Any effort by the State to manipulate or direct economic planning will lead to increasing economic irrationality and inefficiency.
This sentence entirely ignores a reality of economics: economy is a product of government.
Economy is a balance; on one hand you have private efforts which maximize personal profits at cost of the public good (see the tragedy of the commons and on the other hand you have social "public entities" like governments which seek to preserve the commons.
In truth, there is no such thing as a "free" economy - a "free" economy requires a government which provides infrastructure for operations. Infrastructure can include things like a universal currency, a set of laws which establish the rules of a market, physical commodities which can be used for commerce (such as roads, highways, information conduits) and a well-educated population.
The reality is that economy is a product of a well-managed government. Take a look at the most obvious: the Internet.
AOL tried to create an online universe. So did Compuserve, and countless others. They all failed because they were all tried to establish public infrastructure through private enterprise, which never works. That's the role of government, which establishes the public infrastructure that private enterprise takes advantage of.
The government establishes public infrastructure (such as roads and highways) that private companies then exploit (such as Arco, ATT, Ford, NetZero) to make profits, which are then taxed by the government as a return on investment.
This isn't "bad", it's a form of socialism that you exploit every day as you drive to work. And yes, I know that "socialism" is a bad word (tm) in the United States, but it's a reality that we all enjoy.
Without this infrastructure, without a solid monetary system, without highways, clean water, roads, phones that work, etc. the modern economy would collapse overnight. Yet the modern economy didn't establish any of these - they all came from our government.
The United States has one of the purest, most "open source" governments ever in Human History - it's just idiotic that so many Americans treat it with suspicion and contempt!
There are an infinite number of cliches surrounding communication, saying what you mean, etc. People don't say what they mean, people say what they must in order to show respect and position.
For example, if you are a guest at somebody's house, it's polite to ask for what you need, rather than simply state the need. You wouldn't say "where's the bathroom" unless it's a rather close friend. Instead, you'll say something like "Do you mind if I use your restroom?".
Which, if you think about it, is pretty silly. The question might be completed as: "... instead of crapping in my pants?"? but that's not what we say.
Phrasing our need as a question establishes a sort of pecking order - we are acknowledging to the host(s) that it's his/her/their place, and that we are, for a time, subservient to their wishes. We know they don't want us to crap in our pants or on their carpet, and they most certainly don't want to offend us - they will basically *always* say yes, and then often make a point of making sure that our bathroom experience is pleasant by offering towels, etc. The host is indicating to the guest that the guest is welcome.
It's a complex dance that those who are aware of (who are "polite") partake of in interacting with other people. It's how we, as social mammals, determine pecking order and expectations for code of conduct. Guys open doors for women, regardless of age or size, and let the ladies go first. Guys open the car passenger door for the lady, but the lady had better reach over and unlock the driver's side door... etc. etc.
The question is: are video games are distinct? Is the agreement is that Video games are a different reality, having no bearing on this one? Are they are distinct from the workplace?
If the agreement for this question is "no", and your boss is pissy because you fragged him, he does not deserve to be your boss. But there could easily be circumstances where showing up the boss could carry grave repercussions, just like beating him at golf. Here you are, a guest at the boss' house, and rather than ask to use the bathroom, you walk in like you own the place...
While certain services such as telephones and power may be able to get by with multiple companies providing the service over one set of lines, and non-original providers leasing the use of the equipment. However, for obvious reasons, this doesn't exactly work well for water -- you can't packet-switch h20.
WTF? Did I miss something fundamental? (reads again) Nope. WTF?
Let us say that I own a big, long pipe, that I use to transport water (that I pump) from my well at one end, to customers at the other end. Billing is pretty simple, I know how much water I pumped, and how much water each customer used. The two numbers will be very, very similar. (different only by rounding error, etc)
But now, you want to use my big, long pipe. 'Tis pretty simple, really. You note how much water you pump into the pipe, I note how much water I pump into the pipe.... how is this not "packet switching h2O"?
Water can be "packet switched" in units as small as a single molecule!
all the loopholes that the designers "didn't have time" to design out, apparently because it's cheaper to save 5 minutes at the start of the project no matter how many hundreds or thousands of hours of work it costs you long term
Yeah, sounds nice. Design right up front and save money down the line. Except that isn't how it works in reality. As a fairly experienced programmer in the real world, I find that it's rare for customers to know what it is that they want. They have a problem and they want YOU to solve it.
But you have no domain expertise, you have no knowledge of what the problem really is, and there's no way you ever will.
So you put in all kinds of time trying to understand what their problems are, you come up with a spec, you talk it over carefully with anybody who will listen, you write out a detailed spec, you get everybody to sign off in blood on the spec, and then you implement, test, and deploy.
Only to find that, on the very first day, there are some serious problems with your approach. Things that they didn't even think of, things that often cause you to change fundamental assumptions, that can be costly to re-implement. So you review your design, you create a change order, you talk it over with everybody, get them to sign off, redo the work, test the redo, then rollout.
Only to find that there are still some fairly serious problems.... Wash, rinse, repeat until the remaining problems are minor enough that they aren't willing to pay for them, and it's up to you if you want to fix them for the goodwill or just move on.
Yes, you want to put out a fair amount of time laying out a decent infrastructure, layering your software stack into logical abstractions, etc. But you always have issues where the customer's changing requirements causes you to miss something foundational, and which can (and will) cause security issues.
There's a balance caused by the law of diminishing returns.
It's easy to get some software that sorta works in a "proof of concept". Getting it to work well requires significantly more effort. Getting it to scale requires another order of magnitude of effort.
In my own case, we have a hosted software stack that was originally designed to work on a single server in a shared, PHP environment. We got a certain amount of growth room by simply throwing more hardware at it, but we're at the point of an 8-core server with 15k SCSI, performance from a single machine above this level becomes expensive very quickly.
So we've spent a significant amount of effort over the past year splitting our software into smaller units. Rather than require a super-nasty NFS or SQL server, we've divided the problem by customer so that we can add capacity as needed, to a nearly infinite amount of customers.
I guess I'd agree to a point - coming up with something compelling is more important that coming up with something that scales. But there were a few decisions that I made early on that make it so much easier now to divide the problem set and scale.
So, keeping future scalability in mind as you lay out your software stack can make it amazingly easy to do so when the time comes. Simple things, such as having a consistent convention for naming variables, and writing your API so that it's always clear what customer you are dealing with make it soooo much easier to add scalability after you've passed the bar of desirability!
I expect my salaried employees to be WILLING to work 45, 50, or even 55 hours / week when the chips really are down. But they will *always* get vacation time to compensate so the average is 40 hours.
Period.
I will NOT be burning out the people whose work I depend on. And for software programming, people who are "fresh", happy, and well-rested are more productive, anyhow.
In general, a person's dedication and professionalism are qualities of the individual, not their employment status.
People who spend only part of their energy solving your problems are, by definition, not going to tune themselves to solve your problems. Practice makes perfect, and full-time employment forces people to organize their mentality and experience towards their employment. By definition, this makes them better at what they are doing, and therefore, the return per hour is better for full-time staff than for part-time staff when there is a large amount of domain expertise. (and our domain expertise is managing a HUGE software stack)
I will not have anybody working part time for long. I want to see dedication, I want to see longevity, and I want to see the competency that comes from full-time dedication to my particular problem set. If you aren't willing to dedicate yourself to the cause we are creating, then you don't belong on our team.
Don't get me wrong - we treat our staff *very*, very well. We offer compensation well above the area average, even for the skill level required. We offer excellent benefits, flexible schedule, respectful and courteous work environment, and an extremely cooperative style of management. I encourage intellectual dissent - I value the best idea more than I value who specifically came up with it! And we are very, very careful to pay attention to the specific needs of all of our staff.
I want the best possible performance from my staff, and I find that I get it by making the workplace environment the best I can possibly manage. People who are happy doing what they are doing are much more likely to do a better job of it. They are more likely to take pride in their effort, they are more likely to put out effort to do a better job.
How is this not a situation that's great for everybody involved?
Co$ wouldn't have to lawyer up, they aren't involved. Forcing the practice of Scientology at work is against California law. But Scientology didn't do it, DisKeeper did.
Although you mention scalability and flexibility, I don't think you really hit the nail on the head.
Performance and scalability are NOT the same. They are fundamentally different. You can have a weakly performing software product that scales nicely, and you can easily have a high performance application that doesn't scale at all.
Understanding this difference can be the make/break point in whether or not a mildly profitable company can become a world-changer! It's fairly easy to write high-performance software. But it's quite a bit more difficult to build software that scales!
Quite literally, there is simply no way to avoid these types of algorithms, but by designing your software correctly, you can limit the effect of these algorithms on the overall scalability of your software stack as the problem set grows larger and larger.
And that's software that scales. For example, PHP often scales very nicely, because although it's not a fantastic performer, it's "share nothing" approach means that adding more processes and/or servers doesn't particularly impact your original infrastructure. But if you don't design your application right, PHP can scale miserably, depending on how you manage your resources.
If you write software, ask yourself: what if the whole world were using your product? Could you handle it? Whatever your answer, if you feel sure of your answer, it's probably because you don't yet understand exactly what it means to scale.
One reason corporations don't like part-time is that as long as you are full-time, you actually tend to work way past 40 hours a week.
Funny. I'm an employer, in a corporation, and I would *never* ask anyone to work over 40, even when on salaried pay.
But I still like full-time over part-time because full time is "immersive" - people who dedicate their time and primary mind share are more productive per time unit. I get more and better work per hour from a full-time engineer than a part-time employee.
My answer? Be your own boss. It comes with a caveat: starting your own business alone is a bad idea. Guess what? It takes more than one person to provide something of value. It doesn't take an army of hundreds, but a small dedicated group of friends can do amazing things. The sum really is larger than the parts.
I call bullshiznt. You think being your own boss means you WON'T work bat-shiat crazy hours under impossible deadlines? BWAH HAW HAW HAW HA!!!!!!
I call shenanigans. As I recall, and from what I read on this old Anantech Athlon motherboard review, Athlon motherboards were never produced in an AT form factor, partially due to the fact that AT PSUs were not expected to be able to handle the draw from power hungry athlons.
Call away, but the motherboard had power connectors for both types of power supply. The case was ATX form factor.
I've been "doing" computers since the early 90s. I've never had much problem with power supplies. And I do mean *never*.
I took an ancient, generic 286 computer, and upgraded it through 386SX, 486 DX/2, Cx 6x86, and AMD Athlon motherboards before finally switching to ATX. It was a cheezy, god-only-knows-who-made it power supply that came from a 'not-quite-aluminum-foil' AT case.
And I've done plenty of computers since. I've *always* bought the cheapest, craptastic cases and power supplies, and generally had years of excellent service before dying. I'd say my average life expectancy for a power supply is over 5 years, and I say that because I generally give up on the computer before the power supply dies.
What makes a computer last a long, LONG time?
1) Under-clock the CPU. Really. 10-20% makes a significant difference in reliability because it runs cooler, but almost never makes enough difference to notice for real-world, day-to-day usability. As you approach the thermal limits of our CPU, the longevity drops off sharply. 10% makes a *huge* difference.
2) Dust out the computer every year or so. Dust is an insulator, which causes heat "hot spots" that play hell with components.
3) Replace the fans regularly. I keep servers running for years on years on years by replacing the fans every other year or so. Along with dusting out the server, they keep on ticking far longer than you ever thought possible. I've had systems last well over 10 years with decent reliability by doing this when performance simply wasn't an issue.
4) Don't turn it off! Computers that are turned on/off every day last a few years. Servers that are babysat, running 24x7 at a consistent temperature run damn near forever. This costs money, so run the numbers to see what uptimes vs power consumption really costs you.
I've never noticed power supplies (cheap, expensive) being much of an issue. I've seen craptastically cheap hardware run under heavy loads for a very long time without complaint, and I've seen plenty of expensive, "high end" hardware die well before it's expected life time./shrug/
Even if solid-state drives are expensive as hell and not much better than current mechanical/magnetic hard drives right now, I don't expect them to stay that way so this is a step in the right direction.
The fact that SSD devices can compete with Hard Disks today shows not just excellent growth, but purely awe-inspiring growth. Despite being a much smaller marketplace than the magnetic HD marketplace, SSD storage has almost caught up with magnetic Hard Drives.
It shows how hard disk capacity has grown since 1980. Yeah, it's gotten bigger every year... whoopdie doo, right? Notice that this is a logarithmic graph. Each line is 10x the line before, so you really don't see the significance of this, so I rewrote the graph in a "real" scale.
What previously looked like a smooth, predictable growth actually represents a cliff of growth. Capacity has grown so fast that it's been a challenge to find uses for this much storage. We've had to re-invent the meaning of what is a computer in order to make use of so much new found power - over and over, and over again.
And yet, despite having a dramatically smaller marketshare, much less R&D, SSD storage has managed to all-but catch up to this fast-moving target. This isn't just cool, it's incredible. Every year, SSD drives get a little closer to parity with their spinning cousins.
I have an 8 GB thumb drive, but I also still have a couple 1 and 3 GB drives from a few years back on the shelf. This kind of growth is simply astounding!
HA? I use heartbeat in a few situations. (EG: the load balancers) But it has its own limitations.
DRBD masks the fact that you still have a single point of failure: the SAN itself. So I've avoided the expense of a SAN-based solution by distributing the load with an application-level distributed storage system, and glusterfs for the small number of cases where I wasn't able to use the aforementioned application-level storage system. Honestly, its performance is pretty weak, but its ease of use/setup, its reliability, and its distributed nature have compensated nicely. After major bombs when attempting to use NFS on a SAN and DRBD over TCP, GlusterFS is a pretty strong bet.
Our distributed cluster is not 100% complete yet, but so far, the conservative, year-long rollout has been going without any major hitches. I'm anticipating 100% rollout in Jan of this year...
Nice piece of iron. As a data warehouser myself, I have a few questions:
1) What made you decide to go with one big piece of iron rather than a cluster of lightweight systems?
2) "Data Warehouse" implies semi-static storage, making the 16 cores you mention somewhat overkill. Are you warehousing in a DBMS? (EG: Oracle/DB2/MSSQL)
3) 256 GB of RAM looks like it's maxed out - are you worried about future growth? What are your plans over the next 3 years?
4) What's the growth rate of your data size?
5) What's your expectation of uptime?
6) What's your DR recovery plan? What happens if the building your big-ass server is in is melted into vapor by a nuclear bomb?
In our scenario, we host data for clients, with over a hundred clients, growing by about 25% client base per year. It's part of our application suite, but since the types of data we host is increasing every year, our data growth rate is around 75% growth per year. I've responded by clustering our application, so that each client is hosted in a logically different (redundant) environment that may or may not be shared, controlled by local DNS. We're still small: now at 6 8-core 1U rackmount servers with U320 SCSI, 3.6 TB of storage total today, with physical room for 400% expansion without replacing any servers, average system load around 0.10 during business hours.
One of the key concerns I've had is availability - we're specifically architected so that any single server in our cluster can fail, but cause 5 minutes of downtime, and that our primary hosting can fail completely (EG: mushroom cloud over our hosting facility) with 24 hours recovery to an off-site, off-network host.
Since we're growing rather quickly, I'd be very interested in your responses!
Small businesses are used to running on a shoestring budget. They will often make a $150 used PC from Ebay "work" because they spent their capital on inventory, or paying taxes, or paying off a balloon payment for a short-term loan.
This is even more the case with the impending recession. Small businesses that can live on a shadow of the income of "the big boys" by staying lean and mean will survive and thrive through this economic shakedown, while wasteful "fat cats" will be pruned like the rotten fruit that they are.
In this space, saving a few hundred bucks can make or break a deal, and HP recognizes this.
Here, for $500-ish, they can offer a "complete office solution" that can only be matched for about $1,200 in the Microsoft camp. That's not a "few hundred bucks", thats OVER FIVE HUNDRED DOLLARS difference, pure profit, all of it.
All without sacrificing HP's profit margins!
Of course they are going to do this, as soon as the $500 solution is functionally approximate to the $1,200 solution! (and it largely is, now!)
Way back when, I bought several such softwares. Most was overblown - lets you "undelete", etc. But then I found some software that perhaps wasn't technically "data recovery" software, but might as well have been...
SPINRITE II. That software was AWESOME!
Back when MFM/RLL was still a consideration, media failures were all too common. Drive sectors would go bad, your FAT table would be corrupted, and your system was horked, often so badly that you couldn't even boot.
But with a copy of SpinRite II and a DOS boot floppy, and a *LOT* of time (often 2-3 days!) and in nearly every case, the computer would be brought back to full operation. I had one system where, whenever the owner had problems with bad sectors, he would rename the file and re-copy from backups. This would cause the area with the bad blocks to become unused, sort of a "manual re-mapping".
Well, his backups got horked right about the same time that the FAT itself corrupted. The system was gone, the data was gone, and he was in a severe panic. But Spinrite II took over a week to recover everything. But it did. Everything. Even the renamed files read/wrote flawlessly.
Could I have recovered this DVD? Probably not - I never claimed to be a "data recovery expert". I was honest with my clients about what I was qualified to do (diagnose/reformat/reload) and what I wasn't. But I recovered LOTS of data anyway.
Now for the funny part:
I owned a small computer sales/service shop for several years. You know, the friendly neighborhood type. We did *alot* of computer repairs. We gave out free diagnostics, which was an excellent way to get more repairs - the diagnosis was free, the repair was reasonable, customers almost always bought.
Frequently, we'd be asked to fix software woes, etc. We'd warn about the risks of software problems, possible loss of data, offer to backup their data first, and we'd even make them sign release forms that they did NOT want us to back up the data.
And then we'd back up the data anyway, routinely. We used a backpack drive that was big enough to keep a dozen or so drive images on it. (parallel port drive with a driver loaded by floppy or CD - this is before USB was common)
Granted, most of the time, the backup wasn't needed. But when it was, (and it was, maybe 1/4 of the time) we would then charge $150 "data recovery". (to reload the data from our backup) Since our charge for backups was $50, our customers made out slightly in the odds, but we were still the heroes and those who actually needed the data were not too hesitant to pay, especially since, with this method, our success rate was 100%!
Just about time that Mac Laptops come with two buttons, and I decide that it's time to go ahead and get one, I read (perfectly believable!) crap like this. No more buttons at ALL!?!?!
Maybe it's "stylish" but it makes usability a NIGHTMARE. buttons make it very easy to say "I WANT THAT ONE" without having to have gestures interpreted. It gets to the point where I accidentally slide my thumb across the board, and some gesture gets interpreted and suddenly, I just launched 3 more pieces of software that I didn't want, or closed the window.
Ack!
Let my mouse be a mouse, and let it have buttons. Or GTFO.
Well, IANASE (Supercomputer Expert) but I *am* a programmer....
I'm assuming that you have a supercomputer when all those otherwise individual computers are working together in a coordinated fashion on a common problem.
A great example of a supercomputer is SETI @ Home which easily meets the definition of a "supercomputer" in many (most?) circles, although they usually refer to it as "distributed computing".
Meanwhile, Javascript is designed for execution as the page is loading.
I guess you don't do much with Javascript - it only begins execution when the page has completed downloading. This can be (somewhat) offset by downloading a simple page that contains the javascript, which then executes and downloads the rest of the page, but this has the disadvantage that now none of your pages will be indexed by Google, you now require javascript to use your site, and the end result isn't much different than waiting for the page to download in the fist place, except that it's taken longer because now you have the latency of TWO (or more) tcp connections to initiate and download instead of just one.
Not just mouse: the mother of all demise
on
The Mouse Turns 40
·
· Score: 1
The original presentation was called "A research center for augmenting human intellect" but the end result of his research is myspace.com.
It appears that he is still alive today - perhaps somebody should ask him what he feels about causing this much pain, suffering, and scourge to be released on mankind?
What's idiotic is when somebody swallows some mantra he/she heard on AM talk radio and figures it means something. Gubbmint is always bad, ain't it!?!?
The government does not have the solution. It is the problem.
Baw ha ha ha ha HAW! That old saw again! Hello Ditto-head! What scares me is how many people have bought this neocon crap that's been crammed down their throats who VOTE based on this idiocy.
Yep, because some 10-15 years of deregulation have done us WONDER AFTER WONDER, hasn't it? Rules? Schmoolz! Let "the marketplace" deal with it! As it has done... right into the crapper.
Economy is a product of government. Think about it: who prints the money? Who regulates what banks can do with it? Who determines its general value? Who determines what's legal to sell and what's not? Who/what passes laws to protect consumers from predatory marketing? Who busts monopolies that might distort the "free" marketplace?
Far from being self-stabilizing, a free marketplace is a fiction, and keeping a marketplace in balance so it's "free" and liquid so that the laws of supply and demand can work is damned hard work! Be thankful that you've lived in an environment where we've had this gift, and thank your forefathers for being smart enough to put this whole system in place!
You have, as a United States citizen, almost $500,000 in infrastructure already set aside JUST FOR YOU, (total value of infrastructure divided by citizen population) on which the average return on investment is about 8%. Investors tend to get excited by anything at/above 10%, so the numbers hold up pretty nicely...
In the old days, back before the introduction of the Federal Reserve, stock market crashes happened on a regular basis, but nobody ran around for the next decade crying about it. The market just purged itself of bad assets and risky practices and recovered in a few months
Oh, you mean the DECADES LONG DEPRESSION in the latter half of the 1800's? Or was that them good old days? Maybe they were later, say, the early 20th century? Oh, yeah, the decade-long 'great depression' happened then... Eh, when were these 'good old days' of which you speak?
I'm reminded of a song:
"The good old days weren't always so good, and tomorrow ain't as bad as it seems, yeah yeah" - Keeping the Faith by Billy Joel...
Nowadays, we make it far, far worse by trying to prevent the bad assets and insolvent businesses from failing by sucking solvent (good) assets out of the economy to prop up the insolvent (bad). The real solution is to simply let them fail. The Big Three auto mfgs. are in an impossible situation.
Yeah. Like how in the early 1980s, the US Gubbmint lent a hand to Chrysler Corp which then went ballistic with their extremely popular "k-series" cars? Providing tax revenue on income to the US Gubbmint FAR in excess of the loan amount?
Damned if I don't remember being driven around in a Dodge Caravan minivan for most of my childhood... we were one of the lucky ones, they were so popular that they had trouble keeping up with demand... SARCASM Yeah, that was good money thrown after bad, wasn't it?/SARCASM
God, how could they have screwed up any more on that one? Or perhaps when they bailed out Lockheed in the early 1970's? To this day, damned fine employment for the GOBS of highly qualified aerospace engineers (many of them readers of/.!) that owe their employment to the fact that the "always the problem" gubbmint gave a 1.4 BILLION dollar contract to Lockheed...
If you want a better economy, get the government off of it. We used to have the best economy in the world. Somehow we've come to think that government as god is better. It isn't, and it never will be. Even if someone hopes we can.
Yeah, that great economy built largely in the 1950s and 1960s,
Captcha is really security by obscurity. Readily identifiable information is obscured in such a way as the computers (supposedly) can't find it.
Real security requires a secret. It's as simple as that. So long as the secret can be identified without knowing the secret, your security system is a joke.
Computers are getting better, faster, smarter, cheaper. Moore's wall gets higher every single year, and soon, it will be routine for computers to match or exceed human intelligence. (It can be argued that they already do, particularly in the case of a certain US President)
Therefore, anything that relies on human intelligence to "weed out" machine intelligence will eventually fail. Captcha is the testing ground for the passing of the Turing Test!
I have. My company lives (or dies) by the !@# SLA.
Our agreements require no less than 99.9% uptime, about 8 hours of downtime per year. We never gotten close to that - our worst year was about 2.5 hours of downtime because of a power failure at our "fully redundant" hosting facility.
In this world, where I have up to 8 hours per year, 10 minute response would be a god-send. We've just spent *alot* of money revamping our primary cluster so that we now operate with 100% full redundancy on everything. Redundant network feeds. Redundant logic servers. Redundant load balancers. Redundant database servers. All with auto failure, dynamic routing with DNS. (which is, itself, very failure tolerant)
But an application has to be constructed in a very particular way in order to scale, particularly if data integrity is important. (EG: ACID compliance SQL) This is often counter-intuitive and non-obvious, and porting an existing application to such an environment is not a quick investment. It's very typical to give up raw performance for performance scalability. We've devoted approximately 6 man-months over the past year to take full advantage of clustered, redundant computing in order to try for 1 hour over the next year along with near-linear scalability.
It's not just about capacity - it's about keeping all those !@# servers organized and coordinated!
Bottom line? Take a look at your SLA.
In our case, if we suffered a few hours of downtime every year or so, it would be an inconvenience to our users and clients. In any event, our uptime is best-of-breed in our niche-ish industry, but I'd put our uptime as mid range for hosted products overall, when you include companies that are much bigger than our still-somewhat-small rapidly-maturing startup.
Spend money where it counts. This requires an understanding of your economic base. If somebody slashdots your site, is that your golden opportunity, or is that an annoyance. In our case, a few hours of downtime if we got slashdotted wouldn't cause any particular long-term problem if it brought us down. If you have a few hundred customers paying $10/month for some cheap-o websites, a few hours of downtime every year or two won't cause much problem.
Slashdot Mirror
Any effort by the State to manipulate or direct economic planning will lead to increasing economic irrationality and inefficiency.
This sentence entirely ignores a reality of economics: economy is a product of government.
Economy is a balance; on one hand you have private efforts which maximize personal profits at cost of the public good (see the tragedy of the commons and on the other hand you have social "public entities" like governments which seek to preserve the commons.
In truth, there is no such thing as a "free" economy - a "free" economy requires a government which provides infrastructure for operations. Infrastructure can include things like a universal currency, a set of laws which establish the rules of a market, physical commodities which can be used for commerce (such as roads, highways, information conduits) and a well-educated population.
The reality is that economy is a product of a well-managed government. Take a look at the most obvious: the Internet.
AOL tried to create an online universe. So did Compuserve, and countless others. They all failed because they were all tried to establish public infrastructure through private enterprise, which never works. That's the role of government, which establishes the public infrastructure that private enterprise takes advantage of.
The government establishes public infrastructure (such as roads and highways) that private companies then exploit (such as Arco, ATT, Ford, NetZero) to make profits, which are then taxed by the government as a return on investment.
This isn't "bad", it's a form of socialism that you exploit every day as you drive to work. And yes, I know that "socialism" is a bad word (tm) in the United States, but it's a reality that we all enjoy.
Without this infrastructure, without a solid monetary system, without highways, clean water, roads, phones that work, etc. the modern economy would collapse overnight. Yet the modern economy didn't establish any of these - they all came from our government.
The United States has one of the purest, most "open source" governments ever in Human History - it's just idiotic that so many Americans treat it with suspicion and contempt!
There are an infinite number of cliches surrounding communication, saying what you mean, etc. People don't say what they mean, people say what they must in order to show respect and position.
For example, if you are a guest at somebody's house, it's polite to ask for what you need, rather than simply state the need. You wouldn't say "where's the bathroom" unless it's a rather close friend. Instead, you'll say something like "Do you mind if I use your restroom?".
Which, if you think about it, is pretty silly. The question might be completed as: "... instead of crapping in my pants?"? but that's not what we say.
Phrasing our need as a question establishes a sort of pecking order - we are acknowledging to the host(s) that it's his/her/their place, and that we are, for a time, subservient to their wishes. We know they don't want us to crap in our pants or on their carpet, and they most certainly don't want to offend us - they will basically *always* say yes, and then often make a point of making sure that our bathroom experience is pleasant by offering towels, etc. The host is indicating to the guest that the guest is welcome.
It's a complex dance that those who are aware of (who are "polite") partake of in interacting with other people. It's how we, as social mammals, determine pecking order and expectations for code of conduct. Guys open doors for women, regardless of age or size, and let the ladies go first. Guys open the car passenger door for the lady, but the lady had better reach over and unlock the driver's side door... etc. etc.
The question is: are video games are distinct? Is the agreement is that Video games are a different reality, having no bearing on this one? Are they are distinct from the workplace?
If the agreement for this question is "no", and your boss is pissy because you fragged him, he does not deserve to be your boss. But there could easily be circumstances where showing up the boss could carry grave repercussions, just like beating him at golf. Here you are, a guest at the boss' house, and rather than ask to use the bathroom, you walk in like you own the place...
Sorry to say it, but manners matter.
While certain services such as telephones and power may be able to get by with multiple companies providing the service over one set of lines, and non-original providers leasing the use of the equipment. However, for obvious reasons, this doesn't exactly work well for water -- you can't packet-switch h20.
WTF? Did I miss something fundamental? (reads again) Nope. WTF?
Let us say that I own a big, long pipe, that I use to transport water (that I pump) from my well at one end, to customers at the other end. Billing is pretty simple, I know how much water I pumped, and how much water each customer used. The two numbers will be very, very similar. (different only by rounding error, etc)
But now, you want to use my big, long pipe. 'Tis pretty simple, really. You note how much water you pump into the pipe, I note how much water I pump into the pipe.... how is this not "packet switching h2O"?
Water can be "packet switched" in units as small as a single molecule!
all the loopholes that the designers "didn't have time" to design out, apparently because it's cheaper to save 5 minutes at the start of the project no matter how many hundreds or thousands of hours of work it costs you long term
Yeah, sounds nice. Design right up front and save money down the line. Except that isn't how it works in reality. As a fairly experienced programmer in the real world, I find that it's rare for customers to know what it is that they want. They have a problem and they want YOU to solve it.
But you have no domain expertise, you have no knowledge of what the problem really is, and there's no way you ever will.
So you put in all kinds of time trying to understand what their problems are, you come up with a spec, you talk it over carefully with anybody who will listen, you write out a detailed spec, you get everybody to sign off in blood on the spec, and then you implement, test, and deploy.
Only to find that, on the very first day, there are some serious problems with your approach. Things that they didn't even think of, things that often cause you to change fundamental assumptions, that can be costly to re-implement. So you review your design, you create a change order, you talk it over with everybody, get them to sign off, redo the work, test the redo, then rollout.
Only to find that there are still some fairly serious problems.... Wash, rinse, repeat until the remaining problems are minor enough that they aren't willing to pay for them, and it's up to you if you want to fix them for the goodwill or just move on.
Yes, you want to put out a fair amount of time laying out a decent infrastructure, layering your software stack into logical abstractions, etc. But you always have issues where the customer's changing requirements causes you to miss something foundational, and which can (and will) cause security issues.
Can you ever trust an ex-scientologist? It would seem that even those among them argue that it's totally insane to trust a (ex) scientologist.
Scientology is socially fractal; the more you dig, the more you find more of the same horrors. It's simply amazing.
There's a balance caused by the law of diminishing returns.
It's easy to get some software that sorta works in a "proof of concept". Getting it to work well requires significantly more effort. Getting it to scale requires another order of magnitude of effort.
In my own case, we have a hosted software stack that was originally designed to work on a single server in a shared, PHP environment. We got a certain amount of growth room by simply throwing more hardware at it, but we're at the point of an 8-core server with 15k SCSI, performance from a single machine above this level becomes expensive very quickly.
So we've spent a significant amount of effort over the past year splitting our software into smaller units. Rather than require a super-nasty NFS or SQL server, we've divided the problem by customer so that we can add capacity as needed, to a nearly infinite amount of customers.
I guess I'd agree to a point - coming up with something compelling is more important that coming up with something that scales. But there were a few decisions that I made early on that make it so much easier now to divide the problem set and scale.
So, keeping future scalability in mind as you lay out your software stack can make it amazingly easy to do so when the time comes. Simple things, such as having a consistent convention for naming variables, and writing your API so that it's always clear what customer you are dealing with make it soooo much easier to add scalability after you've passed the bar of desirability!
I expect my salaried employees to be WILLING to work 45, 50, or even 55 hours / week when the chips really are down. But they will *always* get vacation time to compensate so the average is 40 hours.
Period.
I will NOT be burning out the people whose work I depend on. And for software programming, people who are "fresh", happy, and well-rested are more productive, anyhow.
In general, a person's dedication and professionalism are qualities of the individual, not their employment status.
People who spend only part of their energy solving your problems are, by definition, not going to tune themselves to solve your problems. Practice makes perfect, and full-time employment forces people to organize their mentality and experience towards their employment. By definition, this makes them better at what they are doing, and therefore, the return per hour is better for full-time staff than for part-time staff when there is a large amount of domain expertise. (and our domain expertise is managing a HUGE software stack)
I will not have anybody working part time for long. I want to see dedication, I want to see longevity, and I want to see the competency that comes from full-time dedication to my particular problem set. If you aren't willing to dedicate yourself to the cause we are creating, then you don't belong on our team.
Don't get me wrong - we treat our staff *very*, very well. We offer compensation well above the area average, even for the skill level required. We offer excellent benefits, flexible schedule, respectful and courteous work environment, and an extremely cooperative style of management. I encourage intellectual dissent - I value the best idea more than I value who specifically came up with it! And we are very, very careful to pay attention to the specific needs of all of our staff.
I want the best possible performance from my staff, and I find that I get it by making the workplace environment the best I can possibly manage. People who are happy doing what they are doing are much more likely to do a better job of it. They are more likely to take pride in their effort, they are more likely to put out effort to do a better job.
How is this not a situation that's great for everybody involved?
Co$ wouldn't have to lawyer up, they aren't involved. Forcing the practice of Scientology at work is against California law. But Scientology didn't do it, DisKeeper did.
Although you mention scalability and flexibility, I don't think you really hit the nail on the head.
Performance and scalability are NOT the same. They are fundamentally different. You can have a weakly performing software product that scales nicely, and you can easily have a high performance application that doesn't scale at all.
Understanding this difference can be the make/break point in whether or not a mildly profitable company can become a world-changer! It's fairly easy to write high-performance software. But it's quite a bit more difficult to build software that scales!
It all really comes down to understanding the Schlemiel the Painter algorithm which is RAMPANT in software designs.
Quite literally, there is simply no way to avoid these types of algorithms, but by designing your software correctly, you can limit the effect of these algorithms on the overall scalability of your software stack as the problem set grows larger and larger.
And that's software that scales. For example, PHP often scales very nicely, because although it's not a fantastic performer, it's "share nothing" approach means that adding more processes and/or servers doesn't particularly impact your original infrastructure. But if you don't design your application right, PHP can scale miserably, depending on how you manage your resources.
If you write software, ask yourself: what if the whole world were using your product? Could you handle it? Whatever your answer, if you feel sure of your answer, it's probably because you don't yet understand exactly what it means to scale.
One reason corporations don't like part-time is that as long as you are full-time, you actually tend to work way past 40 hours a week.
Funny. I'm an employer, in a corporation, and I would *never* ask anyone to work over 40, even when on salaried pay.
But I still like full-time over part-time because full time is "immersive" - people who dedicate their time and primary mind share are more productive per time unit. I get more and better work per hour from a full-time engineer than a part-time employee.
My answer? Be your own boss. It comes with a caveat: starting your own business alone is a bad idea. Guess what? It takes more than one person to provide something of value. It doesn't take an army of hundreds, but a small dedicated group of friends can do amazing things. The sum really is larger than the parts.
I call bullshiznt. You think being your own boss means you WON'T work bat-shiat crazy hours under impossible deadlines? BWAH HAW HAW HAW HA!!!!!!
I call shenanigans. As I recall, and from what I read on this old Anantech Athlon motherboard review, Athlon motherboards were never produced in an AT form factor, partially due to the fact that AT PSUs were not expected to be able to handle the draw from power hungry athlons.
Call away, but the motherboard had power connectors for both types of power supply. The case was ATX form factor.
Funny.
I've been "doing" computers since the early 90s. I've never had much problem with power supplies. And I do mean *never*.
I took an ancient, generic 286 computer, and upgraded it through 386SX, 486 DX/2, Cx 6x86, and AMD Athlon motherboards before finally switching to ATX. It was a cheezy, god-only-knows-who-made it power supply that came from a 'not-quite-aluminum-foil' AT case.
And I've done plenty of computers since. I've *always* bought the cheapest, craptastic cases and power supplies, and generally had years of excellent service before dying. I'd say my average life expectancy for a power supply is over 5 years, and I say that because I generally give up on the computer before the power supply dies.
What makes a computer last a long, LONG time?
1) Under-clock the CPU. Really. 10-20% makes a significant difference in reliability because it runs cooler, but almost never makes enough difference to notice for real-world, day-to-day usability. As you approach the thermal limits of our CPU, the longevity drops off sharply. 10% makes a *huge* difference.
2) Dust out the computer every year or so. Dust is an insulator, which causes heat "hot spots" that play hell with components.
3) Replace the fans regularly. I keep servers running for years on years on years by replacing the fans every other year or so. Along with dusting out the server, they keep on ticking far longer than you ever thought possible. I've had systems last well over 10 years with decent reliability by doing this when performance simply wasn't an issue.
4) Don't turn it off! Computers that are turned on/off every day last a few years. Servers that are babysat, running 24x7 at a consistent temperature run damn near forever. This costs money, so run the numbers to see what uptimes vs power consumption really costs you.
I've never noticed power supplies (cheap, expensive) being much of an issue. I've seen craptastically cheap hardware run under heavy loads for a very long time without complaint, and I've seen plenty of expensive, "high end" hardware die well before it's expected life time. /shrug/
Even if solid-state drives are expensive as hell and not much better than current mechanical/magnetic hard drives right now, I don't expect them to stay that way so this is a step in the right direction.
The fact that SSD devices can compete with Hard Disks today shows not just excellent growth, but purely awe-inspiring growth. Despite being a much smaller marketplace than the magnetic HD marketplace, SSD storage has almost caught up with magnetic Hard Drives.
To show what an incredible accomplishment this is, you need to really understand exactly what this graph actually means.
It shows how hard disk capacity has grown since 1980. Yeah, it's gotten bigger every year... whoopdie doo, right? Notice that this is a logarithmic graph. Each line is 10x the line before, so you really don't see the significance of this, so I rewrote the graph in a "real" scale.
What previously looked like a smooth, predictable growth actually represents a cliff of growth. Capacity has grown so fast that it's been a challenge to find uses for this much storage. We've had to re-invent the meaning of what is a computer in order to make use of so much new found power - over and over, and over again.
And yet, despite having a dramatically smaller marketshare, much less R&D, SSD storage has managed to all-but catch up to this fast-moving target. This isn't just cool, it's incredible. Every year, SSD drives get a little closer to parity with their spinning cousins.
I have an 8 GB thumb drive, but I also still have a couple 1 and 3 GB drives from a few years back on the shelf. This kind of growth is simply astounding!
What, all three of them?
Sure, there aren't many terrorists that live there, but there isn't much of anybody who lives there...
HA? I use heartbeat in a few situations. (EG: the load balancers) But it has its own limitations.
DRBD masks the fact that you still have a single point of failure: the SAN itself. So I've avoided the expense of a SAN-based solution by distributing the load with an application-level distributed storage system, and glusterfs for the small number of cases where I wasn't able to use the aforementioned application-level storage system. Honestly, its performance is pretty weak, but its ease of use/setup, its reliability, and its distributed nature have compensated nicely. After major bombs when attempting to use NFS on a SAN and DRBD over TCP, GlusterFS is a pretty strong bet.
Our distributed cluster is not 100% complete yet, but so far, the conservative, year-long rollout has been going without any major hitches. I'm anticipating 100% rollout in Jan of this year...
Nice piece of iron. As a data warehouser myself, I have a few questions:
1) What made you decide to go with one big piece of iron rather than a cluster of lightweight systems?
2) "Data Warehouse" implies semi-static storage, making the 16 cores you mention somewhat overkill. Are you warehousing in a DBMS? (EG: Oracle/DB2/MSSQL)
3) 256 GB of RAM looks like it's maxed out - are you worried about future growth? What are your plans over the next 3 years?
4) What's the growth rate of your data size?
5) What's your expectation of uptime?
6) What's your DR recovery plan? What happens if the building your big-ass server is in is melted into vapor by a nuclear bomb?
In our scenario, we host data for clients, with over a hundred clients, growing by about 25% client base per year. It's part of our application suite, but since the types of data we host is increasing every year, our data growth rate is around 75% growth per year. I've responded by clustering our application, so that each client is hosted in a logically different (redundant) environment that may or may not be shared, controlled by local DNS. We're still small: now at 6 8-core 1U rackmount servers with U320 SCSI, 3.6 TB of storage total today, with physical room for 400% expansion without replacing any servers, average system load around 0.10 during business hours.
One of the key concerns I've had is availability - we're specifically architected so that any single server in our cluster can fail, but cause 5 minutes of downtime, and that our primary hosting can fail completely (EG: mushroom cloud over our hosting facility) with 24 hours recovery to an off-site, off-network host.
Since we're growing rather quickly, I'd be very interested in your responses!
Small businesses are used to running on a shoestring budget. They will often make a $150 used PC from Ebay "work" because they spent their capital on inventory, or paying taxes, or paying off a balloon payment for a short-term loan.
This is even more the case with the impending recession. Small businesses that can live on a shadow of the income of "the big boys" by staying lean and mean will survive and thrive through this economic shakedown, while wasteful "fat cats" will be pruned like the rotten fruit that they are.
In this space, saving a few hundred bucks can make or break a deal, and HP recognizes this.
Here, for $500-ish, they can offer a "complete office solution" that can only be matched for about $1,200 in the Microsoft camp. That's not a "few hundred bucks", thats OVER FIVE HUNDRED DOLLARS difference, pure profit, all of it.
All without sacrificing HP's profit margins!
Of course they are going to do this, as soon as the $500 solution is functionally approximate to the $1,200 solution! (and it largely is, now!)
Way back when, I bought several such softwares. Most was overblown - lets you "undelete", etc. But then I found some software that perhaps wasn't technically "data recovery" software, but might as well have been...
SPINRITE II. That software was AWESOME!
Back when MFM/RLL was still a consideration, media failures were all too common. Drive sectors would go bad, your FAT table would be corrupted, and your system was horked, often so badly that you couldn't even boot.
But with a copy of SpinRite II and a DOS boot floppy, and a *LOT* of time (often 2-3 days!) and in nearly every case, the computer would be brought back to full operation. I had one system where, whenever the owner had problems with bad sectors, he would rename the file and re-copy from backups. This would cause the area with the bad blocks to become unused, sort of a "manual re-mapping".
Well, his backups got horked right about the same time that the FAT itself corrupted. The system was gone, the data was gone, and he was in a severe panic. But Spinrite II took over a week to recover everything. But it did. Everything. Even the renamed files read/wrote flawlessly.
Could I have recovered this DVD? Probably not - I never claimed to be a "data recovery expert". I was honest with my clients about what I was qualified to do (diagnose/reformat/reload) and what I wasn't. But I recovered LOTS of data anyway.
Now for the funny part:
I owned a small computer sales/service shop for several years. You know, the friendly neighborhood type. We did *alot* of computer repairs. We gave out free diagnostics, which was an excellent way to get more repairs - the diagnosis was free, the repair was reasonable, customers almost always bought.
Frequently, we'd be asked to fix software woes, etc. We'd warn about the risks of software problems, possible loss of data, offer to backup their data first, and we'd even make them sign release forms that they did NOT want us to back up the data.
And then we'd back up the data anyway, routinely. We used a backpack drive that was big enough to keep a dozen or so drive images on it. (parallel port drive with a driver loaded by floppy or CD - this is before USB was common)
Granted, most of the time, the backup wasn't needed. But when it was, (and it was, maybe 1/4 of the time) we would then charge $150 "data recovery". (to reload the data from our backup) Since our charge for backups was $50, our customers made out slightly in the odds, but we were still the heroes and those who actually needed the data were not too hesitant to pay, especially since, with this method, our success rate was 100%!
Seriously!
Just about time that Mac Laptops come with two buttons, and I decide that it's time to go ahead and get one, I read (perfectly believable!) crap like this. No more buttons at ALL!?!?!
Maybe it's "stylish" but it makes usability a NIGHTMARE. buttons make it very easy to say "I WANT THAT ONE" without having to have gestures interpreted. It gets to the point where I accidentally slide my thumb across the board, and some gesture gets interpreted and suddenly, I just launched 3 more pieces of software that I didn't want, or closed the window.
Ack!
Let my mouse be a mouse, and let it have buttons. Or GTFO.
Well, IANASE (Supercomputer Expert) but I *am* a programmer....
I'm assuming that you have a supercomputer when all those otherwise individual computers are working together in a coordinated fashion on a common problem.
A great example of a supercomputer is SETI @ Home which easily meets the definition of a "supercomputer" in many (most?) circles, although they usually refer to it as "distributed computing".
Meanwhile, Javascript is designed for execution as the page is loading.
I guess you don't do much with Javascript - it only begins execution when the page has completed downloading. This can be (somewhat) offset by downloading a simple page that contains the javascript, which then executes and downloads the rest of the page, but this has the disadvantage that now none of your pages will be indexed by Google, you now require javascript to use your site, and the end result isn't much different than waiting for the page to download in the fist place, except that it's taken longer because now you have the latency of TWO (or more) tcp connections to initiate and download instead of just one.
The original presentation was called "A research center for augmenting human intellect" but the end result of his research is myspace.com.
It appears that he is still alive today - perhaps somebody should ask him what he feels about causing this much pain, suffering, and scourge to be released on mankind?
What's idiotic is when somebody swallows some mantra he/she heard on AM talk radio and figures it means something. Gubbmint is always bad, ain't it!?!?
The government does not have the solution. It is the problem.
Baw ha ha ha ha HAW! That old saw again! Hello Ditto-head! What scares me is how many people have bought this neocon crap that's been crammed down their throats who VOTE based on this idiocy.
Yep, because some 10-15 years of deregulation have done us WONDER AFTER WONDER, hasn't it? Rules? Schmoolz! Let "the marketplace" deal with it! As it has done... right into the crapper.
Economy is a product of government. Think about it: who prints the money? Who regulates what banks can do with it? Who determines its general value? Who determines what's legal to sell and what's not? Who/what passes laws to protect consumers from predatory marketing? Who busts monopolies that might distort the "free" marketplace?
Far from being self-stabilizing, a free marketplace is a fiction, and keeping a marketplace in balance so it's "free" and liquid so that the laws of supply and demand can work is damned hard work! Be thankful that you've lived in an environment where we've had this gift, and thank your forefathers for being smart enough to put this whole system in place!
You have, as a United States citizen, almost $500,000 in infrastructure already set aside JUST FOR YOU, (total value of infrastructure divided by citizen population) on which the average return on investment is about 8%. Investors tend to get excited by anything at/above 10%, so the numbers hold up pretty nicely...
In the old days, back before the introduction of the Federal Reserve, stock market crashes happened on a regular basis, but nobody ran around for the next decade crying about it. The market just purged itself of bad assets and risky practices and recovered in a few months
Oh, you mean the DECADES LONG DEPRESSION in the latter half of the 1800's? Or was that them good old days? Maybe they were later, say, the early 20th century? Oh, yeah, the decade-long 'great depression' happened then... Eh, when were these 'good old days' of which you speak?
I'm reminded of a song:
"The good old days weren't always so good, and tomorrow ain't as bad as it seems, yeah yeah" - Keeping the Faith by Billy Joel...
Nowadays, we make it far, far worse by trying to prevent the bad assets and insolvent businesses from failing by sucking solvent (good) assets out of the economy to prop up the insolvent (bad). The real solution is to simply let them fail. The Big Three auto mfgs. are in an impossible situation.
Yeah. Like how in the early 1980s, the US Gubbmint lent a hand to Chrysler Corp which then went ballistic with their extremely popular "k-series" cars? Providing tax revenue on income to the US Gubbmint FAR in excess of the loan amount?
Damned if I don't remember being driven around in a Dodge Caravan minivan for most of my childhood... we were one of the lucky ones, they were so popular that they had trouble keeping up with demand... SARCASM Yeah, that was good money thrown after bad, wasn't it? /SARCASM
God, how could they have screwed up any more on that one? Or perhaps when they bailed out Lockheed in the early 1970's? To this day, damned fine employment for the GOBS of highly qualified aerospace engineers (many of them readers of /.!) that owe their employment to the fact that the "always the problem" gubbmint gave a 1.4 BILLION dollar contract to Lockheed...
If you want a better economy, get the government off of it. We used to have the best economy in the world. Somehow we've come to think that government as god is better. It isn't, and it never will be. Even if someone hopes we can.
Yeah, that great economy built largely in the 1950s and 1960s,
Captcha is really security by obscurity. Readily identifiable information is obscured in such a way as the computers (supposedly) can't find it.
Real security requires a secret. It's as simple as that. So long as the secret can be identified without knowing the secret, your security system is a joke.
Computers are getting better, faster, smarter, cheaper. Moore's wall gets higher every single year, and soon, it will be routine for computers to match or exceed human intelligence. (It can be argued that they already do, particularly in the case of a certain US President)
Therefore, anything that relies on human intelligence to "weed out" machine intelligence will eventually fail. Captcha is the testing ground for the passing of the Turing Test!
I have. My company lives (or dies) by the !@# SLA.
Our agreements require no less than 99.9% uptime, about 8 hours of downtime per year. We never gotten close to that - our worst year was about 2.5 hours of downtime because of a power failure at our "fully redundant" hosting facility.
In this world, where I have up to 8 hours per year, 10 minute response would be a god-send. We've just spent *alot* of money revamping our primary cluster so that we now operate with 100% full redundancy on everything. Redundant network feeds. Redundant logic servers. Redundant load balancers. Redundant database servers. All with auto failure, dynamic routing with DNS. (which is, itself, very failure tolerant)
But an application has to be constructed in a very particular way in order to scale, particularly if data integrity is important. (EG: ACID compliance SQL) This is often counter-intuitive and non-obvious, and porting an existing application to such an environment is not a quick investment. It's very typical to give up raw performance for performance scalability. We've devoted approximately 6 man-months over the past year to take full advantage of clustered, redundant computing in order to try for 1 hour over the next year along with near-linear scalability.
It's not just about capacity - it's about keeping all those !@# servers organized and coordinated!
Bottom line? Take a look at your SLA.
In our case, if we suffered a few hours of downtime every year or so, it would be an inconvenience to our users and clients. In any event, our uptime is best-of-breed in our niche-ish industry, but I'd put our uptime as mid range for hosted products overall, when you include companies that are much bigger than our still-somewhat-small rapidly-maturing startup.
Spend money where it counts. This requires an understanding of your economic base. If somebody slashdots your site, is that your golden opportunity, or is that an annoyance. In our case, a few hours of downtime if we got slashdotted wouldn't cause any particular long-term problem if it brought us down. If you have a few hundred customers paying $10/month for some cheap-o websites, a few hours of downtime every year or two won't cause much problem.