Each drive / manufacturer is going to have differently rotational latency along with mechanical latency.

This causes each different drive to have to buffer more or less depending on that latency. (access time)

It rather minimal between drives that are 7200RPM's.

But drives that are a mix of 5400 and 7200 or 10000RPM's may have a very large difference.

SSD's are even faster that 10000RPM drives and can cause a very exaggerated difference in access time.

But for the most part if you stick with only 7200's or only 5400's or only 10000's then your pretty safe.

Here are my initial performance numbers from running a mix of 3 vendors in a RAIDZ2 array:

Code:

%dd if=/dev/zero of=/mnt/storage/movies/ddtest bs=1024k count=20000
20000+0 records in
20000+0 records out
20971520000 bytes transferred in 89.074932 secs (235436834 bytes/sec)
%dd of=/dev/zero if=/mnt/storage/movies/ddtest bs=1024k count=20000
20000+0 records in
20000+0 records out
20971520000 bytes transferred in 50.796097 secs (412856917 bytes/sec)

Call me crazy but I don't see any performance problems due to 5400 and 5900 spin drives.:):):)

 

[matthewowen01]

Call me crazy but I don't see any performance problems due to 5400 and 5900 spin drives.:):):)

The drive speed only effects the seek time, how long before the spot that needs to be read can get under the read head. Faster drive means less seek time. If you were too read from random places on the drive, you would see a performance hit vs a 7200 rpm or 10k rpm drive. But the large amount of cache really soothes that out. (all of my drives are 5400 rpm or 5900 rpm, not ragging on them)