I broke Oracle8i on Solaris 7 (patched) and Veritas LVM 3.0.x this past summer on a Netra t1405. I can dig up the script that provoked it, maybe, if I'm lucky. I still trust Oracle pretty well, though. It may have been a Solaris issue, but that would surprise me. I've only heard of one other coredump with Oracle 8.x.x on a primary development platform (Solaris and maybe HP-UX?).
I didn't actually break Informix; one of my friends working as a DBA for Cornell broke that in production, was corroborated, and that's good enough for me. That was on AIX. (Are they still in business? Their webpage is up but that's not really a perfect indicator of solvency.)
Sybase was on Linux. I'll admit, I cheated a bit there. I'd be happy to try and break DB2 UDB on something like an S/390 G6 but I don't have one of those lying around...
What do you pay for? Well, mostly you're paying to see your pillow more often, or at least that's what the vendor is trying to sell you -- peace of mind.
I have (lately -- SQL 7, not just SQL 6.5) seen MS SQL databases that buy the farm every night and get rebooted, and this is referred to as "more stable than before". I've watched MySQL crumble under heavy concurrent read/write loads on large tables. I still haven't figured out an easy way to carve up PostgreSQL backend storage so it scales more like Oracle. And I've thus far managed to make every (yes, EVERY) database I've ever used core dump. I'm working on DB2 at the moment; apparently it runs across x86 Linux clusters, which of course is something I'd like to crash. I got Oracle 8i to crash (finally!) earlier this summer. That was unusually difficult; Sybase and Informix were relatively easier. So, when you go the commercial route, you're paying for features, branding, and a lot of stuff you already get from the free RDBMS packages. Plus, more people tend to know about the bugs (every last RDBMS I have seen has at least some bugs).
BUT when you get right down to it, even Postgresql (my favorite free database) simply does not have the application support or PROVEN OLTP scalability of something like Oracle. Today someone batted around the idea that "DB2 is okay up to around 10 or 15TB, but I talked to some banks and oil companies, and they say that beyond that, it just falls apart". It may be the case that Oracle scales a little better than DB2, but up until recently, the top 10 (size-wise) OLTP databases on earth were running NCR Teradata, IBM DB2, and I think there may have been one Oracle customer in there. UPS's database, the largest active OLTP database, was only around 12TB last time I looked.
12TB of live data is a HUGE amount. EMC's top-of-the-line Symmetrix, a piece of equipment costing a cool $3.2M, just barely fits 12TB if you're running everything mirrored. God forbid you blow out the cache on that monster. And of course there's the question of backups...
Basically, if you're tied into a vendor's specs (eg. "We only support MS-SQL here at Affymetrix" or "ArsDigita only supports Oracle as our ACS backend") then you don't have to choose, just get out the wallet and bend over. No lube for you!
Otherwise, in my neither limited nor hugely abundant experience, MySQL and Postgres work pretty damn well for read-only data marts and low- to mid-volume read-write backends.
Beyond that -- if MySQL can't handle your OLAP needs, and Postgresql can't handle your OLTP needs -- then maybe you have to look at Oracle or DB2 or whatever. But by then I strongly suspect you'll know why they charge so %$@#ing much:-).
The differences are fairly marginal, but if you need them, you need them, and that's the end of it.
The situations where Linux beats up on FreeBSD are rare. I know. I've seen a couple of them in live production. Most people won't. Ever.
You are exaggerating the importance of SMP, I think. Yes, 2.4 (and even 2.2) have Real SMP support, as opposed to FreeBSD 4 and earlier's "one processor for kernel space, one for user space" approach. (If that's not the God's honest truth, then it's somewhat pathetic that a quad-CPU box running threaded apps at 4.0 utilization only utilized two processors in the deployment I'm thinking of) Most users won't have to care about this if they're running x86 hardware. I have seen very few sites that needed scalable SMP and couldn't afford a Sun or Sequent box at the heart of their business. (and I simply don't trust Intel hardware for applications where I can only afford to spec a single unit in production; for clustering and server farms, it's fine)
Moreover, FreeBSD 5's SMP support is likely to be on a par with 2.4's. That's pretty damn good. I haven't seen many commodity 8-way Intel boxes, and I've never seen any non-Sequent 64-way x86 machines. Ever. (I've used plenty of the rest)
On dual-processor machines, especially webservers, I have watched FreeBSD kick the crap out of Linux. On quads, Linux does indeed beat up on FreeBSD. But most people are not purchasing 4- or 8-way chassis with an eye to expansion, the way people do with Suns (eg. 4500's with a single processor board, etc).
What worries me is that the same people who consider your post 'informative' are the ones most likely to come to the wrong conclusion.
Oh well. Anyone who believes everything they read on Slashdot shouldn't be making purchasing decisions anyways...
If you haven't figured it out, I'll let you in on a secret: the various BSD distributions excel at non-desktop uses. Unless you really *ARE* an 31337 h4x0r, there is not much point in running them on the desktop. Right now I'm using Windows on my desktop, in fact (although this is primarily because my Thinkpad's screen is hosed -- I do prefer Linux).
Meanwhile, if you run Linux in a very high-load, high-availability environment, I can almost guarantee that you will have more little problems than with FreeBSD. When every little problem turns into a page or an email in the middle of the night, you have to become a little wary of
Linux. I like Linux. I hack on Linux. I sure as hell don't run web server farms or firewalls on Linux. Choose the right tool for the job.
Your argument about books in the local bookstore is also spurious. It indicates that your bookstore isn't very good. Look for 'The Design and Implementation of the 4.4 BSD Operating System' or 'Building Linux and OpenBSD Firewalls'. All of these are well written and illustrate what's going on. I'll go so far as to say they do it better than any of the Linux books do. Look in the canonical Unix sysadmin's handbook, the purple 'Unix System Administrator's Handbook, 3rd edition'. What representative OSes have they chosen? FreeBSD, Linux, HP-UX, and Solaris. That's pretty mainstream, dude.
I don't think you're wrong to use Linux instead of *BSD on the desktop. I think you're wrong to compare them at all. BSD is for servers.
The Win2K kernel (or NT for that matter) is hardly less complex than Linux. If anything, it is a great deal more complex. This book addresses a need which simply isn't present for NT/W2K installations, because 99.9% of them don't have source and can't change anything internal to their OS. Linux is very different, and you don't seem to understand why. This isn't a book for end users.
If you are looking for a book that compares various OSes, something by Silberschatz or Stallings might be more your speed. I don't really find those authors as helpful in my work, though. Tanenbaum (who influenced Linus' initial design) wrote one called "Modern Operating Systems" which is pretty interesting; it may be the best of that genre. The BSD book and the Linux book are not comparable to Tanenbaum's book (or for that matter, to each other, really).
I own both books (and Tanenbaum's) and I have to admit that I find the BSD book most useful. I think that's because BSD was more "designed" and Linux is more a "big ball of mud" that just got to be how it is by evolution. The types of bugs I've seen over the years in Linux seem to reflect that, while the kernel and VFS design is quite clever (in an engineering sense), there are aspects pertaining to networking and filesystems (most notably) that only get addressed when they become a problem to someone who is willing to solve them (eg. Direct Server Return in 2.2.14). I'm not saying there is no design to Linux, but it is less pronounced than in *BSD. I do *NOT* think that ESR's Cathedral and Bazaar analogy is as apt as many other people seem to believe. Maybe it's just because BSD is easier to understand for me. And I work with it a lot. (and Linux... and Solaris... and NT/2K...) On the other hand, there is a much greater level of detail in Bovet and Cesati's book, line-by-line analysis of much of the kernel's data structures, which I am finding useful in hacking on it.
My bias:
I am an applications programmer, sysadmin, and network administrator, rather than a kernel developer or embedded systems programmer. So I am more interested in network and filesystem details than how certain atomic operations are implemented. I occasionally change little things here and there in the Linux kernel (almost never changing actual code semantics in any of the BSDs), but nothing earth shattering in either case. On the other hand, it is my job to *tune* kernels.
My comparison:
The Linux book seems to concentrate on internal data structures, how pieces of code are implemented, and why; more a tinkering than design perspective. The BSD book is a design-motivating-an-implementation book (as per the title). They're both useful books, but for very different reasons. At only $40, it's worth buying the Bovet and Cesati book for guidance if you hack on Linux AT ALL, especially on a laptop or a disconnected workstation (hopefully with all the HOWTOs and source installed on it!).
When I worked at NBCi, we ran a benchmark of our most heavily trafficked application on Linux, Solaris, and FreeBSD. (this is the part that used to be Xoom, FWIW -- around 3-4 million users per day across 32 servers) FreeBSD was about 20% faster than Linux and marginally faster than the more expensive (per-box) Solaris setup once we had them all tuned. Shockingly, FreeBSD 2.2.8 (tuned all to hell) was a little faster (2% - 4%) than FreeBSD 4.0 or 3.2 -- not sure why.
Anyways, Yahoo uses FreeBSD, we used FreeBSD at NBCi/XOOM, many, many ISPs and ASPs use FreeBSD, and Apple now uses FreeBSD, after a fashion...
The only people I know that don't like FreeBSD, especially after being confronted with hard evidence for its networking prowess, are folks who are unfamiliar with BSD internals. To a man, they've all indicated to me that they're willing to take the performance hit because they "know" Linux better. And I can respect that. But the tone of your article seems to imply that FreeBSD users are few and far between, and that's just bullshit. Give it a try sometime, you might discover that you'd prefer it on your web and mail servers yourself!
Dude, use Emacs and learn what C-x a-i-l means. It turns on abbreviations. When I was writing servlets I never typed more than three
characters to call a method -- I had autoexpansion on.
While you're at it, learn to use the speedbar and autogeneration of skeleton methods in JDE.
Too bad you can't tune any kernels in Java, or I'd probably still be working in it. (and Perl)
I use Vi all day long so, when XEmacs starts up, I have it default to Vi keybindings. There goes another one of your arguments. Oh, and use gnuclient -- start up XEmacs when you start up X and simply run client processes. It's as fast as Vi once it's loaded into memory, and you can use a real debugger (DDD, jdb, pdb, gdb) , make, compile, all that IDE shit... from your editor.
Dude, use Emacs and learn what C-x a-i-l means. It turns on abbreviations. When I was writing servlets I never typed more than three characters to call a method -- I had autoexpansion on.
While you're at it, learn to use the speedbar and autogeneration of skeleton methods in JDE.
Too bad you can't tune any kernels in Java, or I'd probably still be working in it. (and Perl)
This gets right at the heart of why good Oracle DBAs often pull salaries in excess of $150K. At the end of the day, speed and reliability count for a lot more than other peoples' preconceptions.
The Safe Bet: Qmail + mutt + OpenSSH + OpenBSD (+ djbdns if you want DIY DNS service). It would be hard to find a more reliable, secure setup. Not the absolute friendliest, but solid as a rock.
People seem to overlook qmail when setting up a reliable, secure system. Having dealt with Sendmail and Qmail, I would suggest the latter to anyone who cares about security or performance. The same logic applies to BIND vs. djbdns.
When you work from home, for example. As a sysadmin and programmer, it happens plenty. My solution for some time now has been to collect email from various (not publicly available) addresses into an account which I ssh to (as do other users on the box) and read mail at my leisure. I don't engage in any activities nefarious to be more paranoid than that anymore (no gun running, drug manufacture, or espionage, for example). I occasionally chat with people from competing companies or fix up someone's resume, and once in a while I might flame someone.
Basically, I wouldn't work for an employer who was so paranoid that this arrangement made me nervous, and I would encourage others to consider whether they should. I'm a fairly decent systems programmer and administrator, but I don't believe that my leverage with my employers is excessive. On the other hand, I also don't try to rip off my employers or do a substandard job, which sometimes seems like apostasy in modern-day working America, so YMMV.
Uh, yeah, it's kind of hard not to realize that... however, there is a difference between building 1200 identical servers (OS and software) and 1200 servers of various sorts (OS and software). All that needs be done for a Googlecluster is a round of DHCP, see which boxes didn't get an image, try them again, and toss out the stragglers.
Duh. This is different from doing a kickstart or jumpstart install... it's all just one simple boot image to a bunch of boxes. But building 1200 in an afternoon is still pretty cool.
Google is (also?) colocated at Exodus and GlobalCenter; I just walked by their Borregas cage, in fact. The Rackables are now up into the thousands of CPUs (a friend of mine who works at Google built 1200 machines in a single afternoon; yes, they're identical, yes, it's done by DHCP and on a private VLAN). Finally, the chef doesn't cost much relative to their programmers (mostly Stanford and UCSB almost-PhD's).
Google is hella cool and both of my friends that work there, love it. They do work pretty hard, though. Still, when you love what you do...
Postgresql may not have outer joins, but at least its developers and copyright holders have respect for their hackers, users, and corporate partners.
I was thinking of getting into Interbase. This certainly dampened my enthusiasm. I will continue to work on Oracle for high-end (really high-end) work, use PostgreSQL for development, and deploy MySQL for data marts. Watching Informix lay off something like 900 people recently, and Sybase clinging to life, I wasn't too surprised at Interbase getting to be a dead end (for now).
Such a drag. The greedheads and day-trading degenerates seem to infect everything these days.
Why on earth is anyone surprised that a company whose first successful product was the 4.4BSD documentation, and whose headquarters is but a wee bit up the bay from Berkeley, should belly up to the table with a concise handbook to one of the most useful operating systems of our time?
Oh wait, it isn't GPLed, it must not be any good...;-)
Legislation can be passed to prevent this sort of BSD-ish corporate cannibalization. Whether or not Celera invested millions into their research is irrelevant in the court of public opinion, where some enterprising senator will likely spin the issue as 'selling humanity' and nail Celera's more nefarious profit motives to the wall.
It's too easy not to do it (start up a backlash against patenting parts of people, as it were). The religious right will be all over it like a rash -- for once I happen to be pleased that they're a force.
Venter is a great guy, and an engaging public speaker, but he has many reasons to be bitter towards the established academic community of molecular genetics. And that's why I don't trust Celera to resist compromising Venter's stated principles in pursuit of profits.
I do not work for Yahoo and cannot speak for them. Don't work for Google either, although one of my friends now does. You'd probably be surprised to find out who #7 on the web is -- that's us. (MediaMetrix numbers -- we bounce around slightly above Amazon and below Lycos most months) I'm not sure that senior management would be real pleased with me bashing Sun (or MS, for that matter) so I'll leave it to interested parties to figure out the rest. I simply use a Yahoo! email address because I like their spam filtering services.
Yahoo has a long history of avoiding complexity wherever possible, and BSD (along with flat files, manual indexing, and their pile-o-netapps setups) fits in with this strategy. They're cool cats.
Google takes a different approach. They take Linux, hack it up a bit, and scale like you would not believe. My friend who works there called me up a couple weeks ago asking about using IP multicasting to broadcast boot images. That's pretty sick -- think on it for a moment. I hope the stock market treats them as well as they deserve to make out -- we tried to buy them and they told us to go to hell (they want to IPO).
I don't know why Google likes Linux, but between them and Yahoo!, both OSes should be healthy for a long time to come. eg. Yahoo is big enough to (apparently) get Oracle to do a build on FreeBSD, although I can't confirm that (just something one of the Walnut Creek guys passed along way back).
If you don't need FreeBSD's slight performance edge, or need something in the Linux kernel that isn't supported by BSD, then you should use Linux. FreeBSD is just better for networking and I/O, really. But that covers a lot of ground...
Also, I've seen some very strange things happen when using Linux on loadbalanced servers doing DSR. FreeBSD boxes in the same setup have always worked great. Just another minor quibble.
1) the free-er BSD license -- do whatever you want with the code, just give credit where it's due
2) superior networking performance -- I'm an admin for one of the 10 largest sites on the web. We have large numbers of both FreeBSD and Linux (and Solaris) boxes in production. To be perfectly honest, FreeBSD is the fastest, most reliable, most configurable OS I have encountered so far. I run Linux at home for no particular reason other than I'm too lazy to switch; however I have found that FreeBSD offers so many compelling advantages as a single-purpose server that I deploy mostly FreeBSD boxes where there is not a compelling application forcing us to use Linux or Solaris.
And I'll probably switch over to FreeBSD at home, too, maybe waiting until 4.1 comes out to be safe. On laptops and such it makes sense to run Linux; on a firewall, OpenBSD. But for maximum performance on web and mail servers, FreeBSD rocks the house. Don't take my word for it -- try it yourself. Many of the Apache and Qmail developers run FreeBSD as their primary platform, and that's not an easy bunch to impress.
FreeBSD is to Linux as Postgres is to MySQL;-). (which would bring us next to Solaris -> Oracle)
I'm not sure why you used the analogy you did. In the event of a natural disaster, a piece of Big Iron is just as fallible as a PC.
Which is one reason IBM sells clustering solutions for just about everything they make.
This makes me curious -- what would happen if the root A server got totalled? What gets failed over onto? I know I should RTFM, and I will, but my Stevens books are at home.
Slashdot Mirror
I broke Oracle8i on Solaris 7 (patched) and Veritas LVM 3.0.x this past summer on a Netra t1405. I can dig up the script that provoked it, maybe, if I'm lucky. I still trust Oracle pretty well, though. It may have been a Solaris issue, but that would surprise me. I've only heard of one other coredump with Oracle 8.x.x on a primary development platform (Solaris and maybe HP-UX?).
I didn't actually break Informix; one of my friends working as a DBA for Cornell broke that in production, was corroborated, and that's good enough for me. That was on AIX. (Are they still in business? Their webpage is up but that's not really a perfect indicator of solvency.)
Sybase was on Linux. I'll admit, I cheated a bit there. I'd be happy to try and break DB2 UDB on something like an S/390 G6 but I don't have one of those lying around...
What do you pay for? Well, mostly you're paying to see your pillow more often, or at least that's what the vendor is trying to sell you -- peace of mind.
:-).
I have (lately -- SQL 7, not just SQL 6.5) seen MS SQL databases that buy the farm every night and get rebooted, and this is referred to as "more stable than before". I've watched MySQL crumble under heavy concurrent read/write loads on large tables. I still haven't figured out an easy way to carve up PostgreSQL backend storage so it scales more like Oracle. And I've thus far managed to make every (yes, EVERY) database I've ever used core dump. I'm working on DB2 at the moment; apparently it runs across x86 Linux clusters, which of course is something I'd like to crash. I got Oracle 8i to crash (finally!) earlier this summer. That was unusually difficult; Sybase and Informix were relatively easier. So, when you go the commercial route, you're paying for features, branding, and a lot of stuff you already get from the free RDBMS packages. Plus, more people tend to know about the bugs (every last RDBMS I have seen has at least some bugs).
BUT when you get right down to it, even Postgresql (my favorite free database) simply does not have the application support or PROVEN OLTP scalability of something like Oracle. Today someone batted around the idea that "DB2 is okay up to around 10 or 15TB, but I talked to some banks and oil companies, and they say that beyond that, it just falls apart". It may be the case that Oracle scales a little better than DB2, but up until recently, the top 10 (size-wise) OLTP databases on earth were running NCR Teradata, IBM DB2, and I think there may have been one Oracle customer in there. UPS's database, the largest active OLTP database, was only around 12TB last time I looked.
12TB of live data is a HUGE amount. EMC's top-of-the-line Symmetrix, a piece of equipment costing a cool $3.2M, just barely fits 12TB if you're running everything mirrored. God forbid you blow out the cache on that monster. And of course there's the question of backups...
Basically, if you're tied into a vendor's specs (eg. "We only support MS-SQL here at Affymetrix" or "ArsDigita only supports Oracle as our ACS backend") then you don't have to choose, just get out the wallet and bend over. No lube for you!
Otherwise, in my neither limited nor hugely abundant experience, MySQL and Postgres work pretty damn well for read-only data marts and low- to mid-volume read-write backends.
Beyond that -- if MySQL can't handle your OLAP needs, and Postgresql can't handle your OLTP needs -- then maybe you have to look at Oracle or DB2 or whatever. But by then I strongly suspect you'll know why they charge so %$@#ing much
The differences are fairly marginal, but if you need them, you need them, and that's the end of it.
apparently it was JFK. oops... pot, meet kettle.
FDR spouted the "only thing we have to fear" line.
said that. (your sig)
Christ, have some class.
Was that a joke?
The situations where Linux beats up on FreeBSD are rare. I know. I've seen a couple of them in live production. Most people won't. Ever.
You are exaggerating the importance of SMP, I think. Yes, 2.4 (and even 2.2) have Real SMP support, as opposed to FreeBSD 4 and earlier's "one processor for kernel space, one for user space" approach. (If that's not the God's honest truth, then it's somewhat pathetic that a quad-CPU box running threaded apps at 4.0 utilization only utilized two processors in the deployment I'm thinking of) Most users won't have to care about this if they're running x86 hardware. I have seen very few sites that needed scalable SMP and couldn't afford a Sun or Sequent box at the heart of their business. (and I simply don't trust Intel hardware for applications where I can only afford to spec a single unit in production; for clustering and server farms, it's fine)
Moreover, FreeBSD 5's SMP support is likely to be on a par with 2.4's. That's pretty damn good. I haven't seen many commodity 8-way Intel boxes, and I've never seen any non-Sequent 64-way x86 machines. Ever. (I've used plenty of the rest)
On dual-processor machines, especially webservers, I have watched FreeBSD kick the crap out of Linux. On quads, Linux does indeed beat up on FreeBSD. But most people are not purchasing 4- or 8-way chassis with an eye to expansion, the way people do with Suns (eg. 4500's with a single processor board, etc).
What worries me is that the same people who consider your post 'informative' are the ones most likely to come to the wrong conclusion.
Oh well. Anyone who believes everything they read on Slashdot shouldn't be making purchasing decisions anyways...
Come on already.
If you haven't figured it out, I'll let you in on a secret: the various BSD distributions excel at non-desktop uses. Unless you really *ARE* an 31337 h4x0r, there is not much point in running them on the desktop. Right now I'm using Windows on my desktop, in fact (although this is primarily because my Thinkpad's screen is hosed -- I do prefer Linux).
Meanwhile, if you run Linux in a very high-load, high-availability environment, I can almost guarantee that you will have more little problems than with FreeBSD. When every little problem turns into a page or an email in the middle of the night, you have to become a little wary of
Linux. I like Linux. I hack on Linux. I sure as hell don't run web server farms or firewalls on Linux. Choose the right tool for the job.
Your argument about books in the local bookstore is also spurious. It indicates that your bookstore isn't very good. Look for 'The Design and Implementation of the 4.4 BSD Operating System' or 'Building Linux and OpenBSD Firewalls'. All of these are well written and illustrate what's going on. I'll go so far as to say they do it better than any of the Linux books do. Look in the canonical Unix sysadmin's handbook, the purple 'Unix System Administrator's Handbook, 3rd edition'. What representative OSes have they chosen? FreeBSD, Linux, HP-UX, and Solaris. That's pretty mainstream, dude.
I don't think you're wrong to use Linux instead of *BSD on the desktop. I think you're wrong to compare them at all. BSD is for servers.
The Win2K kernel (or NT for that matter) is hardly less complex than Linux. If anything, it is a great deal more complex. This book addresses a need which simply isn't present for NT/W2K installations, because 99.9% of them don't have source and can't change anything internal to their OS. Linux is very different, and you don't seem to understand why. This isn't a book for end users.
If you are looking for a book that compares various OSes, something by Silberschatz or Stallings might be more your speed. I don't really find those authors as helpful in my work, though. Tanenbaum (who influenced Linus' initial design) wrote one called "Modern Operating Systems" which is pretty interesting; it may be the best of that genre. The BSD book and the Linux book are not comparable to Tanenbaum's book (or for that matter, to each other, really).
I own both books (and Tanenbaum's) and I have to admit that I find the BSD book most useful. I think that's because BSD was more "designed" and Linux is more a "big ball of mud" that just got to be how it is by evolution. The types of bugs I've seen over the years in Linux seem to reflect that, while the kernel and VFS design is quite clever (in an engineering sense), there are aspects pertaining to networking and filesystems (most notably) that only get addressed when they become a problem to someone who is willing to solve them (eg. Direct Server Return in 2.2.14). I'm not saying there is no design to Linux, but it is less pronounced than in *BSD. I do *NOT* think that ESR's Cathedral and Bazaar analogy is as apt as many other people seem to believe. Maybe it's just because BSD is easier to understand for me. And I work with it a lot. (and Linux... and Solaris... and NT/2K...) On the other hand, there is a much greater level of detail in Bovet and Cesati's book, line-by-line analysis of much of the kernel's data structures, which I am finding useful in hacking on it.
My bias:
I am an applications programmer, sysadmin, and network administrator, rather than a kernel developer or embedded systems programmer. So I am more interested in network and filesystem details than how certain atomic operations are implemented. I occasionally change little things here and there in the Linux kernel (almost never changing actual code semantics in any of the BSDs), but nothing earth shattering in either case. On the other hand, it is my job to *tune* kernels.
My comparison:
The Linux book seems to concentrate on internal data structures, how pieces of code are implemented, and why; more a tinkering than design perspective. The BSD book is a design-motivating-an-implementation book (as per the title). They're both useful books, but for very different reasons. At only $40, it's worth buying the Bovet and Cesati book for guidance if you hack on Linux AT ALL, especially on a laptop or a disconnected workstation (hopefully with all the HOWTOs and source installed on it!).
When I worked at NBCi, we ran a benchmark of our most heavily trafficked application on Linux, Solaris, and FreeBSD. (this is the part that used to be Xoom, FWIW -- around 3-4 million users per day across 32 servers) FreeBSD was about 20% faster than Linux and marginally faster than the more expensive (per-box) Solaris setup once we had them all tuned. Shockingly, FreeBSD 2.2.8 (tuned all to hell) was a little faster (2% - 4%) than FreeBSD 4.0 or 3.2 -- not sure why.
Anyways, Yahoo uses FreeBSD, we used FreeBSD at NBCi/XOOM, many, many ISPs and ASPs use FreeBSD, and Apple now uses FreeBSD, after a fashion...
The only people I know that don't like FreeBSD, especially after being confronted with hard evidence for its networking prowess, are folks who are unfamiliar with BSD internals. To a man, they've all indicated to me that they're willing to take the performance hit because they "know" Linux better. And I can respect that. But the tone of your article seems to imply that FreeBSD users are few and far between, and that's just bullshit. Give it a try sometime, you might discover that you'd prefer it on your web and mail servers yourself!
Dude, use Emacs and learn what C-x a-i-l means. It turns on abbreviations. When I was writing servlets I never typed more than three
characters to call a method -- I had autoexpansion on.
While you're at it, learn to use the speedbar and autogeneration of skeleton methods in JDE.
Too bad you can't tune any kernels in Java, or I'd probably still be working in it. (and Perl)
I use Vi all day long so, when XEmacs starts up, I have it default to Vi keybindings. There goes another one of your arguments. Oh, and use gnuclient -- start up XEmacs when you start up X and simply run client processes. It's as fast as Vi once it's loaded into memory, and you can use a real debugger (DDD, jdb, pdb, gdb) , make, compile, all that IDE shit... from your editor.
Dude, use Emacs and learn what C-x a-i-l means. It turns on abbreviations. When I was writing servlets I never typed more than three characters to call a method -- I had autoexpansion on.
While you're at it, learn to use the speedbar and autogeneration of skeleton methods in JDE.
Too bad you can't tune any kernels in Java, or I'd probably still be working in it. (and Perl)
This gets right at the heart of why good Oracle DBAs often pull salaries in excess of $150K. At the end of the day, speed and reliability count for a lot more than other peoples' preconceptions.
and start beta testing it. What's that? There isn't one? Well, shit, let's forget about Linux then.
Oh wait, who's paying for your servers and bandwidth? VA Linux you say? Hmm, that's odd.
Relevant URLs: .muttrc doesn't hurt either.
Dan Bernstein's page. Home of Qmail and djbdns.
The OpenBSD and OpenSSH home pages are full of useful information.
PuTTY, a free Windows SSH client Great for on road trips, internet cafe's, consulting, etc.
Mutt, the One True mail client. Takes some getting used to, a good
People seem to overlook qmail when setting up a reliable, secure system. Having dealt with Sendmail and Qmail, I would suggest the latter to anyone who cares about security or performance. The same logic applies to BIND vs. djbdns.
It's a /24 for your stereo. Good god, man.
When you work from home, for example. As a sysadmin and programmer, it happens plenty. My solution for some time now has been to collect email from various (not publicly available) addresses into an account which I ssh to (as do other users on the box) and read mail at my leisure. I don't engage in any activities nefarious to be more paranoid than that anymore (no gun running, drug manufacture, or espionage, for example). I occasionally chat with people from competing companies or fix up someone's resume, and once in a while I might flame someone.
Basically, I wouldn't work for an employer who was so paranoid that this arrangement made me nervous, and I would encourage others to consider whether they should. I'm a fairly decent systems programmer and administrator, but I don't believe that my leverage with my employers is excessive. On the other hand, I also don't try to rip off my employers or do a substandard job, which sometimes seems like apostasy in modern-day working America, so YMMV.
Uh, yeah, it's kind of hard not to realize that... however, there is a difference between building 1200 identical servers (OS and software) and 1200 servers of various sorts (OS and software). All that needs be done for a Googlecluster is a round of DHCP, see which boxes didn't get an image, try them again, and toss out the stragglers.
Duh. This is different from doing a kickstart or jumpstart install... it's all just one simple boot image to a bunch of boxes. But building 1200 in an afternoon is still pretty cool.
Google is (also?) colocated at Exodus and GlobalCenter; I just walked by their Borregas cage, in fact. The Rackables are now up into the thousands of CPUs (a friend of mine who works at Google built 1200 machines in a single afternoon; yes, they're identical, yes, it's done by DHCP and on a private VLAN). Finally, the chef doesn't cost much relative to their programmers (mostly Stanford and UCSB almost-PhD's).
Google is hella cool and both of my friends that work there, love it. They do work pretty hard, though. Still, when you love what you do...
So their tastes are imperative as well as fickle, in some years.
Cilk is pretty cool, actually. When I saw that it won so easily, I downloaded and played with it. Didn't use it for anything really cool, though.
Postgresql may not have outer joins, but at least its developers and copyright holders have respect for their hackers, users, and corporate partners.
I was thinking of getting into Interbase. This certainly dampened my enthusiasm. I will continue to work on Oracle for high-end (really high-end) work, use PostgreSQL for development, and deploy MySQL for data marts. Watching Informix lay off something like 900 people recently, and Sybase clinging to life, I wasn't too surprised at Interbase getting to be a dead end (for now).
Such a drag. The greedheads and day-trading degenerates seem to infect everything these days.
Why on earth is anyone surprised that a company whose first successful product was the 4.4BSD documentation, and whose headquarters is but a wee bit up the bay from Berkeley, should belly up to the table with a concise handbook to one of the most useful operating systems of our time?
;-)
Oh wait, it isn't GPLed, it must not be any good...
Legislation can be passed to prevent this sort of BSD-ish corporate cannibalization. Whether or not Celera invested millions into their research is irrelevant in the court of public opinion, where some enterprising senator will likely spin the issue as 'selling humanity' and nail Celera's more nefarious profit motives to the wall.
It's too easy not to do it (start up a backlash against patenting parts of people, as it were). The religious right will be all over it like a rash -- for once I happen to be pleased that they're a force.
Venter is a great guy, and an engaging public speaker, but he has many reasons to be bitter towards the established academic community of molecular genetics. And that's why I don't trust Celera to resist compromising Venter's stated principles in pursuit of profits.
I do not work for Yahoo and cannot speak for them.
Don't work for Google either, although one of my
friends now does. You'd probably be surprised to
find out who #7 on the web is -- that's us.
(MediaMetrix numbers -- we bounce around slightly
above Amazon and below Lycos most months) I'm not
sure that senior management would be real pleased
with me bashing Sun (or MS, for that matter) so
I'll leave it to interested parties to figure out
the rest. I simply use a Yahoo! email address
because I like their spam filtering services.
Yahoo has a long history of avoiding complexity
wherever possible, and BSD (along with flat files,
manual indexing, and their pile-o-netapps setups)
fits in with this strategy. They're cool cats.
Google takes a different approach. They take
Linux, hack it up a bit, and scale like you would
not believe. My friend who works there called me
up a couple weeks ago asking about using IP
multicasting to broadcast boot images. That's
pretty sick -- think on it for a moment. I hope
the stock market treats them as well as they
deserve to make out -- we tried to buy them and
they told us to go to hell (they want to IPO).
I don't know why Google likes Linux, but between
them and Yahoo!, both OSes should be healthy for a
long time to come. eg. Yahoo is big enough to
(apparently) get Oracle to do a build on FreeBSD,
although I can't confirm that (just something one
of the Walnut Creek guys passed along way back).
If you don't need FreeBSD's slight performance
edge, or need something in the Linux kernel that
isn't supported by BSD, then you should use Linux.
FreeBSD is just better for networking and I/O,
really. But that covers a lot of ground...
Also, I've seen some very strange things happen
when using Linux on loadbalanced servers doing
DSR. FreeBSD boxes in the same setup have always
worked great. Just another minor quibble.
Two very obvious reasons leap to mind:
;-).
1) the free-er BSD license -- do whatever you want
with the code, just give credit where it's due
2) superior networking performance -- I'm an admin
for one of the 10 largest sites on the web. We
have large numbers of both FreeBSD and Linux (and
Solaris) boxes in production. To be perfectly
honest, FreeBSD is the fastest, most reliable,
most configurable OS I have encountered so far.
I run Linux at home for no particular reason other
than I'm too lazy to switch; however I have found
that FreeBSD offers so many compelling advantages
as a single-purpose server that I deploy mostly
FreeBSD boxes where there is not a compelling
application forcing us to use Linux or Solaris.
And I'll probably switch over to FreeBSD at home,
too, maybe waiting until 4.1 comes out to be safe.
On laptops and such it makes sense to run Linux;
on a firewall, OpenBSD. But for maximum performance on web and mail servers, FreeBSD rocks
the house. Don't take my word for it -- try it
yourself. Many of the Apache and Qmail developers
run FreeBSD as their primary platform, and that's
not an easy bunch to impress.
FreeBSD is to Linux as Postgres is to MySQL
(which would bring us next to Solaris -> Oracle)
I'm not sure why you used the analogy you did. In the event of a natural disaster, a piece of Big Iron is just as fallible as a PC.
Which is one reason IBM sells clustering solutions for just about everything they make.
This makes me curious -- what would happen if the root A server got totalled? What gets failed over onto? I know I should RTFM, and I will, but my Stevens books are at home.