Monday, February 22, 2010

(Editorial): When Microsoft Attacks ... Again

Raw nerves. You know you’ve hit one when the entity in question practically jumps through the roof to staunch the pain. In my case, the nerve belonged to Microsoft Corporation. And true to form, the company spent incalculable political capital – and cashed in more than a few favors – in order to orchestrate the most one-sided smear campaign in the history of IT journalism.

What has been said about me personally, or Devil Mountain Software as a company, is irrelevant – all de rigeur for the rabid tabloid crowd. Rather, what is disturbing is the timing of it all. The parties in question only loosed their dogs after this project, the exo.performance.network, hit a bit too close to home. It was our research into Windows 7 performance that prompted Microsoft to call in its chips.

And call them in it did, instructing its media cronies to silence me by dragging my name through the mud and casting doubt about what is by any measure a very successful professional history. And now, with xpnet.com, I’m doing more to expose the inner workings of the Windows community than ever before, putting myself on the line so that the truth about Windows market share composition, usage trends and real-world performance is known.

So, despite the hoopla surrounding the day’s developments, I will continue to flash my light at the dark underbelly of Microsoft’s cash cow. Because, in the end, this isn’t about any one personality or pseudonym. It’s about the data, and how it describes the IT world around us. And whether that data is positive for Microsoft or negative, I will never lose my resolve or allow myself to be cowed into silence by the powers that be.

In the meantime, I’d like to invite individuals and organizations that want to know the truth about Windows to register for a free exo.performance.network account. You’ll gain access to a range of useful tools while helping us continue to deliver valuable insight into Microsoft’s dominant OS platform. Together, we can make Windows a more open environment for everyone.

Randall C. Kennedy

49 comments:

Unknown said...

I wouldn't advise that you try to play this off. Learn from Tiger Woods and apologize/'fess up now, or it'll hurt more later. The people will forgive you, but not if you try to dodge the core problem here.

Randall C. Kennedy said...

Sid,

Apologize? For what? Using a pen name when dealing with an overzealous reporter? Because that't the extend of the "deception" that everyone is so excited about. The company itself exists, has real clients and is profitable. Nothing they can say will change that or other facts, like:

* We have nearly 24,000 users at xpnet.com.
* We collect and analyze over 230 million system metrics records and over 13 billion process metrics records every week.
* We publish our findings and make all of our resources freely available to the IT industry.

People want to skewer me because they don't agree with my point of view. Microsoft wants to skewer me because I hurt sales. IDG wants to skewer me to cover their asses - because, as I pointed out to ZDNet/CNet, they knew about the Craig Barth ruse all along. And they did nothing.

If anyone needs to apologize, it's IDG - but not for the reasons they've stated. It was their hunger for page views that ultimately drove them to turn a blind eye.

Me? I just used a pseudonym in a few email exchanges and during a a couple of phone calls. The rest is all BS and posturing, and they (IDG & ZDNet) know it.

RCK

Daniel H said...

Its really sad that it has come to this, Microsoft fully stabbing someone in the back for telling it like it is, Windows 7 still has much of the same performance problems as Vista had.

I understand it as Microsoft being more than scared white at the thought of Windows 7 getting the same reception as Vista. Thus the strong reaction in this case.

They must have pulled a lot of strings and twisted a few arms for this. Spineless reporters and news agencies are the sole reason i dont buy print magazines anymore.

Randall C. Kennedy said...

Daniel,

Thanks for the support. It's refreshing to encounter someone willing to look through the FUD and see what's really going on here.

But not to worry. This whole episode only emboldens us to try even harder, dig even deeper, and never, ever let them cow us into silence.

RCK

John Rose said...

Weird. Any computer science major taking a 100-level operating systems class learns that "free RAM is wasted RAM."

There's no penalty for caching resources in RAM when the computer is idle. If the OS needs that cached resource later, great, it's already in RAM and that's a huge performance win compared to fetching it from disk. If the OS needs that memory for something else, there is zero cost for simply flushing that cached resource and allocating the memory to another process.

Thinking that all RAM usage is "bad" or "bloated" is a common, and totally understandable misconception that many non-technical users have.

What I'm wondering is if you're intentionally exploiting this understandable misconception, or if you've fallen victim to it yourself.

Tim said...

Randall

I'm worried about what I see. Your posts here and your replies over on Dignan's "expose" just aren't rational responses to the brouhaha. I don't enjoy invoking psychological disorders on people via the internet but you're beginning to sound delusional of the "grandiose" and "persecutory" kind.

Seriously. The goons in black aren't out to get you. You are not on some glorious crusade to reveal the truth. Those are the fictions in your head to justify your duplicitous practices heretofore.

Step down from this for a bit. Get some perspective. Speak to a counsellor if need be. You can turn this situation to your benefit, but only if you stop this denialism.

Sincerely,

Tim

Daniel H said...

@John Booty

Change school, free ram is never wasted. If you have enough ram for your cache, applications and OS there is no point using more just for the fun of it. You gain much more by tweaking your VM, your cache and the footprints of your applications. This magic gain in performance to be had if you use up all the memory for cache is at best a myth but in my mind an outright lie.

The best cache in the world is as small as possible and as accurate as it can be. Superfetch is highly inaccurate and sometimes caches things when it should really not and makes the system swap to disk instead, making overall performance worse than without the caching.

Randall C. Kennedy said...

Tim,

Denialism? I didn't even know that was real word - had to look it up! :-)

But seriously, "just because I'm paranoid doesn't mean there not out to get me."

Words to live by...

RCK

Smriti said...

Randall, I think I fully agree with you on the IDG one. They should stand by you on this one, but when they see that another media powerhouse is after them, they back out. Why don't you expose those backstabbers. I see that you have personally replied to each and every comment so far. Can you please document your email exchange with them here so that the tech world can what a shit house IDG is.

Smriti said...

This is interesting. Why should one be fired just because he has a pen name. Intriguing, or maybe IDG really knew it all along, and didn't find a way out of the mess. IDG sucks. They made Randall a scapegoat. Since when has been having two names wrong.
And the point is not whether he published wrong results or right results. The point is the timing of all this. Why were IDG in such a hurry to sever ties. They could have done it on Monday, nobody was running away. I smell fish in the whole episode, and I have a feeling that IDG is the main culprit here.
Randall has done nothing very wrong, as a journalist. As anti-microsoft, whatever, but not as a journalist per se. Just another scapegoat of the media powerhouse catfight.

Good luck Randal, expose those a*holes at IDG.

Unknown said...

A wild guess - "Daniel H" is your latest "pseudonym", right Randall?

And some comments on the controversy. Seriously, you were employed as a journalist, no? Of course you are aware of the ethical issues of quoting your own work and opinions as quotes from a fictional person.

Whether IDG were aware or not is irrelevant to your personal responsibility to all IDG readers who like myself read that someone named Craig Barth, CTO of Devil Mountain Software stated this or that. Where is your personal apology to me and other readers. "My employers knew" doesn't cut it.

Are you saying that IDG asked you to do what you did, that they explicitly supported it?

Are you saying that Gregg Keizer knew as well? He quotes an email from you where you write "Time to level with you". Did you write that or didn't you?

If you fail to respond to such questions and to explain circumstances plainly and transparently it's going to speak volumes.

Unknown said...

Randall,

The problem for me watching from the side is this:
1. I have no reason not to trust you about the involvement from IDG - but it's a classic 'word against word' situation.
2. You attack Microsoft without any proof at all - show me some evidence of this conspiracy.
3. You need to get some support for your findings and they should probably be independent from your company to be trustworthy.

You might be the victim here but you need to step up and show us some credible evidence - if not I can't really take you seriously.

Regards
/andreas

Wardy said...

The problem is Randall, you lied. And not just a random little white-lie like having a blogging pseudonym, which is forgivable. You've been lying about your online persona for 10 years, in various different ways, at various different places, with varying levels of success. Lying breeds distrust, and IDG wouldn't realistically be able to keep you on if you're name isn't trustworthy anymore, especially under such a conflict of interest.

I would be more willing to stand behind you, and uphold the cries of scapegoat, but the ZDNet exposé that kicked a lot of this off is very disturbing in parts.

Your software privacy policy states that you don't associate uploaded data personally, yet this is exactly what you did to oust Peter Bright's claims over at Ars Techinca. You claim that your software is trustworthy because it's installed on the trading floor at Wall Street, but they don't allow vendors to publicise their purchasing of software. And you claim that all data uploaded by your software is done so securely via SSL, but research has shown that your certificates either don't exist, are out-of-date or apply to the wrong domains.

You've only really got yourself to blame. Maybe IDG are using you as a scapegoat, but you've made that job very, very easy for them.

You're counter-argument is that Microsoft, the media et al are out to get you, but where is the refuting of the serious claims that have been made against you, and the facts to back this up? You can't shout against these legitimate claims Randall, no matter how loudly you try (and how used you are to doing so).

Unknown said...

This is quite amusing.

In the now substantially deleted post he made the claim:

And after reviewing this data, it became clear why our System Monitor widget flagged his system as being low on memory.

So, he apparently checked the data then to see why my system is red.

But then a day or two later:
I think I know why your box isn't thrashing: We never flagged your system as one of the machines in question.

In fact, you shouldn't have been flagged at all - I just assumed that you were since you posted a screenshot of System Monitor with the red bar. But when I run your system through our much more sophisticated Report Card template, you score a meager 69 on the Peak Memory Pressure Index scale.


He tries to blame the red graph on a "misconfiguration" (even though I was using the default configuration for my screenshot). And even after re-checking the configuration to confirm that it is using the settings he says it should be using:
It's still red!

My memory chart is here:
Click (though the file may be served up with the wrong MIME type, which is a hallmark of quality coding)
At the time of writing today's bar is "only" yellow, not red. I have "only" 1.4 GB physical memory available at the moment. Low on memory indeed, huh.

You can't use the counters you are using to make the claims that you are making. Page In activity does not indicate pagefile usage or lack of memory; reading from memory-mapped files, including, most importantly, executables and libraries, also causes page ins.

Unknown said...

No, Randall, I think what people think you should apologize for is lying (at one time or another you have denied that the Craig Brath character was fictitious), and failing to disclose the conflict of interest. If you can't see a conflict between "company that collects dubious data and then makes MS-bashing blog posts", and "your persona as a MS-bashing blogger based on dubious data, promoting the aforementioned company's software" then I don't know what to say. This is just so obviously something that should be disclosed to the readership.

If IDG was aware of your relationship, why are you no longer working for them?

Daniel H said...

@"Daniel"

Nope, sorry to break your illusions but i have nothing whatsoever to do with Randall.

Im just not on the astroturfing payroll.

Randall C. Kennedy said...

"If IDG was aware of your relationship, why are you no longer working for them?"

Because I quit? Which is how it went down, regardless of what Eric Knorr says to the contrary. In fact, here's the time-stamped email I sent him on Saturday (right before they pulled all my content off the site):

From: Randall C. Kennedy [rck@csaresearch.com]
Sent: Saturday, February 20, 2010 9:35 PM
To: 'eric_knorr@infoworld.com'
Cc: Galen_Gruman@infoworld.com; Doug_Dineley@infoworld.com
Subject: RE: to be clear...

Eric,

Then I guess your only remaining option is to come clean, say you were duped along with the rest of IDG, and lay the blame squarely at my feet. If that’s what it’ll take to clear the air, then so be it. It won’t be the first time I’ve been hung out to dry by InfoWorld (though it most definitely will be the last).

And with that, I’d say we’re done here…take care!

RCK

Unknown said...

What's with the weird future times on the comments posts? Where in the world are you? Must be GMT + 7 or something...

Randall C. Kennedy said...

@instig8r,

I'm here.

:-)

RCK

Tim said...

RCK: But seriously, "just because I'm paranoid doesn't mean there (sic) not out to get me."

I thought you might quip this back. I'm afraid it only serves to make my point for me.

RCK: Words to live by...

Words for a wacky t-shirt worn by sophomores maybe, but hardly a credo for a mature IT professional whose company deploys software that monitors other people's systems.

Maybe when this storm dies down you'll find your way back. Sorry that you don't yet see your culpability or how to deal with it.

And with that, I’d say I'm done here.

Randall C. Kennedy said...

@Tim,

Find my way back to where? I'm right where I need to be, and that's free of the page view obsessed IDG culture that drove me to make a caricature of myself into my only public face.

I'm absolutely thrilled that I found the fortitude to refuse their offer to stay in that cesspool. I've never been more proud of myself.

RCK

Oz Kern said...

Pen names are for reporters, not sources. You screwed up. Admit it and move on rather than trying to play the victim. You're not getting any sympathy.

Andy Babiec said...

Smriti said...
"Why should one be fired just because he has a pen name."

How is it appropriate for a journalist to write about a company he owns/works for/whatever without mentioning the affiliation???

And it isn't just once - he wrote about www.xpnet.com several times at InfoWorld without a single mention of his connection to the company.

Coyote said...

It's a breach of journalistic ethics to cite yourself as a source, using a false name, and to promote your own company without disclosing your interest in said company. What you did is wrong. Claiming a conspiracy to silence you does not negate that, it's simply an excuse, and in fact shows that you still don't recognize your own wrongdoing.

Even if you could prove your claim that IDG required you to assume a pseudonym as part of your contract for writing for them, that's simply an "I was only following orders" defense. It doesn't make what you did right, as you didn't need to write for IDG. Morally, you should have refused.

Sad.

Unknown said...

Plenty of tech journalists have companies on the side that deal with their area of expertise. And none of them try to hide it. In fact, most of them use it as a source of credentials for why you should believe what they write. What you did was dishonest and gives tech journalists a bad name.

I agree with one other poster, who said that you need to provide independent evidence to support your claim about the memory. Because it could be that your software is skewing the report somehow. NOTE that I'm not accusing you of such, I'm stating that you need independent confirmation of your findings. Because like it or not, you have a conflict of interest in saying that your software is right--and everyone else is wrong.

@DanielH: The idea that free memory is good and that tweaking your VM and cache are the way to go is misguided. The proof is in the pudding. Anyone with an A+ or other hardware cert will tell you that having empty memory, while retrieving items from the pagefile is SLOWER than having the items in memory (and no "free" memory) and not having to retrieve them from the pagefile.

Also, other people have pointed out that Linux and Mac OS X do the same thing as Windows. They prefetch items that they think you'll need. Or more accurately, they retain some of your recently used items in memory--just in case you need them again. They just aren't as obvious about it (or use as much memory for it) as Windows.

For the record: Am I a Microsoft fanboy? No. Actually I spent a crash-free and slowdown-free week in Linux recently. But, I'm a fan of using the best tool for the job. And until more developers create tools for Linux that meet or exceed the ones available for Windows, I'll be using Windows (I won't use Mac OS because it's more expensive and I can't use it on anything other than a Macintosh--legally).

Have a great day:)
Patrick.

Also for the record, I'm logged in under one of my other Gmail accounts, but I am the owner of http://patscompservices.blogspot.com (in the interest of full-disclosure and not falling into the same trap that Mr. Kennedy found himself in).

Daniel H said...

@coyote Andy, Babiec, Joe 'Zonker' Brockmeier

I find it simply hilarious that you talk about journalistic integrity, especially considering the way Waggener, Edelman, Visible and Federated works.

Daniel H said...

@Patrick

"Anyone with an A+ or other hardware cert will tell you that having empty memory, while retrieving items from the pagefile is SLOWER than having the items in memory (and no "free" memory) and not having to retrieve them from the pagefile."

Yes, the pagefile should be avoided if possible. Putting stuff into the pagefile at the expense of free memory to have a big cache, thats not smart. In that case its much better to fix the cache so that it performs better with less memory.

"Also, other people have pointed out that Linux and Mac OS X do the same thing as Windows. They prefetch items that they think you'll need. Or more accurately, they retain some of your recently used items in memory--just in case you need them again. They just aren't as obvious about it (or use as much memory for it) as Windows."

From what i can gather Linux only prefetch applications when you start them and on boot. As such it doesnt do the whole chrystal ball guessing game, wrongly loading stuff it thinks you are using instead of things you really use. It only loads things you really use or very nerly as its atleast coupled with the applications you yourself start.

If you on your Windows 7 average session loads browser, a couple of heavy applications and MS Office you will have a very slow computer on the occasions you have something else in mind. Your computer is full ahead loading stuff it thinks you will use instead of things you really will use. Last time i checked, Windows didnt come with a fortune teller.

Unknown said...

@DanielH:

I did just what you suggested (about running Windows 7 with other apps going), and I posted it under his article about What took you so long". I had Outlook 2007, Task Manager, Performance Monitor, Google with 7 tabs open, and Sun VirtualBox running an XP vm with 128M of RAM dedicated to it active. And I had a bunch of other things including TweetDeck (which was using 128MB of RAM alone) in my systray. This, on a laptop with 1.5G of RAM (32-bit OS). After VirtualBox fully loaded XP, it started to bog down a little, but definitely not to the point where it was unusable. Up until that point, it was perfectly fine (even though Task Manager said I was at 95% memory used).

I maintain that it's not Windows 7 (as my other computer is running at less than 45% memory used with just Windows 7, Windows Home Server Connector, and the Lexmark Printing Application running on 1G of RAM) that's the hog. It's the other applications that you're running. Of course this has been true with all versions of Windows, and I'm betting with other OS's too.

Have a great day:)
Patrick.

Unknown said...

@Randall:

Weird but appropriate how your "Because I quit?" response has a question mark in it. It seems obvious you are confused. You say you quit yet your "evidence" of that reads like you were informed of being exposed and fired so gave a "fine, whatever!" reply. Where in that did you say "I quit"?

Its kind of sad that you spin of all of this is so transparent to most everyone but yourself.

At least psychologically you're an interesting character and study. Outside of that, this is all just so sad to see.

Greg Folkert said...

Randall...

I have one comment for you:

Remember Ed Curry!

Randall C. Kennedy said...

@Elrond,

It's the "we're done here" part, but then you weren't there. But that won't stop you from calling me a liar so we're at an impasse.

RCK

- said...

Just admit you are wrong and move on.

Unknown said...

Though I do not support you or your software I think the article By ZDNet was out of line. How many people use alias', answer a lot. Personally I feel your alias should have been used for your writing name instead of for your business. Using an alias for the business is what hurts, how can somebody trust a company under a fake owner. We as readers do not always expect the true ID of the writer to be shared, especially in this day in age. The safety of a reporter is always important. As a computer tech I know and fully understand how low MS will go to destroy an individual, especially when you run a computer store and promote Linux over their software. Randall I would apologize for the misunderstanding that people have. Also for the apparent conflict-of-interest. This ultimately acted as a self freebie promotion of your company. I hope you continue your work and try to help resolve the issues of Windows as we all do even though I am not a fan of your software. Happy entrepreneuring and remember MS has more money than you do, and will waste as much as necessary to shut people up.

Andy Babiec said...

Daniel H said...
"I find it simply hilarious that you talk about journalistic integrity, especially considering the way Waggener, Edelman, Visible and Federated works."

Since when is a PR firm equivalent to a journalist?

And since when is improper behavior excused by "someone else is also doing it?" Isn't that the logic of 5 year olds?

Kirk B. said...

Perhaps it is time to have respected computer scientists analyse the exo.performance.network code and data. The report might answer the question as to whether or not the software and/or data correctly support the claims.

Edward Dinovo said...

My problem is that Landy Wang and Mark Russinovich explained these changes in the Windows 7 kernel at PDC09 [1] and this information was not taken into account. I think the Windows kernel design team did a fantastic job dealing with the pertinent tradeoffs for Win7 and Server2008 R2. Their design philosophy was clearly laid out, and a consequence was clearly less RAM marked as free. By all means, Randall should be free to disagree with the design, but his company needs to publish a more detailed analysis than just “less RAM is marked as unused” to make a judgment — modern operating systems are much more complex than that.

[1] PDC09: Windows 7 Developer Boot Camp

Anonymous said...

Your ego and stubbornness is astounding!

Thank you RCK for entertaining me this past week so much whilst I've been bored.

make it go more.. yaaay

Unknown said...

See Randall? This is what you get for suggesting that Firefox is doomed.

Childish Democrat said...

You still haven't responded to the easily provable point that your supposedly secure software uses a normal port 80 connection to send unencrypted data over the internet, and that you can map individual system reports to individual people. In my mind THAT is the key point here and you're obscuring that by making your nom de plume the issue.

Unknown said...

Randall,

The key issue here is your longtime lack of transparency regarding the fictitious name. Creating an alias in order to quote yourself is not going to be a common or accepted practice in any newsroom or blog outlet. It's deliberately misleading, good intentions be damned. The perception is that you're plugging a side business while reporting for an outlet...and misleading readers with a fake name so they won't know you stand to benefit. Not a very good idea.

Don't take my word for it. Describe your situation to any journalism teacher, blogger, reporter, editor, etc and you'll hear the same thing. Transparency is vital.

I urge you to man up, admit that inventing Mr. "Barth" wasn't a great idea, and pledge to do better.

Randall C. Kennedy said...

Sid,

I never said I thought it was a "good idea." I said, and you can check the interviews with eWeek, the WSJ and Internet.com, that I felt I had no other choice. I tarnished my real name by too closely associating it with the crap journalism I was being asked to conduct for InfoWorld. In retrospect, I should have used the pen name for the blog and saved my own identity for handling media inquiries to DMS.

There's no question it was a mistake to invent the Craig Barth persona. And I've already expressed to Gregg Keizer and others my regret at causing them any personal or professional damage. However, I also think the situation is being blown way out of proportion. Not only did I not profit from the ruse (exo.performance.network is a free service, one that I was never compensated for offering through InfoWorld), I actually took money out of my own pocket just to support the site and its rising bandwidth costs.

All of this "Bernie Madoff" comparison stuff is predicated on the belief that there was some sinister business plan in place to dupe people and steal their money. What these idiots fail to realize is that exo.performance.network is a labor of love, one that has never earned a dime for anyone involved.

Actually, that's not true: Both ComputerWorld and InfoWorld drew huge traffic from publishing some of our more controversial findings. In fact, given how big a few of these recent stories blew, I'd guess a fair portion of the IT media has made a tidy buck or two off of the drama surrounding first our data and then the scandal that followed.

And how much of that obscene profit will make its way to my wallet? $0. And yet I will continue to run this site and report my findings freely for as long as I can sustain such.

RCK

Unknown said...

Are you going to answer the accusations regarding your "secure" SSL software using port 80 unencrypted?

WayneDB said...

“Oh what a tangled web we weave, When first we practice to deceive” - Sir Walter Scott

Lord_Unseen said...

@RCK
"The company itself exists, has real clients and is profitable. "

This statement confuses me considering (as far as I can find), your company's only product is the exo.performance.network which according to you,

"What these idiots fail to realize is that exo.performance.network is a labor of love, one that has never earned a dime for anyone involved."

So, let's see if I got this right. exo.performance.network, pretty much the only thing your company does, doesn't make any money whatsoever, yet the company is somehow profitable? How are you turning this profit, Mr. Kennedy? Got some deal with advertisers for data?

Brandon said...

@Daniel H / Randall

Some corrections:

"The best cache in the world is as small as possible and as accurate as it can be. Superfetch is highly inaccurate and sometimes caches things when it should really not and makes the system swap to disk instead, making overall performance worse than without the caching."

This is incorrect. All else being equal, the best cache is the largest cache possible. The reasons for this should be obvious. Larger cache means more likely cache hits.

SuperFetch's prediction logic is provably very accurate. That aside, its greatest value comes from elsewhere. Particularly, SuperFetch works to ensure that pages swapped to disk are swapped back *before* they're needed if physical memory becomes available. SuperFetch actually actively prevents hard faults from happening.

It is impossible for SuperFetch to ever cause swapping. The fact that you claim this demonstrates your ignorance of the subject.


"Yes, the pagefile should be avoided if possible. Putting stuff into the pagefile at the expense of free memory to have a big cache, thats not smart. In that case its much better to fix the cache so that it performs better with less memory."

This is a straw man. Obviously that wouldn't be smart, which is why no OS works as you described. Cache memory is equivalent to free memory and is always used in preference to paging to disk.

Unknown said...

"This is incorrect. All else being equal, the best cache is the largest cache possible. The reasons for this should be obvious. Larger cache means more likely cache hits."
Wrong. Larger cache sizes introduce much larger miss penalty and thus negating the "benefits" a larger cache would bring with it. Read up on your fundamental machine architecture.

Unknown said...

Wrong. Larger cache sizes introduce much larger miss penalty and thus negating the "benefits" a larger cache would bring with it. Read up on your fundamental machine architecture.
Please, not in any meaningful way.

If lookups are hashed, they're constant time (independent of the number of elements in the cash); if they're looked up using a sorted structure of some kind, they're logarithmic (time to look up scales with log(number of elements in cache)).

In both cases, the cache can grow substantially without causing measurable delays in lookup speed.

Remember, these are just software caches; we don't have to worry about sticking transistors on the critical path causing latencies to increase by a single cycle. Lookups in the cache are always going to take hundreds or thousands of cycles.

The impact of cache size on performance is negligible. The impact on cache size on hit-rate is substantial.

Unknown said...

As for the discussion about cache size, I'll throw this at all of you.

Is it quicker to look something up in memory and use it, or is it quicker to look for something in memory, realize that it's on the hard disk (either in a pagefile or not even loaded), then go and grab it from the hard disk?

If the larger cache size is in memory only, then I say it's still quicker (even if it's constant time, logarithmic or hunt and peck) than if the computer has to access the hard disk, locate the information, and retrieve it.

Now, I'm not an expert, nor do I lay claim to be. But, I see it like this. The "cache" is how much room you have to store the things that you need, but are not immediately working on.

So, for the "smaller cache" people, what you're saying is that you feel that the computer should only be able to get small chunks of what's needed, store that in a readily accessible location, and make multiple trips to the hard drive for other chunks.

The "larger cache" people are saying that you should be able to grab bigger chunks, store it in readily accessible locations, and make fewer trips to the hard disk.

Again, I'm not an expert, but I think the larger cache people win. Less wear and tear on the hard disk. Less time taken to find the information needed. Less electricity used (electricity required to spin up the hard drive, move the heads, wait for the piece of information to come around, read, move the heads, and spin down the drive).

Now, someone explain to me how a smaller cache is better, please? Because I don't see it.

Have a great day:)
Patrick.

Oh, and to be blunt about it, as long as my computer doesn't slow down to a crawl, why should I give a rat's a$$ that my memory is 96% used up (assuming that the computer will free up unneeded memory when it needs more).

Unknown said...

The other thing I love about you Randall, is you say you were a "shock jock blogger" but there's a few problems:

1) Anything you ever said never changed anything
2) Anything you ever said never changed anything (i re-emphasise my point)
3) Your "revealed bad underbelly" of Microsoft will have never lost them 1 sale (well, maybe the sale of Windows to you, though)
4) Any no-one actually takes a word of what you say seriously. I enjoy your rants, and that's the reason for everyone else visiting anything you've ever done too, for a good giggle.

Cheers, and good luck.
I suggest you get a real job.

P.S 24,000 Users? LOL, I assume those are people you've scammed into installing the softawre.