Its actually not that hard to partition web load over a number of machines. Most websites of any major size already do this. You stick a virtual IP up at the outside that round robins between a bunch of backend web servers that do the real work.
To me the harder part of it is partitioning load across servers that hold data. Figuring a way to split up your data so all important data is available to all front end machines while still keeping the backend machines within a resonable load level is difficult at best.
I really wish there were better e-mail to sms gateways for phones. I know when I e-mail my "sprint address" I don't get an sms notifcation but instead have to "check" my messages on the sprint web. Probably cheaper for sprint but a pain in the ass. I'd like to type for 20 seconds on my computer (if that) and blast it to my wifes phone. Its faster and easier than a call.
The kicker in the whole argument is that the CLI is only better if you understand the available tools very well. If you don't you will probably be more productive in a GUI because there are more hints provided by the interface on how to get the job done.
I find that if I need to do a job once and won't have to do it again I'm better off just grabbing a GUI program that does it and go. If I have to do it over and over I really prefer the convienence of a batch script or a.NET executable (which I mainly use because of its HUGE coherent central library).
If they put.NET functionality in the command line interpreter and give me access to the central.NET libraries I will be very happy. It will be nice to have a powerful command line interpreter in windows.
Re:"it can pulverize ... jelly fish"
on
Tornado in a Can
·
· Score: 1
Actually jellyfish would be harder to pulverize because its flexible and liquidy. The hard substances (like stone) bash against themselves to pulverise.
I think the article said they were trying to dehydrate the jellyfish and that it was already pureed. I think this may turn out to be a cheap dehydrator as well as a wonderful grinder.
And this kernel is worthless because all the code that runs on the series 2 tivo must pass a rigorous examination by the bios. I can't do squat with software. (hard drive upgrades are allowed but it doesn't solve the problem of wanting a relatively easy archival method)
My question is what speed PowerPC does the tivo use? If the system is 100% open and has the ease of use of the tivo I'd switch.
My only complaint about the tivo is I can't archive the videos on my computer (which means on dvd-r). I don't see how this is different than saving to a vhs tape other than the dvd-r would have a lower quality recording (tivo is pretty poor at encoding).
In the end the big bonus of this system over existing PVR solutions would be openness. I hope the openness isn't paired with low quality, hardly usable, half baked software. (sorry I'm jaded, if open source was as "good" as closed why is tivo very good, and I can't find an open PVR for my pc that works as well.)
Be careful, in my experience in a large datacenter that newer scsi disks (the 10k rpm variety) only survive a power cycle 50% of the time. By that I mean a proper power down and start up (in my case shutting down windows and then powering the machine off and then powering the machine on in a resonable amount of time). Granted I only have experience with about 40 drives and 2 different power cycles (our servers stay powered on about a year at a time).
Anyways this leads me to my actual point. Hard drives are bad in a backup situation because the most likely time for failure is spinning up or down. If you are backing up and removing, you are putting the drives through many more spinup/down cycles than they were probably designed for.
I actually think my datacenter has a process where the first teir backup is across the network to a separate and identicle server which provides "instant restoration" by a quick ip flip at the router. That second server is backed up daily or bidaily to tape. That said the tape backup is not considered long term, we don't archive the tapes, so if the tape lasts a few weeks we are happy.
I really don't know a good archival storage that is high capacity. Why can't we get some 100gig optical disks out? That would be nice.
I like the idea on a desktop where the connections per second is easily less than 1.
In the datacenter I work at we handle 2000 transactions per second per machine on average with peaks reaching 10000 transactions per second. Not every transaction requires a new connection because of caching in our software but we create far more than 1 new connection per second.
The problem that probably happens most often now is a buffer overflow deep inside. The input buffer is dynamic and checked and safe, but somewhere deep down inside the code someone decided to cast to a char array and pass the string into strcpy and bam you have a place you can easily buffer overflow (2 in fact). Its hard to catch this in review and programmers often do something like this early in development to "make it easy" promising to come back later and fix it. Its a harder problem to solve than you think, I say remove all static buffers from the C specification (and C++):)
Its still more difficult to script against windows. In unix every configuration is in a text file and the location is easy to find. In windows you need to interact with a complex configuration object model whos documentation is sparse and spread out. Its not as easy as unix by a long shot. There are some actions that I don't think you can safely do without the GUI because they chain a number events in 1 click and if you miss one of them your server is configured in a bad non-state.
As far as windows being more stable. Um no, its pretty stable but doesn't come close to matching the stability of the better *nix's that are around. Windows has more code, thus it has more bugs.
That said, the article was a cost benifit analysis of each alternative and in the end I thought the conclusion was that even with the costs associated with a move to Windows as a platform for Hotmail the benifits outweighed them. How that can be an attack on MS I don't know. Granted it makes cutting points at the OS but in the end its what was chosen.
Maybe because the "problems" listed in this article are mostly oriented to a single market (a server farm of whatever type that is staffed with competent yet independant technicians) which is not MS's main target for the OS. The article even addresses specific differences between the two and states reasons why they are different. For instance apache is compact, stable and fast and IIS is big, slow and has bugs because apache was designed to be a compact, stable and sleek web server and IIS was designed to be a huge, easily configurable web server with billions of features.
In the end I think that ALL the article says is that Hotmail chose unix initially because windows was the wrong tool. It then seemed to go on and say there were some benifits that could be gained for microsoft by moving to Windows for some of the servers at hotmail and listed those benifits (better throughput, better development tools, easier tuning).
I've never once said one OS or the other was the best tool for every job and neither does this article. I think the only important news to the/. crowd is that MS doesn't think thier OS is the best tool for every job either.
It looks like an official document to me. Not that its huge news, it just shows that MS isn't stupid even though thier marketing makes them come off as such numerous times.
The expensive part of that 3rd bin is controlling how many cases get the empty hard drive, and keeping the different computers separated and organized once they come off the line. When you are making hundreds of thousands of computers this is no small task and no small cost. Dell felt itself pretty wonderful when they showed off thier manufacuturing/inventory control system and they spend millions on it all the time, add more complexity (minor but its there) and you increase that cost.
Basically what my main point is that you will see computers with other os's (or no os) if #1 there is enough demand to justfy cost and #2 now that MS can't gouge them for doing so. Don't expect the linux or empty computer to be cheaper than the MS equivalent though.
True, the licensing agreement would be one way that the cost was kicked WAY up for selling non-ms os's. The court remedy specifically addresses this and I noted it in my discussion. I hated the fact that MS did this because I think windows would sell on network effect and merit alone. I also think that the people that don't want to use windows (I find I don't want windows on about 1/3 of my computers depending on what I use them for) should have other choices.
I hate when people here argue about the "microsoft tax" feeling that if windows wasn't bundled the computer they were purchasing would be $200 cheaper.
First off I think OEM's pay something like $15 per liscense for windows. That means if there were any discount it would only be $15 at most.
Second you need to look at the fact it costs far more to make a computer to different specs. That means a computer with no OS costs MORE to make, that means that you actually get a discount for ordering your computer the same way that a billion other users ordered it.
Last you should examine the price, how many $999 or $1199 or $1399 computers have you see on the market? You think that is a direct refelection of the cost of goods? No not really, computer companies pick numbers that look good from a marketing stanpoint when they set retail prices. So if computers all came OS free then you would see computers for the same price as when you buy windows.
With that said quit whining "microsoft tax" and realize that you pay nothing to get that bit of plastic and have the computer ready to boot. If you don't want windows its just as fast to install linux whether windows is on the machine or not.
One thing the anti-trust case will do is allow some vendors that have enough demand to start up an assembly line that installs linux as the last step instead of windows. This didn't happen before because there was a cost to the company because of microsofts anti-competative licence as well as the cost to start up the alternative assembly line to install the alternate OS. Now you just have the cost of that assembly line and we will see if that drops under the bar to get the PC's produced. I would say that if it does you will see Compaq and Dell computers with linux within 6 months. Otherwise you will know its too expensive for those companies (for one reason or another) to start installing alternate OS's and they don't think they can recoup the costs.
If this patent is proven valid it will destroy a huge number applications in the world. I don't know the specifics of the patent of course, but what is COM if not a plug-in technology? I guess the open-source zealots will love this because they will say "if you have the source you don't need plugins" but it will destroy a great deal of commercial software (btw, kernel modules are plug-ins too if I understand them right).
I think they are talking about the type of AI employed in clippy and not the microsoft implementation in particular which has one major flaw, MS didn't have the time to spend to teach clippy how to respond to intermediate and advanced users. Clippy is tuned to beginners only and thats why anyone that has half a clue is totally annoyed.
All in all the technology is good, MS's implementation is flawed (though to note it exactly filled MS's needs by reducing the number of beginner tech support calls by like 70%).
And then fixing a security exploit is a freaking nightmare (I know what of I speak as I am helping develop a.NET assembly that needs to be updated on a multitude of boxes soon) because even though you install the new assembly the app still has a copy of the old one and uses it. Then if you delete all the old assemblies your apps break because the new one isn't versioned right. Finally you figure out to use your machine.config to bind to the correct assembly and you pretty much have the same problem with assemblies as you just discussed with dll's.
My point is that its a hard problem that doesn't have an ultimate silver bullet (unless you like compiling EVERY application on your system for EVERY bug fix found or something like that).
The problem being 2-fold here. #1 I don't think the FDIC insurance protects against someone withdrawing money fraudlently in your name, and #2 the account is in PayPal's name so if paypal were to withdraw the money and use it you wouldn't get any money from the FDIC insurance.
I have read prolly 30 posts stateing this story is a dupe. Granted Lik-Sang has been back for a while but it was only very recently that they had mod-chips posted on thier site. The post on/. did a poor job of explaining this so maybe it should be pointed out.
I'm all for paying for the amount of bandwidth used as long as its clear in the contract ahead of time, and as long as there is a way to roll over unused bandwidth to the next month or get a discount for unused bandwith.
I think if a network is worried about your peak usage rather than total usage they should put a lower threshold on your bandwidth. If you are really only paying for half the bandwidth you are promised then that has to be some sort of fraud. They shouldn't be able to advertise unlimited connections when they really aren't unlimited.
I have no problem with a company deciding to cap connections in one way or another, but at least be honest in your advertising and mention that you are capped.
Seems the next logical step to use this as an input device would be to make 2-state paper that can easily return to its first state. Basically paper than you write on with the special pen and it changes to black, but you press a button and it changes back to clear. Maybe like etch-a-sketch type but with ink and static charge or something. That way the paper would be a bit more expensive but it would be reusable.
Slashdot Mirror
To me the harder part of it is partitioning load across servers that hold data. Figuring a way to split up your data so all important data is available to all front end machines while still keeping the backend machines within a resonable load level is difficult at best.
I really wish there were better e-mail to sms gateways for phones. I know when I e-mail my "sprint address" I don't get an sms notifcation but instead have to "check" my messages on the sprint web. Probably cheaper for sprint but a pain in the ass. I'd like to type for 20 seconds on my computer (if that) and blast it to my wifes phone. Its faster and easier than a call.
I find that if I need to do a job once and won't have to do it again I'm better off just grabbing a GUI program that does it and go. If I have to do it over and over I really prefer the convienence of a batch script or a .NET executable (which I mainly use because of its HUGE coherent central library).
If they put .NET functionality in the command line interpreter and give me access to the central .NET libraries I will be very happy. It will be nice to have a powerful command line interpreter in windows.
I think the article said they were trying to dehydrate the jellyfish and that it was already pureed. I think this may turn out to be a cheap dehydrator as well as a wonderful grinder.
And this kernel is worthless because all the code that runs on the series 2 tivo must pass a rigorous examination by the bios. I can't do squat with software. (hard drive upgrades are allowed but it doesn't solve the problem of wanting a relatively easy archival method)
My only complaint about the tivo is I can't archive the videos on my computer (which means on dvd-r). I don't see how this is different than saving to a vhs tape other than the dvd-r would have a lower quality recording (tivo is pretty poor at encoding).
In the end the big bonus of this system over existing PVR solutions would be openness. I hope the openness isn't paired with low quality, hardly usable, half baked software. (sorry I'm jaded, if open source was as "good" as closed why is tivo very good, and I can't find an open PVR for my pc that works as well.)
Anyways this leads me to my actual point. Hard drives are bad in a backup situation because the most likely time for failure is spinning up or down. If you are backing up and removing, you are putting the drives through many more spinup/down cycles than they were probably designed for.
I actually think my datacenter has a process where the first teir backup is across the network to a separate and identicle server which provides "instant restoration" by a quick ip flip at the router. That second server is backed up daily or bidaily to tape. That said the tape backup is not considered long term, we don't archive the tapes, so if the tape lasts a few weeks we are happy.
I really don't know a good archival storage that is high capacity. Why can't we get some 100gig optical disks out? That would be nice.
In the datacenter I work at we handle 2000 transactions per second per machine on average with peaks reaching 10000 transactions per second. Not every transaction requires a new connection because of caching in our software but we create far more than 1 new connection per second.
The problem that probably happens most often now is a buffer overflow deep inside. The input buffer is dynamic and checked and safe, but somewhere deep down inside the code someone decided to cast to a char array and pass the string into strcpy and bam you have a place you can easily buffer overflow (2 in fact). Its hard to catch this in review and programmers often do something like this early in development to "make it easy" promising to come back later and fix it. Its a harder problem to solve than you think, I say remove all static buffers from the C specification (and C++) :)
As far as windows being more stable. Um no, its pretty stable but doesn't come close to matching the stability of the better *nix's that are around. Windows has more code, thus it has more bugs.
That said, the article was a cost benifit analysis of each alternative and in the end I thought the conclusion was that even with the costs associated with a move to Windows as a platform for Hotmail the benifits outweighed them. How that can be an attack on MS I don't know. Granted it makes cutting points at the OS but in the end its what was chosen.
In the end I think that ALL the article says is that Hotmail chose unix initially because windows was the wrong tool. It then seemed to go on and say there were some benifits that could be gained for microsoft by moving to Windows for some of the servers at hotmail and listed those benifits (better throughput, better development tools, easier tuning).
I've never once said one OS or the other was the best tool for every job and neither does this article. I think the only important news to the /. crowd is that MS doesn't think thier OS is the best tool for every job either.
It looks like an official document to me. Not that its huge news, it just shows that MS isn't stupid even though thier marketing makes them come off as such numerous times.
Basically what my main point is that you will see computers with other os's (or no os) if #1 there is enough demand to justfy cost and #2 now that MS can't gouge them for doing so. Don't expect the linux or empty computer to be cheaper than the MS equivalent though.
True, the licensing agreement would be one way that the cost was kicked WAY up for selling non-ms os's. The court remedy specifically addresses this and I noted it in my discussion. I hated the fact that MS did this because I think windows would sell on network effect and merit alone. I also think that the people that don't want to use windows (I find I don't want windows on about 1/3 of my computers depending on what I use them for) should have other choices.
wasn't bundled the computer they were purchasing
would be $200 cheaper.
First off I think OEM's pay something like $15 per liscense for windows. That means if there were any discount it would only be $15 at most.
Second you need to look at the fact it costs far more to make a computer to different specs. That means a computer with no OS costs MORE to make, that means that you actually get a discount for ordering your computer the same way that a billion other users ordered it.
Last you should examine the price, how many $999 or $1199 or $1399 computers have you see on the market? You think that is a direct refelection of the cost of goods? No not really, computer companies pick numbers that look good from a marketing stanpoint when they set retail prices. So if computers all came OS free then you would see computers for the same price as when you buy windows.
With that said quit whining "microsoft tax" and realize that you pay nothing to get that bit of plastic and have the computer ready to boot. If you don't want windows its just as fast to install linux whether windows is on the machine or not.
One thing the anti-trust case will do is allow some vendors that have enough demand to start up an assembly line that installs linux as the last step instead of windows. This didn't happen before because there was a cost to the company because of microsofts anti-competative licence as well as the cost to start up the alternative assembly line to install the alternate OS. Now you just have the cost of that assembly line and we will see if that drops under the bar to get the PC's produced. I would say that if it does you will see Compaq and Dell computers with linux within 6 months. Otherwise you will know its too expensive for those companies (for one reason or another) to start installing alternate OS's and they don't think they can recoup the costs.
If this patent is proven valid it will destroy a huge number applications in the world. I don't know the specifics of the patent of course, but what is COM if not a plug-in technology? I guess the open-source zealots will love this because they will say "if you have the source you don't need plugins" but it will destroy a great deal of commercial software (btw, kernel modules are plug-ins too if I understand them right).
All in all the technology is good, MS's implementation is flawed (though to note it exactly filled MS's needs by reducing the number of beginner tech support calls by like 70%).
My point is that its a hard problem that doesn't have an ultimate silver bullet (unless you like compiling EVERY application on your system for EVERY bug fix found or something like that).
Blame /. for selling out to doubleclick, they are the ones that actually serve up the MS ads.
The problem being 2-fold here. #1 I don't think the FDIC insurance protects against someone withdrawing money fraudlently in your name, and #2 the account is in PayPal's name so if paypal were to withdraw the money and use it you wouldn't get any money from the FDIC insurance.
You can use a simple text editor to edit .NET software and compile with a command line compiler (distributed with the .NET SDK).
I have read prolly 30 posts stateing this story is a dupe. Granted Lik-Sang has been back for a while but it was only very recently that they had mod-chips posted on thier site. The post on /. did a poor job of explaining this so maybe it should be pointed out.
Because the mods are back, the first time you saw that Lik-Sang was back they had no mods.
I think if a network is worried about your peak usage rather than total usage they should put a lower threshold on your bandwidth. If you are really only paying for half the bandwidth you are promised then that has to be some sort of fraud. They shouldn't be able to advertise unlimited connections when they really aren't unlimited.
I have no problem with a company deciding to cap connections in one way or another, but at least be honest in your advertising and mention that you are capped.
Seems the next logical step to use this as an input device would be to make 2-state paper that can easily return to its first state. Basically paper than you write on with the special pen and it changes to black, but you press a button and it changes back to clear. Maybe like etch-a-sketch type but with ink and static charge or something. That way the paper would be a bit more expensive but it would be reusable.