Most people don't run enough data over 3G or 4G on their smartphones to hit even the 2GB limit. I never have, and I run a lot of data over my smartphone.
The reason is simple: Your smartphone is connected up to WiFi most of the time, particularly when you are sitting still watching video (at home, at the coffee shop, etc), and not eating from the 3G/4G plate.
Of all the services you might use on the road, internet radio is probably going to be the biggest, but I really doubt it will be a problem for most people (not everyone, but most everyone). The price point for internet bandwidth has quickly moved more towards video distribution and further away from low-bandwidth and medium-bandwidth services. That trend will continue.
This is a non-issue. I think people have gotten a little too comfortable with this expectation of unlimited data. Nobody has unlimited data, not really.
You are assuming that your internet provider will allow you to downlink at full speed 24x7, which generally isn't true. I hit up against carrier bandwidth massaging issues all the time. I've never been able to maintain full downlink rates 24x7, or anything close to full downlink rates. My uplink rate is cut in half after 10 seconds of full bandwidth use and my downlink rates are extremely variable, depending on how busy the neighborhood is.
I've also had long downlink connections cut (by the carrier's software), as in TCP RST. AT&T's U-Verse is particularly bad doing this because AT&T's routers try to track every single connection and blow up any links they've lost track of (and they lose track of links quite often).
I've been able to work around some of the issues by using a virtual network link to a nearby colo'd machine to run all my traffic through. AT&T and Comcast can no longer sniff my traffic or track connections. Comcast still caps the bandwidth, however, and AT&T's backhauls often get filled up and start dropping packets due to other traffic on the backhaul.
This just doesn't work in practice. The stock market isn't that predictable. If it were then all investors would dive in with their algorithm that takes advantage of the predictability and BANG, suddenly the algorithm won't work any more because everyone's sentiment is the same.
When computer trading first really gained a foothold in the 80's and early 90's there was some low hanging fruit, for example people holding options forgetting to exercise them around an ex-date. New algorithms were created every month and made lots of money for a little while, then everyone else figured out the algorithm and worked their own algorithms and blew up the trend the original algorithm was working, meaning it stopped making a profit.
These days there are still these sorts of algorithms in play. All the obvious conditions (like idiots who place market orders on stocks with large spreads) are instantly gobbled by the computers to the investor's detriment. The algorithms do still work but they are immensely complex and, really, they can only skew the risk metrics just a tiny bit. Ultimately the money they make is based more on the leverage employed and less on the merits of the algorithm.
No retail investor can ever employ the amounts of leverage that companies like Goldman are able to employ. Even the leverage one can employ with margin or other borrowing is an order of magnitude too small. Retail investors have virtually no understanding of how to hedge a leveraged investment anyway and almost universally screw themselves even just using margin.
There is no magic bullet here. Thinking there is is one of the most fundamental mistakes an investor can make.
Oops, I apologize, I misread your idea. However, your idea is even worse than what I thought you were saying originally. If you only sell the losers then you only lose money in the long run.
The basic problem here is that even the best company's stocks have significant volatility. The more risk an investment has, the more volatility it tends to have (with leveraged investments having the worse volatility).
It is hard to tell the normal fluctuations of the market apart from longer term trends for your investments. So what will happen is that your stop-loss sale will hit a normal fluctuation, you will sell at a loss, then the stock will bounce back above your original sell point.
On top of that a stop-loss sale is a market order. If the market glitches the sale can wind up occurring at a much lower price than you might expect.
Stop-loss selling is virtually guaranteed to lose you money. It's a bad idea and no real investor EVER uses stop-loss orders. Not ever.
This doesn't work in practice because any investment entails a certain degree of risk for both the upside and the downside. If you cap your upside as you have described, then you are shifting most of the risk of your investment to the downside. In other-words, over time you will lose, and lose badly.
There's a secondary issue here in that you are making the assumption that the risk you carry is in that one investment but anyone who seriously invests knows that EVERYTHING in their brokerage account is an investment of one sort or another, even cash, and unless you leave most of your holdings in cash then most of your holdings at any given moment ARE going to be invested in something. It's all in managing the risk.
As an axillary point here the investment philosophy you are suggesting here is actually very similar to the covered-call options play, where you purchase an underlying stock and then write (sell) an option to someone to buy it from you for a slightly higher price in the future. There are even mutual funds that specialize in these sorts of plays but they don't have the gains you probably would expect. This sort of play works in a flat market but fails horribly in either a down market (where it doesn't mitigate enough of the damage) or an up market (where you lose out on a ton of profit potential due to the shifting of risk you undertook).
Remember that the whole point of investing in stocks is to make a better yield over a long period of time than 'safer' investments such as CDs would give you. If you invest in equities in a manner that has no chance of producing that better yield then you might as well not invest in equities at all because your other options (such as CDs) will be better deals.
The law is pointless. You don't need to carry around actual coinage. In fact, carrying around actual coinage (or keeping gold coins in your home) carries a significant risk that it will get stolen.
What you can already do (what ANYONE can already do) is simply invest in a commodity ETF like GLD... something that tracks the price of gold and is backed (by a reasonable percentage) of gold. You don't have to hold the actual stuff in your hands. You then simply sell (convert to dollars) enough to cover your monthly cash flow needs. Even more to the point you can just pay for stuff with your credit card and then settle the card (in USD) once a month by selling enough GLD to pay it off.
Problem solved. Why do people think they are chained to the U.S. Dollar? It just isn't true.
You can do this right now. It's even possible with only a modest savings (no point doing this if you don't have any savings to do it with, BTW)... to reduce transaction costs to reasonable levels.
But, you know, I think people trying to do this will find out very quickly that keeping all ones assets in a metals-backed play, particularly gold and silver, is no panacea. Value fluctuations for metals are extremely volatile relative to cash. It's a fool's errand, really. A diversified investment portfolio has much better odds of not blowing up in your face. Gold (and Silver) is subject to bubbles just like everything else, such as has occurred on multiple occasions for both metals.
Now now, it is just unfair to bring up Warcraft gold since as we all know warcraft gold actually does have an intrinsic value in the amount of human labor required to mine it (read: Chinese prisoners forced to play the game).
Even so, there seems to be an overabundance of human labor to do this activity, which coupled with gold inflation in-game has led to an epidemic of carpal-tunnel syndrome in China and very low prices in the U.S. Just 2 years ago 1000G would cost you $100 USD. Now you can get 10000G for $100 USD.
If you were a warcraft gold farmer you might even think that the U.S. dollar had appreciated significantly against the Yen for the value placed on all the hard work required. But, alas, it wouldn't be so.
I'm not actually jesting here, but I am making a point. Warcraft gold has had a steady, known inflation rate over the years that is actually less volatile than bitcoin over the same period. Inflation and volatility are two different things, and for a trading medium volatility is the bigger problem. The typical Chinese gold farmer can place a real value on his work because despite the inflation the price is still stable enough for the transitory nature of the trading medium (warcraft gold). The gold is mined and is converted to cash less than 2 days later.
There is no real price for any commodity (including any currency). That is, there is no 'standard' by which you can base all the worlds commodities and currencies that gives you a measure of value that matches your lifestyle, because everyone has a different lifestyle and it is the mix of expenses each person incurs that governs the basis upon which they feel that things are getting more costly or less costly. The value society places on commodities (and currencies) fluctuates all the time.
If you never own gold you aren't going to care what the price of gold is. If you commute a lot then the price of gas will matter a lot to you (but nevertheless you might decide to stay in a car and not buy as motorcycle to reduce your gas costs). If you don't commute then gas could be $10/gallon and you probably won't care. If you rent its one thing, if you own a home its another. And so on and so forth. Supply and demand not only dictates relative value but it also causes populations to shift usage. If gas becomes ultra expensive more people will ride public transit or buy motorcycles and fewer will buy cars. And so on.
A currency does not have to have any intrinsic value to work as a trading medium but it does have to have universal acceptance. Come on folks, doesn't anyone remember history 101? The early united states, the civil war? Every state had its own currency back then, some of which worked about as well as bitcoin does. Trust is a key component. The problem with bitcoin in a modern setting, however, is that it is competing with plenty of currencies which are highly trustworthy as a trading medium too. I really doubt bitcoin will ever gain much traction.
I suppose one could try to pay things with piles of manure instead of cash but except for the occasional farmer the volume of barter you are able to do with manure is going to be several orders of magnitude less than the volume of barter you are able to do with cash. Similarly for any commodity, even exchange traded commodities. Hey, you could hold up your phone at the store and transfer 4 shares of GLD as payment, and that's already going to be a far better currency than bitcoin will ever be, but I really doubt stores are going to fall head over heals to accommodate payments in gold.
This just isn't true. Joe average anyone can open up a brokerage account and start investing. Learning how to invest is not something that only 'The Rich' have time to do... it isn't rocket science. In the simplest terms it's the same old vice and greed lesson that every single person has to learn, just with a lot more dangling carrots to avoid. If you can discipline yourself against the vice you'll do just fine.
What you need to have is a dedication to saving some of your cash flow, not just for a vacation, not just for that occasional restaurant meal, but permanently. You can rationalize what to do with the dividends/interest/gains but if you don't save any money in the first place then you can't take advantage of the markets.
Anyone can do this, but it takes several years to really learn and understand the risks and rewards (no matter how little or how much money you have, the learning curve is the same). You don't have to trade every day or even every week... successful investors might make a few dozen trades in an entire year. Getting started rules are simple: (1) Real investors don't ever carry a credit card balance. (2) Avoid leverage like the plague. (3) Don't day-trade, and (4) Diversify. Patience is also key. Despite what you hear on T.V. the day-to-day moves of the market are irrelevant to most good investors.
(Leverage here is using things like margin, options, leveraged ETFs, highly leveraged mutual funds, and so forth. Use of margin is the #1 killer of investment portfolios).
There, I summarized how to start investing in two paragraphs. NOT rocket science.
That's a very misleading statement, you talk about the dollar as if it were a physical asset. All world currencies are trading mediums, not physical assets. All of them. Hard, soft, whatever you want to call them. The value of the trading medium itself is almost irrelevant, but at least a degree short-term stability is always desirable.
An easy way to understand this concept is to understand the term 'cash equivalence'. The term comes up all the time in financials and even the person on the street almost universally works in a cash equivalent form and not in actual dollar bills. Cash sitting in a bank account is in a cash equivalent form. Cash invested in short-term treasury bonds is in a cash equivalent form. Cash invested in a money market is in a cash equivalent form. These forms earn interest, and interest is why the 'lost 95% of its value' statement is extremely deceptive.
Now, I am not implying that the interest earned is able to make up for inflation. The interest earned from holding a cash equivalent form is usually somewhere near par, with bank savings accounts below par, money markets slightly better, treasuries slightly better, and then taking on risk to get better yields to get above the inflation rate (bonds, equities, etc).
However, cash taken out for daily activities is essentially not going to be subject to inflation because it only stays in a pure non-earning cash state for ~1-2 days. You are free to calculate the cost of ~1-2 days worth of inflation. In most economies it isn't an issue. Take Germany after world war 1 and it can become a big issue. Other economies over the years have had hyperinflation as well but it really isn't likely in the U.S. High inflation, sure (80's again). Hyperinflation, not likely.
So... you really have to think about cash as a trading medium and not a measure of absolute value. This isn't just a concept for investors... every single person needs to understand this.
How you invest your cash, whether it be in a cash equivalent form or in the form of an equity (a piece of a company), a bond, or other form, is entirely up to you. Nobody is holding a gun to people's heads and forcing them to keep their cash in bank savings accounts and money markets which (today) yield 0%. I haven't had a single penny in a savings account for 20 years and only a small percentage of my portfolio is ever in the most liquid of cash equivalent forms (a money market).
I have 1-2 months worth of cash in a cash equivalent form, sitting in my checking account ready to go out in bill-pay, and it's staged there, well, just 1-2 months at most. One can calculate 1-2 months worth of inflation and count it as the cost of doing business. Again, cash as a trading medium.
Gold is a trading medium too, by the way, as are many other commodities. There is nothing really special about trading mediums except to those people who insist to think of them as investments instead of trading mediums. On the risk scale gold is no safer than, say, throwing your money into a basket of medium-yielding stocks, and in fact gold can be considerably more volatile (remember the late 80's crash anyone?). Ditto on silver.
Ditto on ANY investment. Everything has a risk, including cash. The risk for holding cash is that '95% since...', for anyone who just stuffs it under their pillow or leaves it in a stupid savings account (though, hey, that 1927 dollar bill is probably worth something to a coin dealer!).
So what does this mean? This means when it comes to bitcoin you really have to think of it as no more than a trading medium. Inflation or deflation is irrelevant... bitcoin is not superior in any way to dollars or any other currency. What IS relevant is volatility, and I would say that single-day fluctuations in the double digit percentages makes bitcoin a non-starter for any serious investor not only as an investment but also as a trading medium.
Similarly to those folks worried about dollar devaluation, consider the fact that for your every day
The Core i5-2515E and the i7-2715QE, but I dunno re: mobo compatibility. Those are the only non-xeon Intel cpus which support ECC insofar as I know (using wiki as a reference).
AMD ECC support tends to be more a function of the BIOS, since all their cpus now support ECC. ASUS has always had good BIOSes so it is no surprise there. Gigabyte will be a bit more spotty, though I think the GA-MA770T-UD3P specifically supports it. The more popular 880GMA-UD2H's can take ECC sticks but it is unclear whether the mobo turns on the ECC and the BIOS doesn't have ECC options.
Also firmware in numerous cheap AMD mobos can be pretty horrible. The ZOTEC mobos, for example, appear to load older 3GBit/sec SATA-II firmware into the newer AMD chipsets that fully support SATA-III.
I've had a hard time finding cheap SandyBridge mobos that support ECC and SATA-III. Go ahead and newegg that. The moment you ask for 6GBit/sec SATA ports w/Intel the mobo prices shoot up. And as far as I can tell none of the low/medium-end SandyBridge mobos/cpus support ECC memory (whereas nearly all the AMD mobos do). Intel only has two non-Xeon cpus which support ECC (The i5-2515E and the i7-2715QE), and no broad mobo compatibility.
Those two issues, particularly the ECC issue, are major roadblocks for me. I fully intended to get at least one Intel i5-based box in my last set of purchases and if not for the lack of ECC memory I would have. I'm sorry, but there's just no point whatsoever throwing 16G of ram into a box without ECC, no matter how fast the cpu is.
This is more of the same from Intel. They are using their near monopoly to essentially force people into buying their higher-end (and far more expensive) cpus just to get basic features such as ECC. The artificial differentiation really pisses me off.
I run both 'server' and 'consumer' style machines. I have two xeon 2U servers and one 48-core opteron box, and a ton of consumer boxes. And you know what?
There is virtually no difference between the 'consumer' motherboards and the 'server' motherboards any more. They use virtually the same chipsets, and unless you need lots of cores (12+) for virtualization (using e.g. multi-socket opteron or xeon servers) the only real difference between a consumer box and a server box is a redundant power supply, extra fans, and a higher memory capacity.
In fact, the single-chip multi-core consumer cpus are considerably faster than their multi-chip multi-core opteron and xeon cousins, for any situation on the server where the cpu suds for an Intel-i7 or PhenomIIx6 is sufficient for the task (which is most situations). It isn't consumer-vs-server any more, it's the advances in technology that drive performance.
The integrated chipsets are no longer a differentiating factor. They ALL sport multi-gigabit interchip links and 16-32 PCIe lanes right off the core chips, and integrated memory controllers, giving even the cheapest consumer mobo 5 SATA-III ports and 30GBytes/sec of dram bandwidth.
Consumer boxes these days support 16-32G of ram (and all AMD mobos and all high-end Intel consumer mobos support ECC). Servers support 1TB+ but the harsh reality is that pretty much everything you'd need a server for these days doesn't actually need 1TB+ of memory, nor do you particularly need power-hungry redundant power supplies when the box is operating in a cluster.
We run 60G data sets out of our boxes at full network bandwidth without them breaking a sweat. I haven't found a need for 10GigE yet, 1GigE is plenty for our cost structure, the 40-120G SSD we throw into every single one of our boxes these days (using DragonFly's swapcache) offloads the (fewer) 2TB HDs we populate so much that we can push random data within their capacity out the door at 300MBytes/sec even in the cheapest consumer box with 3G of ram. In otherwords, the measily 100MBytes/sec 1GigE has has become the bottleneck (but going beyond that is driven by cost, not consumer-vs-server since PCIe can accomodate 10GigE without any issues).
So not only do we not buy high end SAS drives (whos only difference vs SATA is the command set and enclosure management, and a ridiculous price tag), we also don't buy high-RPM drives (which tend to wear out a lot more quickly). Sticking with 5600 rpm HDs and using one or more SATA-II or SATA-III SSDs as a 40-120G 'cache' is the way to go.
But I'm sure a lot of people out there still think they have to buy all the expensive stuff to get good performance and reliability. It's a fools game now, and a complete waste of money.
Based on what? You do understand that the only difference between SATA and SAS is the command set don't you? The PHY is the same. The drive internals are also the same (SAS drives aren't any more reliable than SATA drives in case you missed the memo). SAS's enclosure management is important if you are running dozens of disks but if you aren't then SAS is just an expensive footnote in the protocol. It is not any more or less reliable than SATA, and a complete waste of money.
And insofar as SSDs go, you are way, way, WAY out of date there. And, again, there is basically no difference between a consumer SSD (say, an Intel 500 series) and a supposed commercial SSD. There are differences between SLC and MLC related to durability (erase/write cycles), but at today's capacities the MLC drive provides a huge value within a calculated 10-year life span that just can't be ignored.
Well, also remember that Intel has something like 6 (or more) different incompatible cpu socket types in its lineup now, which means you have no real ability to upgrade in place.
AMD is all AM2+ and AM3, and all current cpus are AM3. This socket format has been around for several years. For example, I was able to upgrade all of my old AM2+ Phenom I boxes to Phenom II simply by replacing the cpu, and I can throw any cpu in AMD's lineup into my AM3 mobos. I only have one Phenom II x 6 machine right now but at least four of my boxes can accept that chip. That's a lot of upgrade potential on the cheap.
This will change, AMD can't stick with the AM3 form factor forever (I think the next gen will in fact change the socket), but generally speaking AMD has done a much better job on hardware longevity than Intel has. It isn't just a matter of the price of the cpu. I've saved thousands of dollars over the last few years by sticking with AMD.
SATA-III also matters a lot for a server now that SATA-III SSDs are in mass production. Now a single SSD can push 300-500 MBytes/sec of effectively random I/O out the door without having to resort to non-portable/custom-driver/premium-priced PCIe flash cards. Servers can easily keep gigabit pipes full now and are rapidly approaching 10GigE from storage all the way to the network.
The Phenom II x 6 core chips already run at 3.7 GHz when 3 or fewer cores are in use (that's what the automatic turbo feature does), so the 980's ability to run 4 cores at 3.7 GHz is only a minor improvement since it basically has no turbo mode. The x6 will win for any concurrency workloads that exercise all six cpus. Intel cpus also sport a turbo mode that works similarly.
The biggest issue w/ AMD is memory bandwidth. For some reason AMD has fallen way behind Intel in that regard. This is essentially the only reason why Intel tends to win on benchmarks.
However, you still pay a big premium for Intel, particularly Sandy-Bridge chipsets, and you pay a premium for SATA-III, whereas most AMD mobos these days already give you SATA-III @ 6GBits/sec for free. Intel knows they have the edge and they are making people pay through the nose for it.
Personally speaking the AMD Phenom II x 6 is still my favorite cpu for the price/performance and wattage consumed.
-Matt
Re:At the risk of being modded flamebait, etc
on
OpenBSD 4.9 Released
·
· Score: 5, Informative
ZFS has a large team of people behind it and resources that I don't have. That said HAMMER wasn't really designed to try to compete against it. HAMMER was designed to solve similar problems, but it wasn't designed to replace RAID as ZFS was. But ZFS is no panacea, and anyone who uses it can tell you that. The IP is now owned by Oracle, the license isn't truly open-source. ZFS itself is an extremely heavy-weight filesystem and essentially requires its ARC cache and relatively compatible workloads to work efficiently... and a veritable ton of memory.
HAMMER has a tiny footprint by comparison, gives you fine-grained automatic snapshots, and most importantly gives you near real-time queueless mirroring streams that makes creating backup topologies painless. Among many other features. Frankly ZFS might be the filesystem of choice if you are running dozens of disks but HAMMER is a much better fit otherwise.
People scream the RAID mantra all the time but the vast majority of people in the open-source world don't actually need massive RAID arrays to put together a reliable service. Often it takes just one 2TB HD and one 80G SSD x a few servers and in DragonFly HAMMER + swapcache fits that bill extremely well.
Our ultimate goal is real-time multi-master clustering. HAMMER doesn't get us quite there, primarily owing to the topology mismatch between HAMMER's B-Tree and OS filesystem cache topologies (mostly the namecache), but as the work progresses it will eventually achieve that.
In anycase, there's a huge difference between the people who do the actual design and implementation of these filesystems and the people who merely use them. Our goals as designers and programmers are not necessarily going to match the goals of the typical end-user who wants a magical black box that does everything under the sun with maximal performance in all respects and works without having to life a finger. ZFS can't even achieve that!
Well, you don't run your toaster 24x7. In fact, most residental homes use less than 1000W of power averaged 24x7 for the entire home.
Running 1000W 24x7 is ~$180-$240/month in electricty depending on where you live. Commercial power isn't much cheaper (and due to improvements in density most colo facilities now also charge for power or include only a few amps in the lowest-tier of service).
It adds up fairly quickly. The DragonFly project has 7 core production machines. Six of those in my machine room together eat around ~3.4A of power 24x7 (idle), and a lot more when they are busy. The last one is colocated and eats ~2A. There are another 2-3 essentially dedicated colocated boxes which are donated and another ~12 boxes on the third tier which donate mirroring and bandwidth. And DragonFly is a very small project.
For small projects... and here I'm not talking just about BSD projects but also many Linux projects, running your own machines requires either a fat purse somewhere or a sponser. FreeBSD gets a lot of sponsorship to help cover continuing costs.
For DragonFly we get some sponsership in the form of a few remote colocated boxes with reasonable bandwidth but mostly there are just two of us funding ongoing operations. I also fund getting ~4 new almost-bleeding-edge single-socket machines every year to keep us up-to-date on hardware and post the old boxes to various developers in need as they get replaced by new boxes, in a sort of pipeline. But it's taken 3 years to build that pipeline. New boxes come in and operate as test machines for ~1 year, then production machines for ~1-2 years, then get rotated out.
This situation has gotten a little better over the years as small projects can now run their boxes on real machines at home with a reasonable amount of upstream bandwidth, then drill a VPN through to a colocated IP service to route the IPs without having to deal with ISP filters (ISP-allocated static IPs tend not to work very well because AT&T and COMCAST's stateful filters can mess up your TCP connections when you have a lot of concurrency).
Even so it seems to me that a lot of projects don't even have that... they either rent time on a virtual machines or depend on sharing space with other larger projects. It's possible to do a lot with virtualized resources, up to a point, but rented virtualized resources tend to have very non-deterministic resources and you can wind up in trouble if you get a demand spike.
The basic mobo support for large N-way configurations has gotten cheap. Power management still has a long ways to go on these beasts, though. Our monster.dragonflybsd.org box is using the quad-socket supermicro mobo with four 12-core opterons (48 cores) and 64G of ram, and I think all told cost around $8000 or so.
The limitation for for these sorts of boxes is basically just power now. The 12-core opterons are effectively limited to 2GHz due to power issues, and these big beasts are really only high performers in environments where all the cores can be used concurrently with very little conflict.
By comparison, a PhenomII x 6 or an Intel I7 runs 6 cores for the PhenomII and 4 x 2 cores for the I7 but automatically boosts the base ~3.2GHz clock to almost 4 GHz when some of the cores are idle. These single chip solutions also have a MUCH faster path to memory than multi-chip solutions, particularly the Intel Sandybridge cpus, and much faster bus locked instructions. So if your application is only effectively using ~4-6 cores concurrently it will tend to run at least twice as fast as it would on a high-core-count monster.
That means that for most general server use a single-chip multi-core solution is what you want. The latest single-chip mobos for Intel and AMD support 16G-32G of ram and 5 or more SATA-III (6GHz) ports. Throw in a small SSD and you are suddenly able to push 400MBytes/sec+ in random-accessed file bandwidth out your network using just ONE of those SATA-III ports. That's in a desktop configuration! So today's modern desktop mobos is equivalent to last year's server mobos at 30-50% the power cost.
A modern high-end configuration as above eats ~60W idle where as the absolute minimum power draw on a 48-core Supermicro box w/ 64G of ram (the ram eating most of the power) is ~250-300W. Big difference.
So lots of cores is not necessarily going to be the best solution. In fact, probably the only really good fit for a 48+ core box is going to be for virtualization purposes.
Atomic ops are limited to 64-bits for the most part (though maybe 128 bits w/fp insns we can't really depend on that). There are several subsystems in the kernel which rely on atomic ops to test and manipulate cpu masks which would have to be reformulated.
The main issue there is one of performance. We don't want to have to use a spinlock for cases where cmpxchg solves the problem because spinlock collisions can get VERY expensive once you have more than 8 cpus in contention.
Similarly the stolen bit for the pmap spinlock (reducing the limit from 64 to 63) is there to deal with a race where one thread needs to do a SMP invtlb style operation just as a new cpu tries to switch-in a thread using the same pmap (adding another cpu to the mask of cpus that need their TLBs to be invalidated). It's a fairly rare race but it has to be dealt with properly. Also fixable with some work.
The 512GB memory limit only exists because we still populate the DMAP entries manually and it is currently hard coded for 512G (one DMAP pte). A good programmer could fix that issue in about 2 hours but we're not going to worry about it unless we actually get hardware to test with with > 512G of dram populated. That much dynamic ram is a bit beyond our budget, not to mention the 1000W+ (~8A) of power it would eat.
I get paper bills for the more important things, but use my bank's bill-pay service to pay them. I file the important pages in a binder (one or two for each year) and shred the rest.
When tax time comes around I get a second, smaller binder and I do a run-through and move tax-related items from the main binder to the smaller one. e.g. car registration, property taxes, donations, tax documents, medical bills, and so forth. The smaller binder gets shipped off to my CPA.
For services paid via credit card I carry two cards. One I use for important recurring services and the other I use for every-day purchases and on-line commerce. That way if a fraudulent charge occurs on the second card I can just close it out without having to call people up on the phone to move the recurring charges (which is a hassle). I also lock the credit limit to a lower value to reduce the card's desirability footprint in the system. This system works really well.
The main reason why I still like to receive paper bills is because it forces me to actually look at what services I am paying for and how much and to really think about whether I need the particular service or the level of service.
It's more about the chipset interface... the driver complexity is greatly reduced with a USB-3 chipset. Intel really screwed up the HCI for USB-2 and USB-1, and they barely worked even when properly implemented. The USB-3 HCI is a much cleaner design. There is this Intel commercial sporting the creater of USB being fawned at by all the woman in the office... every time I see it I feel like socking him one for doing such a bad job.
With most devices sporting wifi (let alone ethernet), fewer and fewer people need portable hard drives these days so it is almost irrelevant. It doesn't even matter that wifi is slower, since it tends to be available all the time or nearly all the time. I still use usb disk keys but my self-powered portable usb hard drives have been collecting dust for well over a year now.
USB is basically the interface for flash keys, keyboards, mice, game controllers, printers, scanners, cameras, and other odds and ends (e.g. wifi if you machine doesn't have it built in, serial ports if you still need them since most machines don't have them any more, etc). None of those really requires ultra high speed.
If you want an external drive eSATA or firewire (ignoring expensive ethernet-based network drives) are the only really reliable games in town... only someone who really really wants to lose their data uses USB as a serious hard drive interface.
I agree completely with this assessment. TRIM is no magic wand, or even close to one. You can get the equivalent simply by provisioning a little less space than the SSD reports. e.g. on a 40G factory-fresh Intel SSD I provision 32G and leave 8G untouched. For all intents and purposes that has the same effect as a heavily TRIMed SSD, and is more consistent to boot.
The performance issue with TRIM is actually related to the AHCI command set. TRIM commands cannot be natively command queued due to the massive brain damage Intel injected into the AHCI standard to differentiate it from the SAS standard. Because of that *ALL* active reads and writes have to complete and the TRIM command must be run synchronously with the drive basically quiescent. That makes TRIM a non-starter for anything but bulk clearing operations.
Basically though, Seagate seems to have made no effort to improve the quality of their drives. All manufacturers have had issues at one time or another, but when one of them stops caring about quality entirely... that is a completely different matter.
This kinda reminds me of 'cheap Chinese tools', like screwdrivers, drills, etc. The stuff sucks rocks, but is 1/4 the cost of a well-made American equivalent. They have a market because most people are willing to accept the failure rate for the price.
HD manufacturers are under similar pressures when most people who buy computers throw them away within a few years. They can make lower quality drives for cheap and only get moderate pushback from the consumer because the consumer doesn't hold onto the product long enough to notice the difference.
Those of us who run machines longer than 3 years, however, see the truth... the quality hasn't just gone down for the consumer drives, it has gone down for enterprise drives too. It has gone down across the board and THAT is a real problem. If you think about it there are only two moving parts in a hard drive: the platter and the heads, and over the years those movements and the related stresses have gone down considerably over their ancient siblings. But for all of that the drives don't last any longer.
The very same issue also impacts SSDs. Smaller-feature MLC flash has a rewrite limit approaching 3000 erase cycles and the smallest feature MLC is approaching 1000 erase cycles, while last year's larger feature MLC flash parts get closer to 10000 erase cycles. How are they able to sell such parts? Because consumer devices don't come remotely close to approach rewrite limit of older flash technologies and cost drives that market.
At least here the quality is well understood, but it means that those of us who purchase SSDs with the intent to actually use them have to be ever more careful. I absolutely refuse to purchase any SSD using flash with erase durabilities less than 10,000 cycles.
Depends what they were trying to do. If they were trying to actually run the database on the SSD through a filesystem then the row updates would likely create a very large degree of write amplification, particularly if the updates were fsync()'d on a fine grain (preventing the drive from doing any real write combining). Even aggregation wouldn't help much for random row updates. In simple terms, a 50-byte fsync'd update to a row would result in a 128KB erase/write on the SSD. With aggregation and a SSD which does proper write combining that 50-byte (random) row update would only cost ~512 bytes. Still a very high write amplification cost but nowhere near as high as 128KB:50B.
On the otherhand, just putting the database's replay log on the (MLC based) SSD and fine-tuning aggregation of the transactions to reduce write amplification effects would probably work quite well. No write combining would be needed and write amplification could probably be reduced to 4:1.
Either way SMART should have told (the original author) that the SSD was about to fail. If the SSD failed prematurely before SMART said it was going to then that's an issue with the firmware and not so much an issue with the base technology.
Slashdot Mirror
Most people don't run enough data over 3G or 4G on their smartphones to hit even the 2GB limit. I never have, and I run a lot of data over my smartphone.
The reason is simple: Your smartphone is connected up to WiFi most of the time, particularly when you are sitting still watching video (at home, at the coffee shop, etc), and not eating from the 3G/4G plate.
Of all the services you might use on the road, internet radio is probably going to be the biggest, but I really doubt it will be a problem for most people (not everyone, but most everyone). The price point for internet bandwidth has quickly moved more towards video distribution and further away from low-bandwidth and medium-bandwidth services. That trend will continue.
This is a non-issue. I think people have gotten a little too comfortable with this expectation of unlimited data. Nobody has unlimited data, not really.
-Matt
You are assuming that your internet provider will allow you to downlink at full speed 24x7, which generally isn't true. I hit up against carrier bandwidth massaging issues all the time. I've never been able to maintain full downlink rates 24x7, or anything close to full downlink rates. My uplink rate is cut in half after 10 seconds of full bandwidth use and my downlink rates are extremely variable, depending on how busy the neighborhood is.
I've also had long downlink connections cut (by the carrier's software), as in TCP RST. AT&T's U-Verse is particularly bad doing this because AT&T's routers try to track every single connection and blow up any links they've lost track of (and they lose track of links quite often).
I've been able to work around some of the issues by using a virtual network link to a nearby colo'd machine to run all my traffic through. AT&T and Comcast can no longer sniff my traffic or track connections. Comcast still caps the bandwidth, however, and AT&T's backhauls often get filled up and start dropping packets due to other traffic on the backhaul.
-Matt
This just doesn't work in practice. The stock market isn't that predictable. If it were then all investors would dive in with their algorithm that takes advantage of the predictability and BANG, suddenly the algorithm won't work any more because everyone's sentiment is the same.
When computer trading first really gained a foothold in the 80's and early 90's there was some low hanging fruit, for example people holding options forgetting to exercise them around an ex-date. New algorithms were created every month and made lots of money for a little while, then everyone else figured out the algorithm and worked their own algorithms and blew up the trend the original algorithm was working, meaning it stopped making a profit.
These days there are still these sorts of algorithms in play. All the obvious conditions (like idiots who place market orders on stocks with large spreads) are instantly gobbled by the computers to the investor's detriment. The algorithms do still work but they are immensely complex and, really, they can only skew the risk metrics just a tiny bit. Ultimately the money they make is based more on the leverage employed and less on the merits of the algorithm.
No retail investor can ever employ the amounts of leverage that companies like Goldman are able to employ. Even the leverage one can employ with margin or other borrowing is an order of magnitude too small. Retail investors have virtually no understanding of how to hedge a leveraged investment anyway and almost universally screw themselves even just using margin.
There is no magic bullet here. Thinking there is is one of the most fundamental mistakes an investor can make.
-Matt
Oops, I apologize, I misread your idea. However, your idea is even worse than what I thought you were saying originally. If you only sell the losers then you only lose money in the long run.
The basic problem here is that even the best company's stocks have significant volatility. The more risk an investment has, the more volatility it tends to have (with leveraged investments having the worse volatility).
It is hard to tell the normal fluctuations of the market apart from longer term trends for your investments. So what will happen is that your stop-loss sale will hit a normal fluctuation, you will sell at a loss, then the stock will bounce back above your original sell point.
On top of that a stop-loss sale is a market order. If the market glitches the sale can wind up occurring at a much lower price than you might expect.
Stop-loss selling is virtually guaranteed to lose you money. It's a bad idea and no real investor EVER uses stop-loss orders. Not ever.
-Matt
This doesn't work in practice because any investment entails a certain degree of risk for both the upside and the downside. If you cap your upside as you have described, then you are shifting most of the risk of your investment to the downside. In other-words, over time you will lose, and lose badly.
There's a secondary issue here in that you are making the assumption that the risk you carry is in that one investment but anyone who seriously invests knows that EVERYTHING in their brokerage account is an investment of one sort or another, even cash, and unless you leave most of your holdings in cash then most of your holdings at any given moment ARE going to be invested in something. It's all in managing the risk.
As an axillary point here the investment philosophy you are suggesting here is actually very similar to the covered-call options play, where you purchase an underlying stock and then write (sell) an option to someone to buy it from you for a slightly higher price in the future. There are even mutual funds that specialize in these sorts of plays but they don't have the gains you probably would expect. This sort of play works in a flat market but fails horribly in either a down market (where it doesn't mitigate enough of the damage) or an up market (where you lose out on a ton of profit potential due to the shifting of risk you undertook).
Remember that the whole point of investing in stocks is to make a better yield over a long period of time than 'safer' investments such as CDs would give you. If you invest in equities in a manner that has no chance of producing that better yield then you might as well not invest in equities at all because your other options (such as CDs) will be better deals.
-Matt
The law is pointless. You don't need to carry around actual coinage. In fact, carrying around actual coinage (or keeping gold coins in your home) carries a significant risk that it will get stolen.
What you can already do (what ANYONE can already do) is simply invest in a commodity ETF like GLD... something that tracks the price of gold and is backed (by a reasonable percentage) of gold. You don't have to hold the actual stuff in your hands. You then simply sell (convert to dollars) enough to cover your monthly cash flow needs. Even more to the point you can just pay for stuff with your credit card and then settle the card (in USD) once a month by selling enough GLD to pay it off.
Problem solved. Why do people think they are chained to the U.S. Dollar? It just isn't true.
You can do this right now. It's even possible with only a modest savings (no point doing this if you don't have any savings to do it with, BTW)... to reduce transaction costs to reasonable levels.
But, you know, I think people trying to do this will find out very quickly that keeping all ones assets in a metals-backed play, particularly gold and silver, is no panacea. Value fluctuations for metals are extremely volatile relative to cash. It's a fool's errand, really. A diversified investment portfolio has much better odds of not blowing up in your face. Gold (and Silver) is subject to bubbles just like everything else, such as has occurred on multiple occasions for both metals.
-Matt
Now now, it is just unfair to bring up Warcraft gold since as we all know warcraft gold actually does have an intrinsic value in the amount of human labor required to mine it (read: Chinese prisoners forced to play the game).
Even so, there seems to be an overabundance of human labor to do this activity, which coupled with gold inflation in-game has led to an epidemic of carpal-tunnel syndrome in China and very low prices in the U.S. Just 2 years ago 1000G would cost you $100 USD. Now you can get 10000G for $100 USD.
If you were a warcraft gold farmer you might even think that the U.S. dollar had appreciated significantly against the Yen for the value placed on all the hard work required. But, alas, it wouldn't be so.
I'm not actually jesting here, but I am making a point. Warcraft gold has had a steady, known inflation rate over the years that is actually less volatile than bitcoin over the same period. Inflation and volatility are two different things, and for a trading medium volatility is the bigger problem. The typical Chinese gold farmer can place a real value on his work because despite the inflation the price is still stable enough for the transitory nature of the trading medium (warcraft gold). The gold is mined and is converted to cash less than 2 days later.
-Matt
There is no real price for any commodity (including any currency). That is, there is no 'standard' by which you can base all the worlds commodities and currencies that gives you a measure of value that matches your lifestyle, because everyone has a different lifestyle and it is the mix of expenses each person incurs that governs the basis upon which they feel that things are getting more costly or less costly. The value society places on commodities (and currencies) fluctuates all the time.
If you never own gold you aren't going to care what the price of gold is. If you commute a lot then the price of gas will matter a lot to you (but nevertheless you might decide to stay in a car and not buy as motorcycle to reduce your gas costs). If you don't commute then gas could be $10/gallon and you probably won't care. If you rent its one thing, if you own a home its another. And so on and so forth. Supply and demand not only dictates relative value but it also causes populations to shift usage. If gas becomes ultra expensive more people will ride public transit or buy motorcycles and fewer will buy cars. And so on.
A currency does not have to have any intrinsic value to work as a trading medium but it does have to have universal acceptance. Come on folks, doesn't anyone remember history 101? The early united states, the civil war? Every state had its own currency back then, some of which worked about as well as bitcoin does. Trust is a key component. The problem with bitcoin in a modern setting, however, is that it is competing with plenty of currencies which are highly trustworthy as a trading medium too. I really doubt bitcoin will ever gain much traction.
I suppose one could try to pay things with piles of manure instead of cash but except for the occasional farmer the volume of barter you are able to do with manure is going to be several orders of magnitude less than the volume of barter you are able to do with cash. Similarly for any commodity, even exchange traded commodities. Hey, you could hold up your phone at the store and transfer 4 shares of GLD as payment, and that's already going to be a far better currency than bitcoin will ever be, but I really doubt stores are going to fall head over heals to accommodate payments in gold.
-Matt
This just isn't true. Joe average anyone can open up a brokerage account and start investing. Learning how to invest is not something that only 'The Rich' have time to do... it isn't rocket science. In the simplest terms it's the same old vice and greed lesson that every single person has to learn, just with a lot more dangling carrots to avoid. If you can discipline yourself against the vice you'll do just fine.
What you need to have is a dedication to saving some of your cash flow, not just for a vacation, not just for that occasional restaurant meal, but permanently. You can rationalize what to do with the dividends/interest/gains but if you don't save any money in the first place then you can't take advantage of the markets.
Anyone can do this, but it takes several years to really learn and understand the risks and rewards (no matter how little or how much money you have, the learning curve is the same). You don't have to trade every day or even every week... successful investors might make a few dozen trades in an entire year. Getting started rules are simple: (1) Real investors don't ever carry a credit card balance. (2) Avoid leverage like the plague. (3) Don't day-trade, and (4) Diversify. Patience is also key. Despite what you hear on T.V. the day-to-day moves of the market are irrelevant to most good investors.
(Leverage here is using things like margin, options, leveraged ETFs, highly leveraged mutual funds, and so forth. Use of margin is the #1 killer of investment portfolios).
There, I summarized how to start investing in two paragraphs. NOT rocket science.
-Matt
That's a very misleading statement, you talk about the dollar as if it were a physical asset. All world currencies are trading mediums, not physical assets. All of them. Hard, soft, whatever you want to call them. The value of the trading medium itself is almost irrelevant, but at least a degree short-term stability is always desirable.
An easy way to understand this concept is to understand the term 'cash equivalence'. The term comes up all the time in financials and even the person on the street almost universally works in a cash equivalent form and not in actual dollar bills. Cash sitting in a bank account is in a cash equivalent form. Cash invested in short-term treasury bonds is in a cash equivalent form. Cash invested in a money market is in a cash equivalent form. These forms earn interest, and interest is why the 'lost 95% of its value' statement is extremely deceptive.
Now, I am not implying that the interest earned is able to make up for inflation. The interest earned from holding a cash equivalent form is usually somewhere near par, with bank savings accounts below par, money markets slightly better, treasuries slightly better, and then taking on risk to get better yields to get above the inflation rate (bonds, equities, etc).
However, cash taken out for daily activities is essentially not going to be subject to inflation because it only stays in a pure non-earning cash state for ~1-2 days. You are free to calculate the cost of ~1-2 days worth of inflation. In most economies it isn't an issue. Take Germany after world war 1 and it can become a big issue. Other economies over the years have had hyperinflation as well but it really isn't likely in the U.S. High inflation, sure (80's again). Hyperinflation, not likely.
So... you really have to think about cash as a trading medium and not a measure of absolute value. This isn't just a concept for investors... every single person needs to understand this.
How you invest your cash, whether it be in a cash equivalent form or in the form of an equity (a piece of a company), a bond, or other form, is entirely up to you. Nobody is holding a gun to people's heads and forcing them to keep their cash in bank savings accounts and money markets which (today) yield 0%. I haven't had a single penny in a savings account for 20 years and only a small percentage of my portfolio is ever in the most liquid of cash equivalent forms (a money market).
I have 1-2 months worth of cash in a cash equivalent form, sitting in my checking account ready to go out in bill-pay, and it's staged there, well, just 1-2 months at most. One can calculate 1-2 months worth of inflation and count it as the cost of doing business. Again, cash as a trading medium.
Gold is a trading medium too, by the way, as are many other commodities. There is nothing really special about trading mediums except to those people who insist to think of them as investments instead of trading mediums. On the risk scale gold is no safer than, say, throwing your money into a basket of medium-yielding stocks, and in fact gold can be considerably more volatile (remember the late 80's crash anyone?). Ditto on silver.
Ditto on ANY investment. Everything has a risk, including cash. The risk for holding cash is that '95% since...', for anyone who just stuffs it under their pillow or leaves it in a stupid savings account (though, hey, that 1927 dollar bill is probably worth something to a coin dealer!).
So what does this mean? This means when it comes to bitcoin you really have to think of it as no more than a trading medium. Inflation or deflation is irrelevant... bitcoin is not superior in any way to dollars or any other currency. What IS relevant is volatility, and I would say that single-day fluctuations in the double digit percentages makes bitcoin a non-starter for any serious investor not only as an investment but also as a trading medium.
Similarly to those folks worried about dollar devaluation, consider the fact that for your every day
The Core i5-2515E and the i7-2715QE, but I dunno re: mobo compatibility. Those are the only non-xeon Intel cpus which support ECC insofar as I know (using wiki as a reference).
AMD ECC support tends to be more a function of the BIOS, since all their cpus now support ECC. ASUS has always had good BIOSes so it is no surprise there. Gigabyte will be a bit more spotty, though I think the GA-MA770T-UD3P specifically supports it. The more popular 880GMA-UD2H's can take ECC sticks but it is unclear whether the mobo turns on the ECC and the BIOS doesn't have ECC options.
Also firmware in numerous cheap AMD mobos can be pretty horrible. The ZOTEC mobos, for example, appear to load older 3GBit/sec SATA-II firmware into the newer AMD chipsets that fully support SATA-III.
-Matt
I've had a hard time finding cheap SandyBridge mobos that support ECC and SATA-III. Go ahead and newegg that. The moment you ask for 6GBit/sec SATA ports w/Intel the mobo prices shoot up. And as far as I can tell none of the low/medium-end SandyBridge mobos/cpus support ECC memory (whereas nearly all the AMD mobos do). Intel only has two non-Xeon cpus which support ECC (The i5-2515E and the i7-2715QE), and no broad mobo compatibility.
Those two issues, particularly the ECC issue, are major roadblocks for me. I fully intended to get at least one Intel i5-based box in my last set of purchases and if not for the lack of ECC memory I would have. I'm sorry, but there's just no point whatsoever throwing 16G of ram into a box without ECC, no matter how fast the cpu is.
This is more of the same from Intel. They are using their near monopoly to essentially force people into buying their higher-end (and far more expensive) cpus just to get basic features such as ECC. The artificial differentiation really pisses me off.
-Matt
I run both 'server' and 'consumer' style machines. I have two xeon 2U servers and one 48-core opteron box, and a ton of consumer boxes. And you know what?
There is virtually no difference between the 'consumer' motherboards and the 'server' motherboards any more. They use virtually the same chipsets, and unless you need lots of cores (12+) for virtualization (using e.g. multi-socket opteron or xeon servers) the only real difference between a consumer box and a server box is a redundant power supply, extra fans, and a higher memory capacity.
In fact, the single-chip multi-core consumer cpus are considerably faster than their multi-chip multi-core opteron and xeon cousins, for any situation on the server where the cpu suds for an Intel-i7 or PhenomIIx6 is sufficient for the task (which is most situations). It isn't consumer-vs-server any more, it's the advances in technology that drive performance.
The integrated chipsets are no longer a differentiating factor. They ALL sport multi-gigabit interchip links and 16-32 PCIe lanes right off the core chips, and integrated memory controllers, giving even the cheapest consumer mobo 5 SATA-III ports and 30GBytes/sec of dram bandwidth.
Consumer boxes these days support 16-32G of ram (and all AMD mobos and all high-end Intel consumer mobos support ECC). Servers support 1TB+ but the harsh reality is that pretty much everything you'd need a server for these days doesn't actually need 1TB+ of memory, nor do you particularly need power-hungry redundant power supplies when the box is operating in a cluster.
We run 60G data sets out of our boxes at full network bandwidth without them breaking a sweat. I haven't found a need for 10GigE yet, 1GigE is plenty for our cost structure, the 40-120G SSD we throw into every single one of our boxes these days (using DragonFly's swapcache) offloads the (fewer) 2TB HDs we populate so much that we can push random data within their capacity out the door at 300MBytes/sec even in the cheapest consumer box with 3G of ram. In otherwords, the measily 100MBytes/sec 1GigE has has become the bottleneck (but going beyond that is driven by cost, not consumer-vs-server since PCIe can accomodate 10GigE without any issues).
So not only do we not buy high end SAS drives (whos only difference vs SATA is the command set and enclosure management, and a ridiculous price tag), we also don't buy high-RPM drives (which tend to wear out a lot more quickly). Sticking with 5600 rpm HDs and using one or more SATA-II or SATA-III SSDs as a 40-120G 'cache' is the way to go.
But I'm sure a lot of people out there still think they have to buy all the expensive stuff to get good performance and reliability. It's a fools game now, and a complete waste of money.
-Matt
Based on what? You do understand that the only difference between SATA and SAS is the command set don't you? The PHY is the same. The drive internals are also the same (SAS drives aren't any more reliable than SATA drives in case you missed the memo). SAS's enclosure management is important if you are running dozens of disks but if you aren't then SAS is just an expensive footnote in the protocol. It is not any more or less reliable than SATA, and a complete waste of money.
And insofar as SSDs go, you are way, way, WAY out of date there. And, again, there is basically no difference between a consumer SSD (say, an Intel 500 series) and a supposed commercial SSD. There are differences between SLC and MLC related to durability (erase/write cycles), but at today's capacities the MLC drive provides a huge value within a calculated 10-year life span that just can't be ignored.
-Matt
Well, also remember that Intel has something like 6 (or more) different incompatible cpu socket types in its lineup now, which means you have no real ability to upgrade in place.
AMD is all AM2+ and AM3, and all current cpus are AM3. This socket format has been around for several years. For example, I was able to upgrade all of my old AM2+ Phenom I boxes to Phenom II simply by replacing the cpu, and I can throw any cpu in AMD's lineup into my AM3 mobos. I only have one Phenom II x 6 machine right now but at least four of my boxes can accept that chip. That's a lot of upgrade potential on the cheap.
This will change, AMD can't stick with the AM3 form factor forever (I think the next gen will in fact change the socket), but generally speaking AMD has done a much better job on hardware longevity than Intel has. It isn't just a matter of the price of the cpu. I've saved thousands of dollars over the last few years by sticking with AMD.
SATA-III also matters a lot for a server now that SATA-III SSDs are in mass production. Now a single SSD can push 300-500 MBytes/sec of effectively random I/O out the door without having to resort to non-portable/custom-driver/premium-priced PCIe flash cards. Servers can easily keep gigabit pipes full now and are rapidly approaching 10GigE from storage all the way to the network.
-Matt
The Phenom II x 6 core chips already run at 3.7 GHz when 3 or fewer cores are in use (that's what the automatic turbo feature does), so the 980's ability to run 4 cores at 3.7 GHz is only a minor improvement since it basically has no turbo mode. The x6 will win for any concurrency workloads that exercise all six cpus. Intel cpus also sport a turbo mode that works similarly.
The biggest issue w/ AMD is memory bandwidth. For some reason AMD has fallen way behind Intel in that regard. This is essentially the only reason why Intel tends to win on benchmarks.
However, you still pay a big premium for Intel, particularly Sandy-Bridge chipsets, and you pay a premium for SATA-III, whereas most AMD mobos these days already give you SATA-III @ 6GBits/sec for free. Intel knows they have the edge and they are making people pay through the nose for it.
Personally speaking the AMD Phenom II x 6 is still my favorite cpu for the price/performance and wattage consumed.
-Matt
ZFS has a large team of people behind it and resources that I don't have. That said HAMMER wasn't really designed to try to compete against it. HAMMER was designed to solve similar problems, but it wasn't designed to replace RAID as ZFS was. But ZFS is no panacea, and anyone who uses it can tell you that. The IP is now owned by Oracle, the license isn't truly open-source. ZFS itself is an extremely heavy-weight filesystem and essentially requires its ARC cache and relatively compatible workloads to work efficiently... and a veritable ton of memory.
HAMMER has a tiny footprint by comparison, gives you fine-grained automatic snapshots, and most importantly gives you near real-time queueless mirroring streams that makes creating backup topologies painless. Among many other features. Frankly ZFS might be the filesystem of choice if you are running dozens of disks but HAMMER is a much better fit otherwise.
People scream the RAID mantra all the time but the vast majority of people in the open-source world don't actually need massive RAID arrays to put together a reliable service. Often it takes just one 2TB HD and one 80G SSD x a few servers and in DragonFly HAMMER + swapcache fits that bill extremely well.
Our ultimate goal is real-time multi-master clustering. HAMMER doesn't get us quite there, primarily owing to the topology mismatch between HAMMER's B-Tree and OS filesystem cache topologies (mostly the namecache), but as the work progresses it will eventually achieve that.
In anycase, there's a huge difference between the people who do the actual design and implementation of these filesystems and the people who merely use them. Our goals as designers and programmers are not necessarily going to match the goals of the typical end-user who wants a magical black box that does everything under the sun with maximal performance in all respects and works without having to life a finger. ZFS can't even achieve that!
-Matt
Well, you don't run your toaster 24x7. In fact, most residental homes use less than 1000W of power averaged 24x7 for the entire home.
Running 1000W 24x7 is ~$180-$240/month in electricty depending on where you live. Commercial power isn't much cheaper (and due to improvements in density most colo facilities now also charge for power or include only a few amps in the lowest-tier of service).
It adds up fairly quickly. The DragonFly project has 7 core production machines. Six of those in my machine room together eat around ~3.4A of power 24x7 (idle), and a lot more when they are busy. The last one is colocated and eats ~2A. There are another 2-3 essentially dedicated colocated boxes which are donated and another ~12 boxes on the third tier which donate mirroring and bandwidth. And DragonFly is a very small project.
For small projects... and here I'm not talking just about BSD projects but also many Linux projects, running your own machines requires either a fat purse somewhere or a sponser. FreeBSD gets a lot of sponsorship to help cover continuing costs.
For DragonFly we get some sponsership in the form of a few remote colocated boxes with reasonable bandwidth but mostly there are just two of us funding ongoing operations. I also fund getting ~4 new almost-bleeding-edge single-socket machines every year to keep us up-to-date on hardware and post the old boxes to various developers in need as they get replaced by new boxes, in a sort of pipeline. But it's taken 3 years to build that pipeline. New boxes come in and operate as test machines for ~1 year, then production machines for ~1-2 years, then get rotated out.
This situation has gotten a little better over the years as small projects can now run their boxes on real machines at home with a reasonable amount of upstream bandwidth, then drill a VPN through to a colocated IP service to route the IPs without having to deal with ISP filters (ISP-allocated static IPs tend not to work very well because AT&T and COMCAST's stateful filters can mess up your TCP connections when you have a lot of concurrency).
Even so it seems to me that a lot of projects don't even have that... they either rent time on a virtual machines or depend on sharing space with other larger projects. It's possible to do a lot with virtualized resources, up to a point, but rented virtualized resources tend to have very non-deterministic resources and you can wind up in trouble if you get a demand spike.
-Matt
The basic mobo support for large N-way configurations has gotten cheap. Power management still has a long ways to go on these beasts, though. Our monster.dragonflybsd.org box is using the quad-socket supermicro mobo with four 12-core opterons (48 cores) and 64G of ram, and I think all told cost around $8000 or so.
The limitation for for these sorts of boxes is basically just power now. The 12-core opterons are effectively limited to 2GHz due to power issues, and these big beasts are really only high performers in environments where all the cores can be used concurrently with very little conflict.
By comparison, a PhenomII x 6 or an Intel I7 runs 6 cores for the PhenomII and 4 x 2 cores for the I7 but automatically boosts the base ~3.2GHz clock to almost 4 GHz when some of the cores are idle. These single chip solutions also have a MUCH faster path to memory than multi-chip solutions, particularly the Intel Sandybridge cpus, and much faster bus locked instructions. So if your application is only effectively using ~4-6 cores concurrently it will tend to run at least twice as fast as it would on a high-core-count monster.
That means that for most general server use a single-chip multi-core solution is what you want. The latest single-chip mobos for Intel and AMD support 16G-32G of ram and 5 or more SATA-III (6GHz) ports. Throw in a small SSD and you are suddenly able to push 400MBytes/sec+ in random-accessed file bandwidth out your network using just ONE of those SATA-III ports. That's in a desktop configuration! So today's modern desktop mobos is equivalent to last year's server mobos at 30-50% the power cost.
A modern high-end configuration as above eats ~60W idle where as the absolute minimum power draw on a 48-core Supermicro box w/ 64G of ram (the ram eating most of the power) is ~250-300W. Big difference.
So lots of cores is not necessarily going to be the best solution. In fact, probably the only really good fit for a 48+ core box is going to be for virtualization purposes.
-Matt
Atomic ops are limited to 64-bits for the most part (though maybe 128 bits w/fp insns we can't really depend on that). There are several subsystems in the kernel which rely on atomic ops to test and manipulate cpu masks which would have to be reformulated.
The main issue there is one of performance. We don't want to have to use a spinlock for cases where cmpxchg solves the problem because spinlock collisions can get VERY expensive once you have more than 8 cpus in contention.
Similarly the stolen bit for the pmap spinlock (reducing the limit from 64 to 63) is there to deal with a race where one thread needs to do a SMP invtlb style operation just as a new cpu tries to switch-in a thread using the same pmap (adding another cpu to the mask of cpus that need their TLBs to be invalidated). It's a fairly rare race but it has to be dealt with properly. Also fixable with some work.
The 512GB memory limit only exists because we still populate the DMAP entries manually and it is currently hard coded for 512G (one DMAP pte). A good programmer could fix that issue in about 2 hours but we're not going to worry about it unless we actually get hardware to test with with > 512G of dram populated. That much dynamic ram is a bit beyond our budget, not to mention the 1000W+ (~8A) of power it would eat.
-Matt
I get paper bills for the more important things, but use my bank's bill-pay service to pay them. I file the important pages in a binder (one or two for each year) and shred the rest.
When tax time comes around I get a second, smaller binder and I do a run-through and move tax-related items from the main binder to the smaller one. e.g. car registration, property taxes, donations, tax documents, medical bills, and so forth. The smaller binder gets shipped off to my CPA.
For services paid via credit card I carry two cards. One I use for important recurring services and the other I use for every-day purchases and on-line commerce. That way if a fraudulent charge occurs on the second card I can just close it out without having to call people up on the phone to move the recurring charges (which is a hassle). I also lock the credit limit to a lower value to reduce the card's desirability footprint in the system. This system works really well.
The main reason why I still like to receive paper bills is because it forces me to actually look at what services I am paying for and how much and to really think about whether I need the particular service or the level of service.
-Matt
It's more about the chipset interface... the driver complexity is greatly reduced with a USB-3 chipset. Intel really screwed up the HCI for USB-2 and USB-1, and they barely worked even when properly implemented. The USB-3 HCI is a much cleaner design. There is this Intel commercial sporting the creater of USB being fawned at by all the woman in the office... every time I see it I feel like socking him one for doing such a bad job.
With most devices sporting wifi (let alone ethernet), fewer and fewer people need portable hard drives these days so it is almost irrelevant. It doesn't even matter that wifi is slower, since it tends to be available all the time or nearly all the time. I still use usb disk keys but my self-powered portable usb hard drives have been collecting dust for well over a year now.
USB is basically the interface for flash keys, keyboards, mice, game controllers, printers, scanners, cameras, and other odds and ends (e.g. wifi if you machine doesn't have it built in, serial ports if you still need them since most machines don't have them any more, etc). None of those really requires ultra high speed.
If you want an external drive eSATA or firewire (ignoring expensive ethernet-based network drives) are the only really reliable games in town... only someone who really really wants to lose their data uses USB as a serious hard drive interface.
-Matt
I agree completely with this assessment. TRIM is no magic wand, or even close to one. You can get the equivalent simply by provisioning a little less space than the SSD reports. e.g. on a 40G factory-fresh Intel SSD I provision 32G and leave 8G untouched. For all intents and purposes that has the same effect as a heavily TRIMed SSD, and is more consistent to boot.
The performance issue with TRIM is actually related to the AHCI command set. TRIM commands cannot be natively command queued due to the massive brain damage Intel injected into the AHCI standard to differentiate it from the SAS standard. Because of that *ALL* active reads and writes have to complete and the TRIM command must be run synchronously with the drive basically quiescent. That makes TRIM a non-starter for anything but bulk clearing operations.
-Matt
Basically though, Seagate seems to have made no effort to improve the quality of their drives. All manufacturers have had issues at one time or another, but when one of them stops caring about quality entirely ... that is a completely different matter.
This kinda reminds me of 'cheap Chinese tools', like screwdrivers, drills, etc. The stuff sucks rocks, but is 1/4 the cost of a well-made American equivalent. They have a market because most people are willing to accept the failure rate for the price.
HD manufacturers are under similar pressures when most people who buy computers throw them away within a few years. They can make lower quality drives for cheap and only get moderate pushback from the consumer because the consumer doesn't hold onto the product long enough to notice the difference.
Those of us who run machines longer than 3 years, however, see the truth... the quality hasn't just gone down for the consumer drives, it has gone down for enterprise drives too. It has gone down across the board and THAT is a real problem. If you think about it there are only two moving parts in a hard drive: the platter and the heads, and over the years those movements and the related stresses have gone down considerably over their ancient siblings. But for all of that the drives don't last any longer.
The very same issue also impacts SSDs. Smaller-feature MLC flash has a rewrite limit approaching 3000 erase cycles and the smallest feature MLC is approaching 1000 erase cycles, while last year's larger feature MLC flash parts get closer to 10000 erase cycles. How are they able to sell such parts? Because consumer devices don't come remotely close to approach rewrite limit of older flash technologies and cost drives that market.
At least here the quality is well understood, but it means that those of us who purchase SSDs with the intent to actually use them have to be ever more careful. I absolutely refuse to purchase any SSD using flash with erase durabilities less than 10,000 cycles.
-Matt
Depends what they were trying to do. If they were trying to actually run the database on the SSD through a filesystem then the row updates would likely create a very large degree of write amplification, particularly if the updates were fsync()'d on a fine grain (preventing the drive from doing any real write combining). Even aggregation wouldn't help much for random row updates. In simple terms, a 50-byte fsync'd update to a row would result in a 128KB erase/write on the SSD. With aggregation and a SSD which does proper write combining that 50-byte (random) row update would only cost ~512 bytes. Still a very high write amplification cost but nowhere near as high as 128KB:50B.
On the otherhand, just putting the database's replay log on the (MLC based) SSD and fine-tuning aggregation of the transactions to reduce write amplification effects would probably work quite well. No write combining would be needed and write amplification could probably be reduced to 4:1.
Either way SMART should have told (the original author) that the SSD was about to fail. If the SSD failed prematurely before SMART said it was going to then that's an issue with the firmware and not so much an issue with the base technology.
-Matt