ML150 Slow Copying between HDDs

[Benedek Frank]

Hi

This happens on an ML150 I just recently installed. It has 4 HDDs, two of them are RAID1 for mirroring, and house the OP system. The other two are 500GB HDDs (Drive A and B later on) that are on their own, and used only for storing backups for short term periods.

This is what happens. I have my whole Network of Servers doing backup to Drive A, and on Monday, I transfer the backup to an external Drive, which then goes to a bank safe for outside storage. When its all transferred (200GB approx.), I copy it to Drive B, then delete it from Drive A. The problem is the following.

Transferring the 200GB data From Drive A to External Drive takes little less than 2 hours.

Transferring the 200GB data From Drive A to Drive B, it takes like 7-9 hours.

Drive A and Drive B are both in the HotSwap ports, and they are controlled by the built in HDD Sata raid controller as single drives.

What can I do to increase performance? Copying to the External to Drive A and then from external to Drive B takes half the time than copying directly from Drive A to Drive B. This is puzzling, at least to me.

Thanks for any pointers,

Ben

[Rafael Pedrosa]

I have the same problem. For me is even worse because this is a production server and when copying files the entire server slows down tremendously. Is really affecting my system, this server is the upgrade, the old server works a lot better.

[StephenRJr]

Ben,

We are having the same issue with 2 of our 3 ML 150 G3 units. Heavy read activity seems to work fine; heavy writes seemingly lock the disk subsystem for 3-4x the expected duration.

I have configured an 80GB drive on the first channel (simple volume), and 2 x 250GB on channels 2 & 3 (RAID 1 Mirror). My system is the non-HotPlug variety and is configured with 8GB of RAM running Windows Server 2003.

When the system hangs, the drive activity LED lights up solid. Other inputs from keyboard and mouse work fine - but all applications waiting on disk activity freeze.

I ran a simple test - perform a CHKDSK on the RAID set. This produced the condition as well.

No answers - just more information.

[PussPower]

Enable the write cache in the RAID BIOS.
I have 2 ML150G3 myself with such problem. Without write cache enabled this powerful servers seems as my PDA in disk operations. The HP understanding about disk systems is really unpredictable. They think, that disabling the internal drive's cache is a good idea. Consistency ? How about flow syncronisation with spindle rotation ?
Have the same problem with ML350G5 with E200/128 and SATA drives. In my situation at this moment this not solvable, because on E200 there is no new firmware/sofware, which can enable drives cache on SATA.

[chongkan]

Update the firmware on the HDDs and overall system if posible.. write cache could be enable on both drives ( through OS )

If the performace is not increased then try the following:

1- Copy A on the tape drive ( 2hrs )
2- Copy A from the tape drive to B and measure time. ( haven't tried but I guess it could be a little longer than 2hrs due to writening operations instead of reading.. )
3- Deliver Tape drive.

Time expense 2 + the time to copy A from the tape to B

That would be the workaround... lol..

[kong San]

Hi All,

I have the exact same problem with a production server.

It is a Proliant ML350 G5 with 2 x 160GB SATA drives on a smart array e200i controller configured with RAID 1 + 0 with standard 64MB cache. Cache Ratio is default to 100% Read, 0% write.

The configurations above are all default settings and can not be changed.

I have updated all drivers and firmware however nothing helps.

Problem is identified to be writing to disk. When writing to any of the RAID disk partitions the disk queues and page files max out to a 100% within perfmon and all network application freezes.

I have tested by copying a 1GB file from the local drive to another location on the local drive. When doing so server disk queues and paging files max to 100% till the file transfer completes and during that time all network activity freezes.

However reading from the drives is ok. Copied files from Drive D to external HDD (F). Disk queues and page files only reach 20% max.

I have spoken to HP technical services and they have no idea. I always thought RAID 1 + 0 will only work with a minimum of 4 HDD in the array? HP is saying other wise. They believe that it will work as a mirror if you only have two HDD installed. Can someone verify that it is the case. I have forumed this and all experts are challenging this statement.

I have tried to add additional HDD to the array to see if it solves the problem however because of the hardware limitations of the Smart Array E200I i am unable to. HP advises to purchase the battery back upgrade to 128MB cache to enable the following features.

1) Adding HDD to the array.
2) Changing the cache ratio from 100% read, 0% write to 50% read, 50% write.

Which i believe would both help with performance. What i understand with mirroring RAID 1 does have allot of overhead writing for it has to write to two disks at the same time.

Having 2 more disk added to the array will enable stripping which generally is there for performance. When writing you are able to access 2 HDD therefore eliminating bottleneck effect of writing to one HDD. This will decrease disk queue request.

Secondly having cache enabled on the writing component will allow for a slight increase in performance writing to the disk. This will decrease disk paging file request.

Please if anyone else have anymore thoughts on this please post.

Thanks in advice for any assistance.

Kong.

[McM_1]

Hi All,

I am having the same problem. My ML350 G5 freezes when i copy one file greater than 1 GB, i have 3 SATA HDs with raid0 but each with your raid (i dont know how to explain).

[PC Doc]

Yep. I can confirm I am having the same problem.

I have 4 X 500 Gb SATA2 drives in RAID 0+1 on the E200. Transfers within the drive are very slow (under 10 MB/s), transfers to ESATA drive around 70 MB/s, transfers accross Gigabit LAN to another server (not running E200) 40-60 MB/s and transfers from other server to HP with E200 are between 10-18 MB/s.

We purchased the ML150 G5 to be a VM server, but are forced to use the cheaper, no brand backup server because it is FASTER. It has less ram, cheap RAID card, same HDD configuration, and normal P4 quad core. The ML150 has a Xeon Quad Core.

If we can't get this resolved soon, we are going to be selling the ML150 (if we can) and get a Dell, or just build one from scratch with components that work.

The reasoning behind buying HP for the VM server was that HP are supposed to be good at providing robust, reliable servers.

Any server with an E200 IS NOT ROBUST OR RELIABLE.

FIX THE PROBLEM HP!!!

[TarunJain]

Hi All,
The performance would be significanty increased significantly after installing 128 MB BBWC

[TarunJain]

Sorry for the incomplete post.
It's a pretty old & long thread now.
Let me try to summarize it.

This thread was created for performance issue of ML150 server & the author never mentioned the server generation, Controller's name, Hard Drive (whether SAS or SATA).

However as per description we assumes/know following things:
1. Server is ML150G3
2. HDDs are SATA
3. Controller should be integrated SATA Raid controller (As hp was promoting the ML150 G3 as the most cost effective server at that time & wasnâ t pushing for optional controllers)

PussPower provided the solution to improve the performance of this configuration & also shared his bad experience about E200 controller

The Author (Ben) never confirmed whether the issue resolved or not, in fact this thread was posted on his last contribution date.

After that this thread become Performance issue with E200/E200i controller
E200 is an entry level controller only. Its true performance can only be seen with BBWC.
Therefore it should be installed in every server. HP should include this 100$ thing to save its billion $ image.

Hope that information would be helpful.
& as the original Author is inactive, I propose that any new query should be placed separately. At least I wonâ t scratch this thread now.

Tarun
-----------------------

Specifications of E200/E200i Controller: http://h18004.www1.hp.com/products/quickspecs/12460_div/12460_div.html
ProLiant RAID controllers page with optional upgrade:

http://h18006.www1.hp.com/products/servers/proliantstorage/arraycontrollers/index.html