Topic: Recommendations on a PCI-E USB 3.0 card

[ghorst352]

I appreciate the response but I decided to go the esata route.  Esata is compatible with Centos 5.8 and the speeds are considerably different USB 2.0.  I spoke to an engineer about this and basically what I go out of it is was that Esata for large file transfers such as the case with huge backups, esata is actually better suited for this task than USB 3.0 but I really can't say that's true or not however there are plenty of whitepapers out their and benchmarks that kind of reflect that.  Anyways I think I am satisfied with going this route, time will tell.

[purvis]

That is possibly your best bet if your not that portable, and you can find some esata portable bays.
I do not know know what happened to the esata market. USB 3 is much faster than USB 2.
But the USB 3 should have had SMART support for all equipment.

[idp_qbn]

This may be wandering a bit off track, but there is some buzz in the CENTOS forums about using eSATA hot-pluggable drives. Perhaps somewhere down this path you may find a solution to faster backups.

Cheers
Ian
(No, I have not tried, I do not know if it is possible)

[ghorst352]

I have done a bit of research on this and have come across benchmark after benchmark after benchmark for eSATA vs USB 3 and it's quite remarkable because for large file transfers eSATA tends to win.  I didn't think coming into this that I would decide on going from USB 3 to eSATA but that's what I have chosen.  You never hear any buzz about eSATA, it's always USB 3 this and that which is understandable since USB is the most widely used interface in the computer world.  However I think eSATA takes USB 3 to lunch on huge file transfers but USB 3 I am sure will go through revisions that will change all of this in the near future.

[purvis]

There are some eSata connections on some mother boards. I really have not read about the differences in esata and sata connections on a motherboard and i have too much work to do. But the original poster many want to see what the scoop is all about.

[ghorst352]

I am pretty satisfied with the performance of eSATA.  It is a day and night difference from USB 2.  If you are interested here are some benchmarks I ran:

/dev/sdc:
 Timing cached reads:   13404 MB in  2.00 seconds = 6716.18 MB/sec
 Timing buffered disk reads:  380 MB in  3.01 seconds = 126.14 MB/sec
 
[root@draftdata ~]# dd if=/dev/zero of=/mnt/backupa/output bs=8k count=10k; rm -f /mnt/backupa/output
10240+0 records in
10240+0 records out
83886080 bytes (84 MB) copied, 0.15761 seconds, 532 MB/s

[Stefano]

try the same test using bs=hd cache and a file bigger (some GB)

[ghorst352]

Thanks Stefano,


Totally missed that I wasn't even testing the drive but rather cache.  Have to test higher than the amount of memory installed.


[root@draftdata ~]# dd if=/dev/zero of=/mnt/backupa/tmp.dat bs=2048k count=50k
51200+0 records in
51200+0 records out
107374182400 bytes (107 GB) copied, 1140.98 seconds, 94.1 MB/s

So in less than 20 minutes it wrote nearly a 100GB to the external eSATA drive.  Not too shabby. 

[piran]

So in less than 20 minutes it wrote nearly a 100GB to the external eSATA drive.  Not too shabby. 

In what RAID config is your SME?
Did you get the chance to observe by how much your
writing to the eSATA increased the HTOP utils factor?

[ghorst352]

Piran,

I am running a Mirror config.
I don't know what "HTOP utils factor" is?  I even googled it and nothing came up.  Plz explain.

Sincerely,

[ghorst352]

I think I found what you where asking...I am assuming you are talking about http://software.opensuse.org/package/htop.  I have never used htop so I will check it out.

[ghorst352]

LOL, I just realized HTOP is installed natively.  Wow, learn something new everyday.

[piran]

Sorry, distracted elsewhere.

The HTOP utils is the "load average: ".
< 1.0 SME ticking over proportionately.
1.0 is everything fully loaded but no delays.
>1.0 SME under load and stuff is being delayed.
>>5 SME taking maybe 5x as long to get things done.
>>10 SME coping but potentially in trouble - EVERYTHING delayed.
...and so it gets worse.

Curious to learn how writing to your eSATA affects SME's load factor
(utilisation) and, in particular, under what level of RAID you've in use.
Here we have an 8TB backup (potentially)...
Currently using a NAS box to a dedicated gigabit NIC.

[ghorst352]

Well I happen to use Nagios to monitor the load of my servers and during the time-frame of the benchmark the load did get high, see below:

Notification Type: PROBLEM

Service: Current Load
Host: DRAFTDATA
Address: 192.168.1.10
State: CRITICAL

Date/Time: Tue Apr 23 10:17:20 EDT 2013

Additional Info:

CRITICAL - load average: 11.82, 8.10, 3.60

[ghorst352]

I have to add that I use rsnapshot for my backups which does not stress the load.  So the dd benchmark is not comparative to using rsync just to make clear.  During the entire backup the load was minimal never high.