The Defrag test to end all arguments!

So today I was with a customer having one of those disk Defrag arguments; you know the one where he insists that running defrag will make his crap, slow system fast and reliable.... I told him no....

And then more people joined the argument: "You're using the wrong app, try X and you will be surprised!" they insisted. "Why would they make defrag apps if it did nothing?" they queried (I once saw a memory defrag utility!). "It depends on many things" they muttered.

So, for my own peace of mind, and to end the damn argument once and for all, I am going to test it myself. Every time I am faced with a performance problem I get asked to defrag; and then the argument starts - I really need to know the truth!

I want to test the performance of a heavily fragmented HDD versus a non-fragmented HDD; I do not want to test various products against one another. So what would be an accurate test of tests? Some questions:

1) Which defrag product should I use? Is it safe to assume that a disk with 0% fragmentation is non-fragmented. Should I run several programs just to make sure?
2) What tests do I run? Startup/shutdown, throughput, disk read/write, access. Any other benchmarks/tests?
3) Anything else that will make this test accurate?

Thanks!

It's logical that it will boost performance on HEAVILY fragmented drives, but between a 0% vs 10-20% fragmentation there won't be much difference.

its all those little applications thats making their pc's slow
registery monitor, 3rd party defrag software, spyware cleaners, bulky antivirus, google desktop, every piece of software that claims to make your pc faster, and every piece of hardware loading its own software into startup(i`m looking at you iburst terminal) etc.
Best way to reduce boot time and free some memory is to disable startup items in msconfig that you don't use, and don't install bulls.... software.
Defrag is a waste of time imo

Look at startup items and services running you don't need. Cleaning the registry out while you are at it wont do any harm either.

Yeah your HDD has to be very old and unmonitored for a defrag to make a difference. jabberwocky hit it on the head, it's all the background startup apps that slow it down. Just tell them to install more ram. Works like magic 99% of the time with slow PCs.

I've yet to see real-world results with defragging. The only thing that ever showed real results was degfragging/optimizing of the boot files - had an app that did it for me, forgot the name.
This *should* do the same... Start > CMD > defrag %systemdrive% /b
[-]Try it. Works like a bomb.[/-] Remember to time your system startup/shutdown times before & after!

Edit: Should being the verb. Let me try and find that app again... It did 2 things, defrag %systemdrive% /b was one of them afaik. Now to find that purple-looking-app

Keeper said:

It's logical that it will boost performance on HEAVILY fragmented drives, but between a 0% vs 10-20% fragmentation there won't be much difference.

Click to expand...

Thanks guys, I know how to improve the speed of systems, and I know/suspect that defrag does nothing. What I want to do is test the actual performance benefits of defrag; most of what you see is product verus product, heresay or PR. I need proof

Its logical to some people Some IT managers don't deal in logic; my discussion last week with a customer:

IT MAN: Conrad, my server is slow.
Conrad: Yes, its old and crap; upgrade. 500 users can't work on a system designed for 30.
IT MAN: Can't we defrag?
Conrad: No. Upgrade.
IT MAN: Why can't we defrag? You just want to sell us hardware!
Conrad: And the defrag guys - they giving their stuff away??
IT MAN: I saw a white paper....
Conrad: Advert
IT MAN: ...that said there was, er, pardon?
Conrad: You saw an advert - not a white paper.
IT MAN: OK, I saw an advert, that said there was a 300% increase in performance!
Conrad: Lets say I run the defrag. It will take 20 hours. It won't help squat - and then when I bill you, you will refuse to pay because I could not fix the problem....
IT MAN: But then we'll buy this improved version!
Conrad: It wont work either....
IT MAN: But Conrad, you just don't understand defrag! It takes all the files scattered all over the disk and puts them SEQUENTIALLY on the disk!
Conrad: And windows defrag?
IT MAN: But Conrad, you just don't understand defrag! This special program takes all the files scattered all over the disk and puts them more SEQUENTIALLY'er on the disk than Windows ever could!

And so it goes around in circles....

trust me, if your files are scattered too much laying all over your hard drive then your virtual memory file will need to lay all around too, between all the scattered sectors - not good.

best to keep your virtual memory as one solid chunk. also good to disable it from expanding & shrinking.

now, i'm not saying a defrag will fix it all - disabling stupid bloaty startup programs is where you start first - but a heavilly fragmented drive IS going to slow you down, trust me.

It's logical because your Drive head needs to jump around, here, there, everywhere!

Your HDD doesn't need every single file in place but for heavens sake defrag atleast once every year or two - just leave it for a night or so....

The okes at microsoft didn't add it into windows for the joke of it.

also, it will expand your Disk Life. (And defraging it every second day will shorten it.)

Contig is a tool here to defrag a single file:
http://technet.microsoft.com/en-us/sysinternals/bb897428.aspx
Autoruns will tell you EVERYTHING that starts:
http://technet.microsoft.com/en-us/sysinternals/bb963902.aspx
VMMap gives the memory usage of each application, and feature/file. So you can see how much memory a word document for example uses!
http://technet.microsoft.com/en-us/sysinternals/dd535533.aspx
And then (unrelated) desktops does does what Linux does by allowing many different desktops!
http://technet.microsoft.com/en-us/sysinternals/cc817881.aspx

You gotta love sysinternals

Just gooi some more ram in their machines LOL

Keeper said:

now, i'm not saying a defrag will fix it all - disabling stupid bloaty startup programs is where you start first - but a heavilly fragmented drive IS going to slow you down, trust me.

Click to expand...

Thats my point, it may help; but how much? If it does help, then at what %fragmentation should you run it? 10% 20% 60%? What sort of benefits do you get in performance 5%, 10%? I need a test that answers these questions with proof - tests of my own doing that I can trust

ld13 said:

Edit: Should being the verb. Let me try and find that app again... It did 2 things, defrag %systemdrive% /b was one of them afaik. Now to find that purple-looking-app

Click to expand...

BootVis?

PeterCH said:

BootVis?

Click to expand...

No. That ain't it. /me hits himself on the head

Edit: YAY!! Went looking through screenshots in Google Image search ... Found it. Freaking first image result lol.

Here we go: http://www.driverheaven.net/dforce/
It is called TuneXP 1.5 by DriverHeaven.net. Under "Memory & File System" > "Ultra Fast Booting (rearrange boot files)". It is the only thing I used this app for. Worked wonders.

Depends really on the type of I/O you have. I've never seen any benefit on a PC or Laptop, but when you go up to the enterprise, especially SAN attached volumes, defragging can make a difference. If you have multiple users attached to one volume, performance is already taking a hit. If you have large I/O block sizes, 32K plus and your volume is heavily 70% or more fragmented, you will see a performance dip and it can be a major one. This is because large blocks that are sequential can be read and cached quicker.

Smaller I/O doesn't really matter. For this, Windows defrag usually does the trick. There isn't a need for all the 3rd party software out there.

But if you're talking PC's....as mentioned, kill all the smaller apps - it'll give you better performance than defragging

I decided to buy another sound card one day... thought I'd break from the norm and not get the one brand that 99% of PC's usually get...

It lopped a good 20 seconds off the time from windows login to the time that I could ACTUALLY start something.

I was amazed! That's just those stupid little programs that the sound card manufacturers THINK that you want running so that you can look at pretty blinking lights and fancy looking 3d sound whatsits while you playing music.

Old driver size... over 400MB's. Loaded over 50megs into memory at startup for volume control apps, mic monitors, speaker optimization profiles and so forth. And that's without the "environmental effects" stuff enabled.

New card... and the new drivers... 6MB's!!!!
They do exactly what a sound card is meant to do... PLAY STUFF OUT OF THE SPEAKERS!
The card is better all round (better DSPs and higher quality components) so there's no sacrifice in quality...
BUT THE REAL SURPRISE comes from my systems all round improved performance from just getting rid of those cumbersome drivers!

Moral of the story?

Go back to basics... trim the fat off. Update schedulers, file/usage monitors, desktop overlays, intricate themes and so on.
Do you really need that crap? NO!

The other biggie is the mother of all performance killers! ANYTHING from that company that starts with an "S" and ends in a "ymantic".
No matter what it tells you... ITS NEVER REALLY REALLY OFF!

Use something "leaner and meaner" instead of those "One Stop" programs.
AND...
Make sure that they don't conflict with one another... for example: if your anti-virus program has a net-filter or firewall, TURN THE WINDOWS FIREWALL OFF!

Get rid of the language bar and... dare I say it... that program that "continuously" increases performance by "continuously" defrag-ing the drives. All it does is "continuously" chow resources.

I find the odd defrag to cut disk read/write times if you let the disk get really cluttered. Especially if the disk is filled with gazzillions of small files.

Defrag-ing used to be AWESOME when we all had 120MB drives that used to read REALLY slowly, and we all had 16MB of RAM.
I remember the days of my 486 DX2-66... "Doom" would stutter unless I ran a defrag before i started it up.

I think defrag-ing is a remnant of those days... when it actually made a difference.
With SATA-2 drives that can read/write at over 60MB/s... You'll see a greater difference if you just disable MSN.

My 2c.
I used to be a firm believer in Defrag-ing. Now I'm a firm believer in "leaner and meaner" more efficient programs that don't try and be everything all at once.

'spose i should mention the random access time of any modern drive is probably on average below 8ms.

8ms. to randomly move to any sector on the disk, on average.

think.

defrag was really handy, back in the days, when it was called norton speedisk, and you had to make sure your hard drive interleave was set right too (anybody remember that?)

i dont deny it will make a difference, but i certainly oppose any suggestion that it will make a noticeable measurable difference on 99.9% of typical modern desktop machines.

ask the customer...do you run XP? if so, and your drive is bigger than 32gb, it will be ntfs.
then point them here
http://ixbtlabs.com/articles/ntfs/ and scroll down to the defragmentation section.

Does the machine have a RAID5 setup? If so, then I really don't think a defrag can do anything!

So there has been mention of SAN disks and RAID5. Strange that Diskkeeper provides a very strong argument as to why defrag is MORE important on a RAID volume than a single disk? Will need to add RAID to the test....

Fragmentation on SAN volumes. This is also an area of doubt for me: On a file server, as a basic example, you would have multiple random reads and writes. You would scale your spindle count to meet these IO demands; would laying all the files out neatly in line really benefit the multiple random IO? Or any other IO? Remembering that sequential data in the form of SQL logs etc that could benefit from being a single contiguous file don't fragment?

Will need to test enterprise platforms too since this is where most of my work is done....

Great, I just wasted 4 hours defragging 3 drives last night...

Conradl said:

On a file server, as a basic example, you would have multiple random reads and writes. You would scale your spindle count to meet these IO demands; would laying all the files out neatly in line really benefit the multiple random IO? Or any other IO? Remembering that sequential data in the form of SQL logs etc that could benefit from being a single contiguous file don't fragment?

Click to expand...

You make some interesting points.

1. On a file server I cannot see any point to defragging. As you say, read/writes are totally random, excerbated as the user count increases. IMHO the ONLY way to improve this is to add memory which is used for disk cache. This can be on the disk controller and/or in the server.

2. For best performance on a stressed server, SQL logs should be kept on a separate drive from data, better still on a separate controller. A mirrored drive would improve throughput. All of this assumes that the box is dedicated to SQL and not shared as a file or Application server, which lets out SBS. A similar argument could be applied to Exchange.