Very slow OS drive RAID1

[tjader]

 Hi,

In my Z800, I have two mirrored drives C: (OS) and S: (DATA). The performance on S is fine but terrible on C: .

Below I have attached a number of screenshots which might help.

What can possibly be causing this?

Thanks,

Lars

Here are the test results from a benchmark I ran.

[BambiBoomZ]

Lars,

A RAID 1 is making an exact copy of the source drive in real time to protect the contents in the event of a hard drive failure.  that process would be very slow of a C: drive there is a delay in reading the source and writing to the mirror.  For a C: drive, it's necessary to combine the RAID 1 with a RAID 0 so as to divide the read / write of the source (C:) over two drives and then the mirror drives  can copy it at much higher speed.  But, all those drives- 4 or 5 will only have the capacity of one of the drives.

For a z800, consider storing a system image on a spare drive in a USB enclosure.  If the the C: drive fails or is infected with a virus, the system image can quickly restore the exact C: drive with all programs, settings, bookmarks-everything intact. The bonus is that the system restore will be finished in 1/10 time- or less than rebuilding a RAID 1.  The elderly office server has a RAID 1 on 146GB 15K SAS drives and it takes about 14 hours to rebuild.  A system image could restore that size drive in perhaps one or one and a half hours.

The z800 is running an SATAII 3GB/s disk system- and that can be significantly improved. Consider adding a PCIe RAID controller, for example an LSI 9240, which will convert the disk substystem to SATAIII 6GB/s. A PERC H310 controller was added to a Dell Precision T5500  (similar to HP z600) and without any other changes the H310 changed the Passmark disk score from 1970 to 2649.

Install an SSD large enough to contain the C: Drive and a partition for active files. The z420 I'm using at the moment is using an Intel 730 480GB.  The C: drive is 152GB,  but  a 250GB drive would still work as the active files partition contains only 54GB.  Then configure a RAID 1 of mech'l drives for libraries, data storage, archive & etc. 

A safer configuration is to have the SSD OS/active files drive and single storage drive. Then a drive equal to the storage drives is in a USB 3 enclosure and that drive is run only when backing up.  That isolates the backup from system failures and malware viruses.  Becase the drive is run so little, there are seven and eight-year-old drives in perfect health.

Cheers,

BambiBoom

[tjader]

Hi BambiBoom,

Great proposals and I am thinking about having two SSD's. However, I would also like to understand why is the C drive so slow vs the S drive? They should be almost equal.

Could it be a driver issue and also why is the C drive not flagged as an OS drive and bootable?

Thanks,

Lars

[BambiBoomZ]

Lars,

That's an interesting situation. or rather two situations. 

Firstly, I think the performance deficit has to be the result of the inherent RAID 1 parity check latency. 

But , it's an independent problem if the RAID of C is not recognized as the boot drive.  Check the boot sequence in BIOS and see where the RAID 1 occurs. If the RAID 1 was configured after installing Windows that could be a problem as the RAID may either not be recognized or is possibly running off it's own BIOS and the OS may have to installed in effect to the RAID controller.  I've never used a software RAID so I'm not sure.  But, if a system acquires a hardware RAID controller, the controller has to be recognized and Windows would need to be installed with the controller listed in proper order in the boot sequence.  After checking the boot sequence and the RAID configuration setup, I'd suggest reviewing the procedure for setting up a RAID in the z800 manual.

Cheers,

BambiBoom

[tjader]

I have run the HP Performance Advisor software (excellent tool) and I can now see that all disks are set to NEGOTIATED SPEED 1.5 GBPS. Why, it should be 3 GBPS???

Below is information

Device Slot Number: 0
Friendly Name: OS
Instance Id - Hardware Key: IDE\DISKOS1.0.00__\4&23DDAE51&0&0.0.0
Device Enumerator: IDE
Device Service Name: disk
Size: 298.09 GB
Physical Disk Instance Name: 0 C:
Number of Partitions: 2
Status: OK
Media Type: Fixed hard disk media (self-encrypting drive)
C: Bytes Per Logical Sector: 512
C: Bytes Per Physical Sector: 512
C: Bytes Offset for Sector Alignment: 0
Device Negotiated Link Speed: GBPS_1_5
Device Maximum Supported Speed: GBPS_1_5

[BambiBoomZ]

tjader,

Ah, this is a bit complicated and will take further research.

The story so far:

1.   It appears that the Port Link settings are mismatched  beacause the EX-series Switch is configured incorrectly.

2.   As a result of  that incorrent configuration, the  Link Partner's link mode is designated UNKNOWN so the EX link mode is defaulted to HALF Duplex. This means the drives' throughput is running on only half the PCIe lanes- 1.5GB/s instead of 3.0GB/s.

3.  If the Link Mode of partner is showing UNKNOWN try:

__ A. setting the link mode on EX to full-duplex manually or:

__ B. disable auto-negotiation and set the speed and link mode manually to match with remote side.

It may be possible to solve this situation by removing the current RAID configuration, resetting all system configurations to default, and reconfiguring.

However, my tendency at this level of technicality is to consider redirecting the effort of restoring the SATAII speed instead to improving the disk subsystem to SATAIII.  There is a learning curve - and cost- also in installing and configuring a 6GB/s controller and SSD, but the long term performance benefit would be significant.  Consider using a much faster system restore image of the OS and programs instead of a RAID 1 and incremental or differential backup to an external drive to protect data.

Cheers,

BambiBoomZ

[tjader]

Hi BambiBoomZ,

Unfortunately, I don't have the knowledge required to understand your answer.

Where do I set the link mode and what is EX?

Yesterday I installed the Intel RAID Web Controller and it says the firmware on my controller is version 01.25.03.00 and I should upgrade to 06.22.00.00

As I have not seen this recommendation before I am interested in hearing your thoughts. Is it safe to upgrade the firmware and have anyone done this earlier?

Thanks,

Lars

[BambiBoomZ]

Lars,

I am also going outside my experience and can only make general suggestions. I've never had to reset the link mode manually and unfortunately, without seeing the sequence of screens and options, I don't know how to list the steps. 

First though:

If the firmware updated software is the correct one for the z800's controller, then yes, that would be a good start.  The link mode is supposed to have an automatic negotiation mode and the updated firmware may get that to work.

If you're not certain about their status, update the BIOS, chipset driver, and any drivers related to storage including the Intel Rapid Storage Technology Utility and Driver.  See:

https://support.hp.com/us-en/drivers/selfservice/hp-z800-workstation/3718645

Earlier, you pasted the current configuration that inlcuded:

Device Negotiated Link Speed: GBPS_1_5
Device Maximum Supported Speed: GBPS_1_5

So, the exercise is to use the configuation utility to reset those values.

If you can find a setup screen that looks something like this:

Main menu, select an option: [1-99 or e or p or 0 to quit] 13

Link topology: [0=Auto, 1=NL_Port, 2=N_Port, default is 0]

Link speed: [0=Auto, 1=1Gb, 2=2Gb, 4=4Gb, default is 0]

FCP Initiator protocol: [0=Disabled, 1=Enabled, default is 1]

FCP Target protocol: [0=Disabled, 1=Enabled, default is 1]

LAN protocol: [0=Disabled, 1=Enabled, default is 1]

Assignment of Bus/Target IDs: [0=SortByWWN, 1=SortByDID, default is 0]

Immediate Error Reply: [0=Disabled, 1=Enabled, default is 0]

Maintain Logins: [0=Disabled, 1=Enabled, default is 0]

Hard AL_PA: [01 to EF, or FF for Soft AL_PA, default is ff]

Initiator Device Timeout: [0 to 2047, default is 60]

Initiator I/O Pending Timeout: [0 to 127, default is 8]

It will then be possible to see the current settings. It show your Link Speed set as 1.5. and I suggest setting it to Auto.

The above menu is not for the z800 utility.  But the equivalent highlighted settings might solve it.  Disregard all the other settings and values.

I am a bit timid to recommend actions concerning RAID.  Changing RAID modes and certain reconfigurations can delete all the data on the affected volume.  Before changing the consifuration on an existing RAID, be certain to have a complete backup of the affected drives.

Cheers,

BambiBoomZ