Partition Alignment tool

[metamorphosis]

Sent. The paragon system is pretty picky about which emails it'll actually send something through to from my experience. Check your PM's.

[DevonB]

Wow, what a bad white paper. It skips important reasons about the misalignment and SAN/RAID (RAID 5 does 4 IOPS per write, which makes the alignment problem worse with it over RAID 1, which is 2 IOPS), it doens't touch on the drive cache and system cache polution problem with misalignment at all because it has to read in more data than it really needed, and with a SAN with having multi-hundreds if not thousnads of drives defined, that problem compounds quickly, it talks about 'pre-Vista', but doens't even tell what Vista/W7 do, it 'claims' to improve desktop performance, but goes over nothing about it and why it thinks it improves it... wow. Not impressed. That's by far the least amount of info I've seen in an alignment paper that I've read to date.

Thanks Susiwong and Meta for sending me links so I could read the paper.

Devon

[susiwong]

Now Devon,
as I'm only dealing with small peer to peer networks (< 5 PCs) and no RAID either (wouldn't touch onboard or software RAID anyway), please comment on the following statement :

In that scenario the only reason for using partition alignment would be one of the new 4k drives, which would suffer from inferior write performance otherwise.
Vista and 7 align new partitions correctly anyway, only XP users need to use one of those tools to align.

Correct ? Wrong ?
At least that's what I read from varying, less than clear sources.

Thanks,
susiwong

[Throbert]

Paragon hosed my laptop but I only think it screwed up my MBR or boot.ini. The only problem was I didn't remember having an admin PW but when I booted my XP disk to recover it asked and I couldn't get in. I pulled the HD and looked at it on another PC and couldn't find any thing wrong or any corruption.
Definitely, you gotta back up or clone before screwing with this stuff.
throb

[susiwong]

Paragon hosed my laptop but I only think it screwed up my MBR or boot.ini. The only problem was I didn't remember having an admin PW but when I booted my XP disk to recover it asked and I couldn't get in. I pulled the HD and looked at it on another PC and couldn't find any thing wrong or any corruption.
Definitely, you gotta back up or clone before screwing with this stuff.
throb

Three typical pitfalls:
- accidentally hidden boot partition
- inactive boot partition
- changed boot.ini
All this can easily happen with partitioning / imaging tools.
Not necessarily always your own fault.
Can't tell you how often I've been there with my multiboot systems.
A few different boot CDs and a Windows CD and the problems can easily be solved.
fixmbr and fixboot from the recovery console, something like Acronis Disk Director (nice: you can edit files like boot.ini even on NTFS disks from this boot CD), maybe a WinPE ...
It helps when you can be sure these disks actually support your hardware because you verified this before.
The simple DOS /IDE /PS2 days won't come back.
Ymmv,
susiwong

[susiwong]

[susiwong]

[susiwong]

[DevonB]

Now Devon,
as I'm only dealing with small peer to peer networks (< 5 PCs) and no RAID either (wouldn't touch onboard or software RAID anyway), please comment on the following statement :

In that scenario the only reason for using partition alignment would be one of the new 4k drives, which would suffer from inferior write performance otherwise.
Vista and 7 align new partitions correctly anyway, only XP users need to use one of those tools to align.

Correct ? Wrong ?
At least that's what I read from varying, less than clear sources.

Thanks,
susiwong

I'm saying you don't need it, period, on a desktop. A typical load doesn't generate enough IO to make a difference. Where this DOES make a difference is in RAID 5 configurations with heavy small random writes. RAID 5 takes 4 drive IOPS to satisfy one host side write IOP. Take all those writes in a heavy multi-user enviornment, the extra overhead from having to load in those 31.5k blocks that aren't needed from a drive misalignment adds up. So like a moderately active environment of say 10k to 20k IOPS, each with an extra 31.k5 is 300 megabytes to 600 megabytes extra per second of host side data that really didn't need to be written to the drive or be poluting controller cache. If this was 50% read and 50% write, that would be 600 to 1200 megabytes of writes and 150 to 300 megabytes of reads for back-end operations. Ouch. Unneccesary cache polution will push good blocks out of cache that should have lived in cache longer, which means more cache read misses, which means more read IOPS... see the vicious cycle?

Absolute best case scenario on a new drive with opimal conditions (heavy seqeuncial read) can do over 1000 IOPS. But a seqeuncial read in that case would be one extra 31.5k block at the beginning and one at the end. Random IO would be less IOPS, and less transfers. Seeing what I'm getting at a little better now of why I'm saying it's not a big deal for an average desktop to have the mis-alignment?

Devon

[susiwong]

Hi Devon,
thanks a lot !
This description is a lot clearer than that paper for sure.
The few systems I have to care about are your typical small studios or similar, 2 to 5 computers, a few network shares, some mapped drives, lots of external backup drives, partially internet via router, single / dual user - child's stuff, really, from an admin's standpoint.
Performance and stability of a DAW is #1 priority, but so far we haven't even remotely reached a bottleneck with HD streaming from single consumer SATA drives, one each for samples and audio, one for OS.
I might start with upgrading my OS drive to a small SSD (Postville or similar) in fall, the others will probably follow.
That's about as "hardcore" as it'll get in the foreseeable future.
Thanks again,
susiwong

[metamorphosis]

Still useful tool for Raid and SSD systems - desktop or otherwise...
and particularly for OS partitions.
Streaming 50-100 24/96 tracks in unison (happens in my studio anyway)
uses a lot of bandwidth. It reaches the IO limit.
I don't think the white paper is bad, it's targeted at average technically competent users, not system tech workers.
Not really a white paper though, more a product guide ;P

[DevonB]

Still useful tool for Raid and SSD systems - desktop or otherwise...

Only if it was formatted with an OS that has the issue in the first place. Again, if the load isn't significant, the problem really isn't much of a problem, even if you never did the alignment the first time. For a desktop, this is purely a 'geek tweak'. It REALLY takes a significant load to really see the problem clearly and have an impact.

Streaming 50-100 24/96 tracks in unison (happens in my studio anyway)
uses a lot of bandwidth. It reaches the IO limit.

Of which, you're describing best case scenario, which is heavy reads, lots of sequencial activity and pre-fetch advantages. Fixing the misalignment issue is going to be next to nil in performance help if you happened to be misaligned in this scenario. It's writes in a RAID 5 environment that start to kill you. That's coming from the disk vendors.

During your heavy streaming, with PermMon, what is Avg. Disk Queue Length, Avg. Disk Sec/Transfer, Current Disk Queue length and Disk Transfers/sec under Physical Disk counter?

I don't think the white paper is bad, it's targeted at average technically competent users, not system tech workers.
Not really a white paper though, more a product guide ;P

Yes, it's a purely a sales document with a bit of technical puffery to sound good. The better white papers on the subject are one's that they aren't trying to sell you something, not to mention, there are tools that are free that do this procedure.

Devon

[metamorphosis]

No, I'm sick of this argument. Most partitioning apps create misaligned partitioned with SSDs - so it's not just the OS's. You seem to be putting up one hell of a fight for a free application that has nothing to do with your line of work (apparently), so I don't know quite why you're so invested in it. Pointless.
Whatever. It works for me, has had a positive impact, particularly using SSD and I'm happy with it.
If it's 1% difference, well, 1% difference then. Several people here would disagree with you if you read back - I don't think it's worth spending time on-
See ya,
m@

[DevonB]

No, I'm sick of this argument.

I'm not arguing, I'm correcting the mis-information. I'm doing exactly what you did at the beginning of this thread with susiwong; stating that the information presented didn't apply to the situation at hand.

You seem to be putting up one hell of a fight for a free application that has nothing to do with your line of work (apparently), so I don't know quite why you're so invested in it. Pointless.

Yes, the tool has everything to do with my line of work, I just wouldn't use it. I'd use mbralign from NetApp or vOptimizer from VizionCore. Not sure where you got that from though. And no, pointless is discussing that synth A is fatter than synth b; that's just an opinion. We're discussing factual information. Considering a lot of other people read these threads, and then look them up via Google later for years to come, technically accurate information is more important.

Whatever. It works for me, has had a positive impact, particularly using SSD and I'm happy with it.

Sure, that's SSD misaligned under XP. That hasn't been the whole of the conversation either. Like it or not (I know I don't), XP is a dead OS now.

If it's 1% difference, well, 1% difference then.

For a normal drive with platters, yes, I'd agree, it's minimal.

Several people here would disagree with you if you read back

They did, and I gave my technical explanation of what I believed the issue was. I still don't believe it had anything to do with the tool, it was the drive.

- I don't think it's worth spending time on-

But you spent the time responding back yet again. I've tried to make the thread as on topic and technically relevant as possible. Storage is my field of expertise, and I try to help as many people on here as I can with it. If you disagree with me, that's fine; I certainly don't know everything there is to know, and the field is constantly changing at a very rapid pace. I've presented my case and tried to present a factual-based discussion asking for very specific metrics. You've just argued and I haven't seen much in the way of technical discussion to back up your arguements. If you believe I'm wrong, then present your case with facts that that can be backed up from documentation or from experience. But you're done discussing... just like you were a few pages back. That's fine too. I still wouldn't mind getting this back on topic and getting answers to some of the questions I asked previously either. Whichever way you'd like to take it, it's in your court.

Devon[/quote]

[metamorphosis]

No, I'm sick of this argument.

I don't think it's worth spending time on-
See ya,
m@

No, I made myself pretty clear. Good luck.
M.



and-

[EDIT: Also for those of you actually wanting to learn and not just swing your e-penis around-
stumbled across this article today while trying to figure out whether the hard drives I'm buying are "advanced format" (4k) or not:
LINK

It shows dramatically the performance difference between aligned and unaligned partitions on 4k-sector drives under XP (the worlds most popular OS), in everyday tasks. Be warned-
Cheers. M@]