I was never stripped of any of my rights, and you weren't either, unless you're in the habit of communicating with suspected terror cells in-country or overseas.
You're arguing that having privacy and liberty "redefined" to be at the discretion of a wannabe socialist dictator (GW), well it really isn't a bad thing, because we've changed the meaning of the words "bad", "privacy" and "liberty". OK, I'm with you so far...
I find it amusing, on one hand, to find that everything that Bush was accused of doing with regard to personal privacy and such, is actually being done by the current administration.
Sorry, this statement adds nothing to your argument. If Obama is following in the footsteps of Bush, that means they're both equally good?
Whoa, getting dizzy. Now, we're led to believe that Bush was a deregulating freedom-fighting poor-hating conservative, but in reality he increased regulations, increased central government, made it possible for anyone to buy a home, explosively expanded spending and engaged in a protracted peace keeping effort. All the while, he was jumping on the (flawed) pro-environment conn ethanol fuel bandwagon. His detractors (his successor included) paint him as a far-right deregulator, but that just doesn't make sense.... get me off this crazy ride.
The only logical conclusion is: Nothing is as it seems.
Likewise, no UI I've seen approaches a command line in functionality. X11 to me is mostly a vehicle to manage dozens of xterms (rxvt's actually).
Want to generate a quick histogram of hits vs. IP from an apache log spanning midnight til 12:59am?
Try: awk '/11.Nov.2000:00:/{print $1;}' access_log | sort | uniq -c | sort
Cooking up a little pipeline like this only takes a few seconds when you're familiar with a cli (far less time than clicking your way to "M$-Histogram LogTool", opening and closing dialogs, aargh!).
As far as representing data with shapes, icons fall short. They generally need to be fairly large (bigger than a word) to be recognizable. Even then, they're usually accompanied with a text description or tooltips (because they simply don't provide enough data for positive identification). I fail to see how some fancy, rotating, textured solid would differ, except in compute reqs.
A cave-man's iconic repesentation of chasing a gazelle requires a few square feet. A textual representation (ie. "Og chase gazelle with spear".) is far more efficient.
Also consider a real-world desk. They're generally flat. You put papers, books, etc. on them. After thousands of years, they're still flat. No real innovation there. No significant transparent multi-layer approaches. Sure, you can use stackable paper trays and file folders, but they're essentially storage. While your paperwork is in these devices, its not being used. When you need to access something in bin-4, you take it out and put it on your flat desktop.
On the other hand, when you don't care too much for resolution in your data, 3-D can help. Our core network topology is fairly complex, 200 machines, routers, a few dozen administrative IP ranges, lotsa' vlans across a few local directors,
and thats before you count desktops/wan/etc. Most people need therapy after looking at the diagram.
Some sort of 3-D visualization would go miles towards deciphering it for the younger folks, though I suspect it might be a bit awkward in Euclid-space.
excerpt from article: "The genius of NetInfo is that it provides a uniform way of accessing and manipulating all system and network configuration information."
Maybe I'm missing something, but I fail to see any ingenuity here. Granted, these guys are good enough to admit they aren't hard-core, old-school sysadmins, but still... NIS/NIS+ have been answering this question for years now. Despite any failings you might cite about yp, netinfo hardly seems like an improvement.
The nsswitch mechanism, present on almost every unix these days, allows you to map {passwd, shadow, group, hosts, services, mail aliases, etc.} against {dns, local files, nis, etc} transparently as you see fit. If your system doesn't support host or password lookups against an LDAP database (as glibc-2.1 now does), there's a good chance you can build a module...
OK, having a central, common, consistent facility for everything sounds "nice", right? This flies in the face of the unix-credo: "Every tool should do one thing well".
When confronted with a scredriver and pliers, do you complain: "You mean this one works by turning and that one works by squeezing?"No. This, to me, is akin to complaining about having multiple formats for '/etc/passwd' and dns zone files.
When I read about doing name-service (esp. passwd) stuff from files in single user mode and via some external service during multiuser mode, I almost choked. Local files aren't consulted when you're connected to a remote netinfo server? (Unix answers the question with the '/etc/nsswitch.conf' entry hosts: files nis or similar.)
This essentially means that some external machine can tell you who root, wheel, localhost and shutdown are. I don't know if this is a horrible oversight, a design flaw, or some kludge to avoid implementing a real nsswitch. This is not a feature, its a bug. It begs questions about what other kludges will be used to patch it up.
It sounds to me like Apple has re-invented the wheel, and in fine tradtion, decided to make it different for the sake of being different.
Unfortunately, Alphas have a reputation of being refridgerator-sized machines with
AlphaLinux.Org A good starting place. They have tons of links to vendors, list-archives, news, FAQ's, etc.
DCG Inc. I've had good luck/great service from these guys. Alpha pricing starts at $1295 for a 533-mhz bare-bones kit.
Compaq's DS10 (21264-500mhz) The 466mhz model does over twice as well as a G4 in SPECfp. I seem to remember stumbling across a sale (from Compaq) for these little monsters for $2999, though I can't find the link now.
Microway Never dealt with them personally, but they have fast machines and a all-around good reputation. They also sell quadputers and compilers
eBay You can often find cheap Alpha hardware on eBay. Over 6 months ago, I put together a PC164-500/64mb system for about $600. Read the AlphaLinux.Org FAQ's, HOWTO's and HCL before you buy anything.
For reference, here are a few (single-cpu) spec*95 figures... (mostly from spec.org) INT- FP-- processor 20.3 13.3 Mac G3/466mhz 22.3 15.1 Intel P-III/550mhz 21.4 20.4 G4-450 Mhz 7400 16.2 23.9 UltraSparc/450mhz 18.0 27.0 Alpha 21164/600mhz (very old now) 24.6 47.9 Alpha 21264/466mhz (new "low-end") 32.1 53.7 Alpha UP2000/667mhz
Competiton is good Simply put, it breeds a better product. S.u.S.e. has enough of a user base in the E.U. that it will not be erased by RedHat. In order to compete with RedHat, S.u.S.e will likely become more open and "developer-friendly". In order to compete with S.u.S.e., RedHat will likely become more solid.
Growth No, not good, necessary for a business to survive in the computer industry.
World Domination and Microsoft This notion is somewhere between silly and ridiculous. RedHat and Microsoft are both experiencing success and profit, similarities end there. Where MS uses its weight to develop closed "standards" and thwart independance, RH is doing the opposite.
Though not intended as flame-bait, I fear this post might act as such...
I don't think this is an issue. These computers are no more inherently sexist than Lego, Barbie, Hot Wheels, "Plastic Tool-Chest", etc. (all tm someone or another). Its the children themselves who are sexist.
"Social Engineering" and other silly notions will not change the root-cause: human beings have conformity hard-wired into their brain stems, just like most other community-based organisms. Of course (thankfully) there are always exceptions.
The little boy with 200 HotWheels cars sees this computer and his heart leaps into his throat. He'll be proud to show this to his friends. If his parents buy him a "Barbie-Box" and he comes to school with "Bridal Barbie", he won't fit in.
To me, it just looks like these "Compu-Toy" manufacturers are just trying to grab a slice from an existing market.
A simple Dual-PII-450. with any decent multitasking OS (linux, maybe solaris-x86), you'll be faster (and happier (and cheaper than $2200) ) with a dual-cpu machine.
My ancient Alpha-PC164-500mhz The Athlon, touted for superior floating point, still falls short in Spec_FP. You can find PC164 motherboards (with 500mhz cpu) on eBay for about $250. These machines are arguably PC-architecture, having built-in PCI/ISA/IDE. Yes, these are slower on int perfs.
For those unfamiliar with the term DMZ, it stands for De-Militarized Zone. The notion here is that you have:
The Internet.
A firewall
The DMZ -This is where your Webservers go. They should be running minimal, secure services, static (ro) data, cgi's, etc.
Another firewall. - only allow access from the DMZ into your production net where absolutely needed (database, etc.)
Your internal network.
Additional good ideas are:
Use the "--rtfm" flag. There are tons of FAQ's out there that tell you to choose cryptic passwords, turn off services, limit access to needed IP's only.
Use NAT and private IP's. This is not a cure-all, but it is alot more annoying to crack an IP that you can't get a route to.
Disable network access on your routers. Get a serial-console server and place it somewhere well protected.
Sacrifice a goat.
Use sanity-checking application proxies. For example, if your web-servers need "write" access to an oracle database, install a proxy that verifies SQL queries against the set of queries that you've installed on your webserver.
If in doubt about using a restrictive fw-rule or policy, use it. If this breaks your application, you can remove the rule.
Install bogus services (and log activity). Most "original" cracks aren't instantaneous, they usually involve some poking around.
Currently, (x86) NBD is limited to devices of size (2^31)-1 bytes in length. I've been meaning to fix this, but (whine) haven't had the time.
Also, nbd uses both a client and a server process. If either one dies, you're left with a filesystem that fails on all i/o-ops, no way to umount it, and no way to "reconnect".
Other than that, you should be fine. I've built raid-5 filesystems (for fun) over nbd with fair performance.
First off, classifying attractive women as stupid is, in itself, a stupid stereotype. Ugly women like to perpetuate this myth.
Second, the fact that you readily lump them into the "dumb group" shows you guilty of the same "class judgement" that so-called-jocks are.
Third, anyone who chooses a PalmV over a week in Aspen either:
Doesn't like Aspen -or-
Is just plain dumb. (perhaps best employed as a cheerleader)
If "true" geeks are so intellectually superior, be sure of this: no-one without a technical bent will be able to judge it. Those who will be heralded as "net-setter uber-geeks" by the masses will:
Correction, Sun, having a far slower CPU (than Alpha), chooses wisely to emphasize I/O in its adverts. Anyway, the major speed considerations are disk, scsi-adapter and fs.
AlphaLinux supports almost an order of magnitude more adapters.
ext2 is faster than ufs.
Any "modern" Alpha, even the "old" 21164's, have 64-bit PCI slots.
With 25000 users, scaling is your largest issue. For this reason, don't go with WinNT.
First off, storage for your pop-server. Assuming an average 1mb per mbox, you'll need about 25gig of disk. You could stripe (raid0) across 4 disks, but this multiples risk. I'd recommend 4 20gb Ultra/33(or 66) ide drives and a "new" motherboard that supports them. This will give you ~60gb of raid5 storage (and you'll still bottleneck on 100-BT before you bottle on disk-IO). I'd hate to have 25000 users asking me where all of their mail went. (make sure you have a reliable backup solution!)
Next, build a few "outgoing" smtp-boxes (call them {smtp1,smtp2,...,smtpn.yourdomain) Configure sendmail to deliver local-mail to go to pop.yourdomain (the above machine). Single 10-15gb disks _should_ be plenty to hold your mqueue's (you probably don't want them growing that large anyway. When your queues start to back up, there's no way out.)
Lastly, get a Cisco Local Director (or some other form of load-balancing). Configure a virtual called smtp.yourdomain with smtp1,smtp2,...,smtpn as its component real-servers. Advantages here are that you can quickly add new (clone) machines to the pool, create new pools, etc.
You could additionally try getting other pop machines, mounting/var/spool/mail via NFS, creating a virtual in the LD, and putting them behind that.
In our datacenter, we have 150+ machines. Almost every one (for performance reasons) has a caching nameserver running with references to our "real" nameservers in their '/etc/named.{boot,conf}'. Add to that several class C networks, subnet a few of them. Routers and such expect IP addresses, changing the config on our Local Directors would take a few hours in itself... Changing our IP address space (on a moments notice) would be a nightmare.
Also consider that broadcasts (i.e. dhcp/bootp) don't cross networks. You don't want to join all of the nets together, broadcast traffic could eat up a good chunk of your bandwidth (at the very least, increase latency). So, you decide to maintain 5 DHCP servers. Congratulations, you've added 5 new single points of failure. You need redundancy, so you add another 5 machines, replicas of the first. (I could go on...)
What is(are) the "moral(s) of the story"?
bootp+production=bad.
There is no cure-all.
Read before you reply. This guy didn't have alot of time for prep-work.
Does anyone have any numbers on BeOS performance. As those of us old enough to vote know, every commercial OS vendor signs their own praises:
"Blazing New Speeds!!" This means that the new release is either faster than an old one, or that it seems faster due to increasing CPU speeds.
"Powerful Server OS!!" Utterly meaningless in most cases. Maybe they include some primitive, closed web-server?
"Advanced Filesystem" Could mean many things.
"Unique Multimedia Capabilities" They've coded in hardware acceleration for at least two chipsets, support 32-bit color and they have some new widget type hacked into one of their "control-panels".
"Outperforms Brands A and B" After conducting numerous comparitive tests, they found one test which, under certain (silly) circumstances, does indeed outdo the competition.
All very true. In my own defence, look again, I said "...might see (marginally) better...".
A while back I tried raid0 (2D/1C) with EIDE drives on an early UDMA controller and got roughly 1.3:1.0 speedup. Simply put, the drives were a fair amount slower than the controller.
Sorry for the top-level post. Instead of several 2nd level replies, I thought I'd try to answer a bunch of questions at once...
Do you need 3 channels? No. In fact with 2D/1C (2 disks on 1 ctrlr), you will still get getter performance 1D/1C in most cases. Under general use, you're doing small-size reads distributed across the disk, so the real bottleneck is head-seek. Even with big contiguous-block reads, you'll still notice an improvement.
RedHat Kernels and Autodetect First off, RH-kernels are far from stock linux kernels. Do an 'rpm -qpl [file].src.rpm' on one of their kernel SRPM's and you'll see a bunch of (non-dist) patches. Amoung them is the raid patch
Promise Cards Support for the new Ultra/66 hasn't hit the 2.2.x tree yet(I think). Check 2.3.5+ for new Ultra/(33,66) support.
Raid 5 on Three Disks
Doable: yes. Advisable: no. ( I've never tried it ) I suspect that you'd might see (marginally) better read-speeds, but you might even see degradation on write or mixed rdwr perfs ( since every write yanks two out of three heads across the platters )
Two small and one big? This won't work for raid5, not unless you want most of the large disk unprotected. Consider instead striping (for example) hda3+hdb3==md0, and then making a raid0 or raid1 volume md0+hdc3==md1. Better yet, get four disks of the same size...
Identical Drive Myth For hardware raid controllers, yes, go with identical disks. This is not needed for any kind of s/w raid I've dealt with (linux, disksuite, veritas-vm, xlv). For linux s/w-raid, you should be safe making a raid5-vol by mixing two ide-partitions, a scsi-disk, a loopback off of a file and a few NBD's (so long as they are the same size).
Hotswap-IDE Scary, risky and very unwise. So I'm not the only one...
Trying to make any correlation ("psycho killers are attracted to this stuff") frankly offends me. While I was in high-school playing (violent) video games, the children who later became violent criminals had no interest in them. They were more happily occupied torturing insects, small animals and other children, shoplifting and committing random acts of vandalism.
In fact, if you look at murders nationally, the majority falls outside of the Doom/Quake playing populace. You're more likely to find that where:
G is the # of people who play Doom/Quake.
Gm is the # of murders these gamers commit.
N is the # of people who don't play.
Nm is the # of murders committed by them. That (Gm/G) is far lower than (Nm/N). One might even suggest from this that Id Software promotes mental health. After all, if you don't like Doom, something must be wrong with you...
Re:What are the changes from 2.2.5-15?
on
Linux 2.3.0
·
· Score: 1
Its good that you're asking, but you're asking the wrong question.
The question should've been "Where can I find what?" not "What is what" In this case, try here. That overwrought saying "give a man a fish and he eats vs. teach a man how to fish and he eats for a lifetime" applies here in spades.
Read docs and man-pages, do a "grep -i snmp/usr/doc/HOWTO/*", FAQ's are your friends.
On another note, the RedHat-2.2.5 kernel is not a vanilla linux-2.2.5 in the first place. Raid and a few other patches have been applied. To get the complete list of changes, run, mount the source-cd, and run something like this: "rpm -qpl/mnt/cdrom/SRPMS/kernel-2.2.5-16.src.rpm"
Re:finger ftp.kernel.org
on
Linux 2.3.0
·
· Score: 1
Slashdot Mirror
I was never stripped of any of my rights, and you weren't either, unless you're in the habit of communicating with suspected terror cells in-country or overseas.
You're arguing that having privacy and liberty "redefined" to be at the discretion of a wannabe socialist dictator (GW), well it really isn't a bad thing, because we've changed the meaning of the words "bad", "privacy" and "liberty". OK, I'm with you so far...
I find it amusing, on one hand, to find that everything that Bush was accused of doing with regard to personal privacy and such, is actually being done by the current administration.
Sorry, this statement adds nothing to your argument. If Obama is following in the footsteps of Bush, that means they're both equally good?
Whoa, getting dizzy. Now, we're led to believe that Bush was a deregulating freedom-fighting poor-hating conservative, but in reality he increased regulations, increased central government, made it possible for anyone to buy a home, explosively expanded spending and engaged in a protracted peace keeping effort. All the while, he was jumping on the (flawed) pro-environment conn ethanol fuel bandwagon. His detractors (his successor included) paint him as a far-right deregulator, but that just doesn't make sense.... get me off this crazy ride.
The only logical conclusion is: Nothing is as it seems.
And both of you have a speech fimpediment.
There. Fixed that for you.
Hmmm, "flamebait". Really?
Questioning BB is doubleplusungood. How dare you mock his soundbites?
Likewise, no UI I've seen approaches a command line in functionality. X11 to me is mostly a vehicle to manage dozens of xterms (rxvt's actually).
Want to generate a quick histogram of hits vs. IP from an apache log spanning midnight til 12:59am?
Try: awk '/11.Nov.2000:00:/{print $1;}' access_log | sort | uniq -c | sort
Cooking up a little pipeline like this only takes a few seconds when you're familiar with a cli (far less time than clicking your way to "M$-Histogram LogTool", opening and closing dialogs, aargh!).
As far as representing data with shapes, icons fall short. They generally need to be fairly large (bigger than a word) to be recognizable. Even then, they're usually accompanied with a text description or tooltips (because they simply don't provide enough data for positive identification). I fail to see how some fancy, rotating, textured solid would differ, except in compute reqs. A cave-man's iconic repesentation of chasing a gazelle requires a few square feet. A textual representation (ie. "Og chase gazelle with spear".) is far more efficient.
Also consider a real-world desk. They're generally flat. You put papers, books, etc. on them. After thousands of years, they're still flat. No real innovation there. No significant transparent multi-layer approaches. Sure, you can use stackable paper trays and file folders, but they're essentially storage. While your paperwork is in these devices, its not being used. When you need to access something in bin-4, you take it out and put it on your flat desktop.
On the other hand, when you don't care too much for resolution in your data, 3-D can help. Our core network topology is fairly complex, 200 machines, routers, a few dozen administrative IP ranges, lotsa' vlans across a few local directors, and thats before you count desktops/wan/etc. Most people need therapy after looking at the diagram. Some sort of 3-D visualization would go miles towards deciphering it for the younger folks, though I suspect it might be a bit awkward in Euclid-space.
void rbowles(int signature)
{
signal(signature, rbowles);
raise(signature);
Looking back at my comment, I noticed that my "subject line" looks a bit like flame-bait.
Sorry, this was unintentional.
void rbowles(int signature)
{
signal(signature, rbowles);
raise(signature);
excerpt from article:
"The genius of NetInfo is that it provides a uniform way of accessing and manipulating all system and network configuration information."
Maybe I'm missing something, but I fail to see any ingenuity here. Granted, these guys are good enough to admit they aren't hard-core, old-school sysadmins, but still... NIS/NIS+ have been answering this question for years now. Despite any failings you might cite about yp, netinfo hardly seems like an improvement.
The nsswitch mechanism, present on almost every unix these days, allows you to map {passwd, shadow, group, hosts, services, mail aliases, etc.} against {dns, local files, nis, etc} transparently as you see fit. If your system doesn't support host or password lookups against an LDAP database (as glibc-2.1 now does), there's a good chance you can build a module...
OK, having a central, common, consistent facility for everything sounds "nice", right? This flies in the face of the unix-credo: "Every tool should do one thing well". When confronted with a scredriver and pliers, do you complain: "You mean this one works by turning and that one works by squeezing?" No. This, to me, is akin to complaining about having multiple formats for '/etc/passwd' and dns zone files.
When I read about doing name-service (esp. passwd) stuff from files in single user mode and via some external service during multiuser mode, I almost choked. Local files aren't consulted when you're connected to a remote netinfo server? (Unix answers the question with the '/etc/nsswitch.conf' entry hosts: files nis or similar.) This essentially means that some external machine can tell you who root, wheel, localhost and shutdown are. I don't know if this is a horrible oversight, a design flaw, or some kludge to avoid implementing a real nsswitch. This is not a feature, its a bug. It begs questions about what other kludges will be used to patch it up.
It sounds to me like Apple has re-invented the wheel, and in fine tradtion, decided to make it different for the sake of being different.
I'll stick with my round wheels, thank you.
void rbowles(int signature)
{
signal(signature, rbowles);
raise(signature);
Or, if you don't want to worry about loop-device allocation:
mount -t iso9660 -o loop [file] [mnt-pnt]
You might also want to avoid mounting on '/mnt', as subdirs (such as '/mnt/floppy') are used by many distros.
No need to wrap this up in fluff. Subject says it all.
The Simpsons are real. After all, I've seen them (on TV) with my own eyes.
Unfortunately, Alphas have a reputation of being refridgerator-sized machines with
A good starting place. They have tons of links to vendors, list-archives, news, FAQ's, etc.
I've had good luck/great service from these guys. Alpha pricing starts at $1295 for a 533-mhz bare-bones kit.
The 466mhz model does over twice as well as a G4 in SPECfp. I seem to remember stumbling across a sale (from Compaq) for these little monsters for $2999, though I can't find the link now.
Never dealt with them personally, but they have fast machines and a all-around good reputation. They also sell quadputers and compilers
You can often find cheap Alpha hardware on eBay. Over 6 months ago, I put together a PC164-500/64mb system for about $600. Read the AlphaLinux.Org FAQ's, HOWTO's and HCL before you buy anything.
For reference, here are a few (single-cpu) spec*95 figures... (mostly from spec.org)
INT- FP-- processor
20.3 13.3 Mac G3/466mhz
22.3 15.1 Intel P-III/550mhz
21.4 20.4 G4-450 Mhz 7400
16.2 23.9 UltraSparc/450mhz
18.0 27.0 Alpha 21164/600mhz (very old now)
24.6 47.9 Alpha 21264/466mhz (new "low-end")
32.1 53.7 Alpha UP2000/667mhz
Simply put, it breeds a better product. S.u.S.e. has enough of a user base in the E.U. that it will not be erased by RedHat. In order to compete with RedHat, S.u.S.e will likely become more open and "developer-friendly". In order to compete with S.u.S.e., RedHat will likely become more solid.
No, not good, necessary for a business to survive in the computer industry.
This notion is somewhere between silly and ridiculous. RedHat and Microsoft are both experiencing success and profit, similarities end there. Where MS uses its weight to develop closed "standards" and thwart independance, RH is doing the opposite.
Though not intended as flame-bait, I fear this post might act as such...
I don't think this is an issue. These computers are no more inherently sexist than Lego, Barbie, Hot Wheels, "Plastic Tool-Chest", etc. (all tm someone or another). Its the children themselves who are sexist.
"Social Engineering" and other silly notions will not change the root-cause: human beings have conformity hard-wired into their brain stems, just like most other community-based organisms. Of course (thankfully) there are always exceptions.
The little boy with 200 HotWheels cars sees this computer and his heart leaps into his throat. He'll be proud to show this to his friends. If his parents buy him a "Barbie-Box" and he comes to school with "Bridal Barbie", he won't fit in.
To me, it just looks like these "Compu-Toy" manufacturers are just trying to grab a slice from an existing market.
Here are a few examples:
with any decent multitasking OS (linux, maybe solaris-x86), you'll be faster (and happier (and cheaper than $2200) ) with a dual-cpu machine.
The Athlon, touted for superior floating point, still falls short in Spec_FP. You can find PC164 motherboards (with 500mhz cpu) on eBay for about $250. These machines are arguably PC-architecture, having built-in PCI/ISA/IDE. Yes, these are slower on int perfs.
For those unfamiliar with the term DMZ, it stands for De-Militarized Zone. The notion here is that you have:
Additional good ideas are:
Currently, (x86) NBD is limited to devices of size (2^31)-1 bytes in length. I've been meaning to fix this, but (whine) haven't had the time.
Also, nbd uses both a client and a server process. If either one dies, you're left with a filesystem that fails on all i/o-ops, no way to umount it, and no way to "reconnect".
Other than that, you should be fine. I've built raid-5 filesystems (for fun) over nbd with fair performance.
Correction, Sun, having a far slower CPU (than Alpha), chooses wisely to emphasize I/O in its adverts. Anyway, the major speed considerations are disk, scsi-adapter and fs.
With 25000 users, scaling is your largest issue. For this reason, don't go with WinNT.
First off, storage for your pop-server. Assuming an average 1mb per mbox, you'll need about 25gig of disk. You could stripe (raid0) across 4 disks, but this multiples risk. I'd recommend 4 20gb Ultra/33(or 66) ide drives and a "new" motherboard that supports them. This will give you ~60gb of raid5 storage (and you'll still bottleneck on 100-BT before you bottle on disk-IO). I'd hate to have 25000 users asking me where all of their mail went. (make sure you have a reliable backup solution!)
Next, build a few "outgoing" smtp-boxes (call them {smtp1,smtp2,...,smtpn.yourdomain) Configure sendmail to deliver local-mail to go to pop.yourdomain (the above machine). Single 10-15gb disks _should_ be plenty to hold your mqueue's (you probably don't want them growing that large anyway. When your queues start to back up, there's no way out.)
Lastly, get a Cisco Local Director (or some other form of load-balancing). Configure a virtual called smtp.yourdomain with smtp1,smtp2,...,smtpn as its component real-servers. Advantages here are that you can quickly add new (clone) machines to the pool, create new pools, etc.
You could additionally try getting other pop machines, mounting /var/spool/mail via NFS, creating a virtual in the LD, and putting them behind that.
Not necessarily all his fault...
In our datacenter, we have 150+ machines. Almost every one (for performance reasons) has a caching nameserver running with references to our "real" nameservers in their '/etc/named.{boot,conf}'. Add to that several class C networks, subnet a few of them. Routers and such expect IP addresses, changing the config on our Local Directors would take a few hours in itself... Changing our IP address space (on a moments notice) would be a nightmare.
Also consider that broadcasts (i.e. dhcp/bootp) don't cross networks. You don't want to join all of the nets together, broadcast traffic could eat up a good chunk of your bandwidth (at the very least, increase latency). So, you decide to maintain 5 DHCP servers. Congratulations, you've added 5 new single points of failure. You need redundancy, so you add another 5 machines, replicas of the first. (I could go on...)
Does anyone have any numbers on BeOS performance. As those of us old enough to vote know, every commercial OS vendor signs their own praises:
This means that the new release is either faster than an old one, or that it seems faster due to increasing CPU speeds.
Utterly meaningless in most cases. Maybe they include some primitive, closed web-server?
Could mean many things.
They've coded in hardware acceleration for at least two chipsets, support 32-bit color and they have some new widget type hacked into one of their "control-panels".
After conducting numerous comparitive tests, they found one test which, under certain (silly) circumstances, does indeed outdo the competition.
Production is the only true benchmark.
All very true. In my own defence, look again, I said "...might see (marginally) better...".
A while back I tried raid0 (2D/1C) with EIDE drives on an early UDMA controller and got roughly 1.3:1.0 speedup. Simply put, the drives were a fair amount slower than the controller.
No. In fact with 2D/1C (2 disks on 1 ctrlr), you will still get getter performance 1D/1C in most cases. Under general use, you're doing small-size reads distributed across the disk, so the real bottleneck is head-seek. Even with big contiguous-block reads, you'll still notice an improvement.
First off, RH-kernels are far from stock linux kernels. Do an 'rpm -qpl [file].src.rpm' on one of their kernel SRPM's and you'll see a bunch of (non-dist) patches. Amoung them is the raid patch
Support for the new Ultra/66 hasn't hit the 2.2.x tree yet(I think). Check 2.3.5+ for new Ultra/(33,66) support.
( I've never tried it ) I suspect that you'd might see (marginally) better read-speeds, but you might even see degradation on write or mixed rdwr perfs ( since every write yanks two out of three heads across the platters )
This won't work for raid5, not unless you want most of the large disk unprotected. Consider instead striping (for example) hda3+hdb3==md0, and then making a raid0 or raid1 volume md0+hdc3==md1.
Better yet, get four disks of the same size...
For hardware raid controllers, yes, go with identical disks. This is not needed for any kind of s/w raid I've dealt with (linux, disksuite, veritas-vm, xlv). For linux s/w-raid, you should be safe making a raid5-vol by mixing two ide-partitions, a scsi-disk, a loopback off of a file and a few NBD's (so long as they are the same size).
Scary, risky and very unwise.
So I'm not the only one...
Trying to make any correlation ("psycho killers are attracted to this stuff") frankly offends me. While I was in high-school playing (violent) video games, the children who later became violent criminals had no interest in them. They were more happily occupied torturing insects, small animals and other children, shoplifting and committing random acts of vandalism.
In fact, if you look at murders nationally, the majority falls outside of the Doom/Quake playing populace. You're more likely to find that where:
G is the # of people who play Doom/Quake.
Gm is the # of murders these gamers commit.
N is the # of people who don't play.
Nm is the # of murders committed by them.
That (Gm/G) is far lower than (Nm/N). One might even suggest from this that Id Software promotes mental health. After all, if you don't like Doom, something must be wrong with you...
Its good that you're asking, but you're asking the wrong question.
The question should've been "Where can I find what?" not "What is what" In this case, try here. That overwrought saying "give a man a fish and he eats vs. teach a man how to fish and he eats for a lifetime" applies here in spades.
Read docs and man-pages, do a "grep -i snmp /usr/doc/HOWTO/*", FAQ's are your friends.
On another note, the RedHat-2.2.5 kernel is not a vanilla linux-2.2.5 in the first place. Raid and a few other patches have been applied. To get the complete list of changes, run, mount the source-cd, and run something like this: /mnt/cdrom/SRPMS/kernel-2.2.5-16.src.rpm"
"rpm -qpl
Erm... you meant "finger @linux.kernel.org".
( Sorry to nit-pick... )