I feel like my response would be something like "Well, I've never in my life put together a standalone computer system that lasted fifteen years without maintenance, but you've already done it once, so why are you asking *me* for advice. Sounds like I should be asking *you*."
Wasn't it HP-UX years ago where the "@" symbol was some sort of delete key? I remember once it taking me a while to sort out why an employee kept complaining that his password wasn't working only on certain systems.
I heard an interview with John Doe (of X, The Knitters, etc) and he was asked about auto-tuning at one point. He said that even if you're someone who doesn't normally use them, sometimes when you've been in the studio for hours and you're just having one of those days where you can't seem to get a particular part right, you just decide fuck it and use the auto-tuner so that you can record the damn thing and move on.
I've mostly found it a sign of a company's size/age/maturity as to how boring the server names are. Several places I've worked for started out with the admins coming up with their funny/cute/dorky naming schemes, only to eventually have server names be locked down in the name of STANDARDIZATION.
Then you have endless meetings to decide what should be the important components of a system name. Should it indicate the machine's location? It's OS? It's function? Should it even indicate which rack number and elevation slot the system is in? Eventually you end up with racks full of servers named SJC-LX-APPDEV01, NYC-SV-EXCHG02, and LDN-UX-SMTPDR01.
I have to admit, a little part of me misses having room for a little creativity in naming systems, but then the rest of me doesn't miss wasting time trying to come up with names for work systems. I've always got my home network to label with my ever-changing nerdly obsessions.
There are free-as-in-beer email servers, even for very high volumes of mail, that any competent IT staff could maintain with minimal effort and better reliability than GMail. How much money do you think GMail would save? Is that amount of money actually worth the hassle of dealing with GMail?
According to a Forester report, they estimate that it costs on average $25.18 per month per user to provide email services in-house, compared to $8.47 for Gmail.
Interestingly, most people couldn't actually guess what the real cost of providing email services in-house was, many guessing $2-11 per user.
The upshot of Forester's analysis was that up to around 15,000 users, it could be substantially cheaper to outsource email as an infrastructure service.
Admittedly, there can be a lot more to the calculations though. Depending on your business needs or industry, you could have regulatory or compliance requirements that might interfere with an outsourced solution if the vendor can't meet those requirements.
Likewise, after my most recent hiring I was told one of the strongest factors in my favor wasn't my 15+ years of technical experience, it was the hiring manager's sense that I was a low-stress personality type who would not be driven to insanity by the high-stress nature of the job.
This wasn't based on any particular personality test, mind you, just the hiring manager's judgment call based on my performance in the interviews.
Since then I've seen potential candidates for other positions in my group who met the professional qualifications passed up because they seemed wound too tight for the work.
The best that Sun could do is make OpenSolaris as much of a developer workstation OS as they can, in competition with Linux.
I keep thinking that Sun should seriously take a shot at the high-capacity low-cost storage market. They seem to have some really good ideas behind the "Thumper"-type solutions, but they're still priced way WAY too high.
Overpriced products easily duplicated with FreeNAS or any number of products at a fraction of the cost.
Are people really running petabytes of enterprise-class storage on FreeNAS? Don't get me wrong, FreeNAS and others like it are great products for tier 3 and maybe even tier 2 storage, but for serious high-availability storage clusters?
Can we quickly run through which of those other solutions have features equivalent to VMotion?
(Seriously, I seem to recall Xen or KVM was working towards something like that but don't know the status. I don't believe MS's free solution nor Parallel's non-free one has anything remotely Enterprise-suitable.)
I've got Parallels on my Macbook Pro right now, and it's fine for my purposes, but I really can't imagine anyone running production data center virtual machines with it.
Something like 15 years ago when I worked at a job supporting AIX, someone once made an error on an invoice and we received a full copy of AIX on floppies. I can't remember how many total disks there were, but there were multiple boxes of them.
They're both just examples of using the language of inspiration. The difference is one group uses it out of learned experience in inspiring people, the other group once read a few pages of a book on how to inspire people.
What I wonder is at what point, if any, does the temperature of a data center trigger OSHA involvement?
I'm sure most of us who have had to work in a data center with malfunctioning air conditioning can attest to just how uncomfortable it can be to work in a 90+F closed room for very long.
I remember when I worked at a fast food place, we once had some OSHA inspectors (or at least some kind of gummint inspectors) going around placing temperature monitors for a few days. I never found out what it was all about though, but I assumed it had something to do with complaints about it being too hot in the kitchen.
And yet the response is always "just use soap and water".
So why aren't we getting soap-and-water-resistant bacteria? Presumably because such an evolved trait is too "expensive".
A genetics professor of mine once explained that when I asked if bacteria can become resistant to alcohol. (As he was wiping his hands with Purell.) He said, yes, you can induce bacteria to evolve alcohol-resistance in a lab environment, but it's such an expensive adaptation that as soon as the alcohol exposure is reduced, the trait rapidly disappears again.
So the real question would be, is any resistance encouraged by this nano-particle approach an expensive trait or not?
The Nomad was not remotely first either. The PJBox predated it, at the very least.
Also every edition of the Nomad completely sucked. It was nine times larger than it needed to be, just so that it could still look like a CD player for some reason. And a friend of mine that had one was constantly complaining about how long it took to "boot up" when he turned it on.
You really think it's urban areas that are getting wired up first? I lived in San Jose for years and was still using ISDN when my co-workers and friends in outer suburbs were on DSL or Cable.
I always assumed it was because it's easier to wire up a new development out in the suburbs than it is to go into the heart of a city and run new cables/fiber.
My water bill here in Chicago is flat-rate. It's based on the size of my property and the number of outdoor faucets I have, if I recall.
My garbage service I don't pay for directly at all. (It's paid for by the city so obviously I do pay for it, but I don't pay anything based on my usage.) People can leave just about anything out on the curb and the city picks it up. It's not unusual to see large couches or appliances sitting out on trash day.
On the other hand there are toll roads everywhere. Which of course always serve to make traffic actively worse. And the city is upgrading a lot of it's parking meters and paid parking lots to the new electronic ones that prevent you from leaving extra time to the next person.
Out of curiosity, what particular features in ZFS do you want on Linux? I mean, it's a large step upwards from solaris disksuite, but compared to the linux device-mapper/filesystem paradigm it's not a particularly large improvement (if it is one at all)
In a major corporation, fear of massive fines and prosecution is enough to stop them from pirating your software.
As someone who has worked in IT at various companies over the last 15+ years, I have to say, no it isn't. Fear of fines and prosecution is not remotely enough to stop unlicensed use of software. Most of the time, the people who would be afraid of that have no idea how out of compliance they are for some software. Particularly things like copies of Windows, Office, Visio, various little utilities that someone got a legal copy of and then shared it with their team, etc.
Now, having said that, all this license management crap *is* stupid and is more annoying to real customers than it is beneficial to software sales.
Hitchhiker's Guide to the Galaxy had arguably the worst example of that, which you never got the opportunity to scream at.
Very near the end of the game, the entire ship would be eaten by a dog, if you did not happen to think of feeding the dog a sandwich in the very beginning of the game.
At least, this is my vague memory of the game's biggest annoyance. It has of course been many many years since I played it. I even had a "I got the babel fish!" t-shirt.
My understanding of the DHCP protocol was that once a system has obtained a DHCP lease it is expected to keep that same IP address as long as it keeps requesting a lease renewal before the lease expires.
It's not *guaranteed* that the IP address won't change, of course, but if AT&T is handing out different addresses every time a device reboots then that's not normal to me. Maybe they have an extremely short lease configured?
I've had quite a few different routers over the years, mostly Linksys but also one Netgear combo modem/router. The only time I've ever had to reboot one regularly was when I was running DD-WRT on one. When I switched that to Tomato the daily reboots stopped being necessary.
But, I have pretty much always had my router powered through a decent UPS, so that indeed might be relevant.
Even though this is days old and nobody will ever see it, I felt I should record for posterity that I was simply being a moron in my original post and transposed CNAME and A records. The follow-ups who corrected my stupidity are saying what I actually meant.
Are we sure about that? My understanding is that the server should be rejecting the message as undeliverable during the initial SMTP conversation, not sending any bounce messages afterwards.
This would mean that $SPAM_SERVER says "Hello, $LEGITIMATE_SERVER, please deliver this message to $RECIPIENT from $FAKE_SENDER" and $LEGITIMATE_SERVER says "Sorry, I don't know who $RECIPIENT is."
$LEGITIMATE_SERVER never sends a bounce message to anyone. It just says "no such address". What are the odds that a botnet spamming mail server is doing to waste CPU cycles returning undeliverable notices back up the chain that it already knows is fake? What would be the profit in that?
Actually, you should never use NTP in a virtual machine. Virtual machines perceive inconsistent clock ticks from their virtual CPU, which can confuse the holy hell out of NTP as it is constantly trying to predict the clock drift based on ticks.
At least, that's what some VMWare engineers told me at a conference once. But it was consistent with some problems I'd seen with NTP clients in VMs having problems keeping the clock synced.
As for "Where do I stop this virtualization thingie?" You stop it when it makes sense to do so. You probably shouldn't do it for an NTP server, but you should still pick the server you decide to add the NTP service to carefully.
Slashdot Mirror
I feel like my response would be something like "Well, I've never in my life put together a standalone computer system that lasted fifteen years without maintenance, but you've already done it once, so why are you asking *me* for advice. Sounds like I should be asking *you*."
Wasn't it HP-UX years ago where the "@" symbol was some sort of delete key? I remember once it taking me a while to sort out why an employee kept complaining that his password wasn't working only on certain systems.
I heard an interview with John Doe (of X, The Knitters, etc) and he was asked about auto-tuning at one point. He said that even if you're someone who doesn't normally use them, sometimes when you've been in the studio for hours and you're just having one of those days where you can't seem to get a particular part right, you just decide fuck it and use the auto-tuner so that you can record the damn thing and move on.
I've mostly found it a sign of a company's size/age/maturity as to how boring the server names are. Several places I've worked for started out with the admins coming up with their funny/cute/dorky naming schemes, only to eventually have server names be locked down in the name of STANDARDIZATION.
Then you have endless meetings to decide what should be the important components of a system name. Should it indicate the machine's location? It's OS? It's function? Should it even indicate which rack number and elevation slot the system is in? Eventually you end up with racks full of servers named SJC-LX-APPDEV01, NYC-SV-EXCHG02, and LDN-UX-SMTPDR01.
I have to admit, a little part of me misses having room for a little creativity in naming systems, but then the rest of me doesn't miss wasting time trying to come up with names for work systems. I've always got my home network to label with my ever-changing nerdly obsessions.
There are free-as-in-beer email servers, even for very high volumes of mail, that any competent IT staff could maintain with minimal effort and better reliability than GMail. How much money do you think GMail would save? Is that amount of money actually worth the hassle of dealing with GMail?
According to a Forester report, they estimate that it costs on average $25.18 per month per user to provide email services in-house, compared to $8.47 for Gmail.
Interestingly, most people couldn't actually guess what the real cost of providing email services in-house was, many guessing $2-11 per user.
The upshot of Forester's analysis was that up to around 15,000 users, it could be substantially cheaper to outsource email as an infrastructure service.
Admittedly, there can be a lot more to the calculations though. Depending on your business needs or industry, you could have regulatory or compliance requirements that might interfere with an outsourced solution if the vendor can't meet those requirements.
The Forester report: http://www.forrester.com/Research/Document/Excerpt/0,7211,46302,00.html
Arstechnica report on the report: http://arstechnica.com/news.ars/post/20090108-report-gmail-about-one-third-as-expensive-as-hosted-e-mail.html
Well, what about people who do have that sort of personality? What are they supposed to do, starve?
Work in a different industry? There are a lot of IT jobs that aren't in trading/clearing/brokering.
Likewise, after my most recent hiring I was told one of the strongest factors in my favor wasn't my 15+ years of technical experience, it was the hiring manager's sense that I was a low-stress personality type who would not be driven to insanity by the high-stress nature of the job.
This wasn't based on any particular personality test, mind you, just the hiring manager's judgment call based on my performance in the interviews.
Since then I've seen potential candidates for other positions in my group who met the professional qualifications passed up because they seemed wound too tight for the work.
The best that Sun could do is make OpenSolaris as much of a developer workstation OS as they can, in competition with Linux.
I keep thinking that Sun should seriously take a shot at the high-capacity low-cost storage market. They seem to have some really good ideas behind the "Thumper"-type solutions, but they're still priced way WAY too high.
2) NetApp
Overpriced products easily duplicated with FreeNAS or any number of products at a fraction of the cost.
Are people really running petabytes of enterprise-class storage on FreeNAS? Don't get me wrong, FreeNAS and others like it are great products for tier 3 and maybe even tier 2 storage, but for serious high-availability storage clusters?
Can we quickly run through which of those other solutions have features equivalent to VMotion?
(Seriously, I seem to recall Xen or KVM was working towards something like that but don't know the status. I don't believe MS's free solution nor Parallel's non-free one has anything remotely Enterprise-suitable.)
I've got Parallels on my Macbook Pro right now, and it's fine for my purposes, but I really can't imagine anyone running production data center virtual machines with it.
Something like 15 years ago when I worked at a job supporting AIX, someone once made an error on an invoice and we received a full copy of AIX on floppies. I can't remember how many total disks there were, but there were multiple boxes of them.
They're both just examples of using the language of inspiration. The difference is one group uses it out of learned experience in inspiring people, the other group once read a few pages of a book on how to inspire people.
What I wonder is at what point, if any, does the temperature of a data center trigger OSHA involvement?
I'm sure most of us who have had to work in a data center with malfunctioning air conditioning can attest to just how uncomfortable it can be to work in a 90+F closed room for very long.
I remember when I worked at a fast food place, we once had some OSHA inspectors (or at least some kind of gummint inspectors) going around placing temperature monitors for a few days. I never found out what it was all about though, but I assumed it had something to do with complaints about it being too hot in the kitchen.
And yet the response is always "just use soap and water".
So why aren't we getting soap-and-water-resistant bacteria? Presumably because such an evolved trait is too "expensive".
A genetics professor of mine once explained that when I asked if bacteria can become resistant to alcohol. (As he was wiping his hands with Purell.) He said, yes, you can induce bacteria to evolve alcohol-resistance in a lab environment, but it's such an expensive adaptation that as soon as the alcohol exposure is reduced, the trait rapidly disappears again.
So the real question would be, is any resistance encouraged by this nano-particle approach an expensive trait or not?
The Nomad was not remotely first either. The PJBox predated it, at the very least.
Also every edition of the Nomad completely sucked. It was nine times larger than it needed to be, just so that it could still look like a CD player for some reason. And a friend of mine that had one was constantly complaining about how long it took to "boot up" when he turned it on.
You really think it's urban areas that are getting wired up first? I lived in San Jose for years and was still using ISDN when my co-workers and friends in outer suburbs were on DSL or Cable.
I always assumed it was because it's easier to wire up a new development out in the suburbs than it is to go into the heart of a city and run new cables/fiber.
My water bill here in Chicago is flat-rate. It's based on the size of my property and the number of outdoor faucets I have, if I recall.
My garbage service I don't pay for directly at all. (It's paid for by the city so obviously I do pay for it, but I don't pay anything based on my usage.) People can leave just about anything out on the curb and the city picks it up. It's not unusual to see large couches or appliances sitting out on trash day.
On the other hand there are toll roads everywhere. Which of course always serve to make traffic actively worse. And the city is upgrading a lot of it's parking meters and paid parking lots to the new electronic ones that prevent you from leaving extra time to the next person.
Out of curiosity, what particular features in ZFS do you want on Linux? I mean, it's a large step upwards from solaris disksuite, but compared to the linux device-mapper/filesystem paradigm it's not a particularly large improvement (if it is one at all)
Block checksums?
In a major corporation, fear of massive fines and prosecution is enough to stop them from pirating your software.
As someone who has worked in IT at various companies over the last 15+ years, I have to say, no it isn't. Fear of fines and prosecution is not remotely enough to stop unlicensed use of software. Most of the time, the people who would be afraid of that have no idea how out of compliance they are for some software. Particularly things like copies of Windows, Office, Visio, various little utilities that someone got a legal copy of and then shared it with their team, etc.
Now, having said that, all this license management crap *is* stupid and is more annoying to real customers than it is beneficial to software sales.
Hitchhiker's Guide to the Galaxy had arguably the worst example of that, which you never got the opportunity to scream at.
Very near the end of the game, the entire ship would be eaten by a dog, if you did not happen to think of feeding the dog a sandwich in the very beginning of the game.
At least, this is my vague memory of the game's biggest annoyance. It has of course been many many years since I played it. I even had a "I got the babel fish!" t-shirt.
My understanding of the DHCP protocol was that once a system has obtained a DHCP lease it is expected to keep that same IP address as long as it keeps requesting a lease renewal before the lease expires.
It's not *guaranteed* that the IP address won't change, of course, but if AT&T is handing out different addresses every time a device reboots then that's not normal to me. Maybe they have an extremely short lease configured?
I've had quite a few different routers over the years, mostly Linksys but also one Netgear combo modem/router. The only time I've ever had to reboot one regularly was when I was running DD-WRT on one. When I switched that to Tomato the daily reboots stopped being necessary.
But, I have pretty much always had my router powered through a decent UPS, so that indeed might be relevant.
Even though this is days old and nobody will ever see it, I felt I should record for posterity that I was simply being a moron in my original post and transposed CNAME and A records. The follow-ups who corrected my stupidity are saying what I actually meant.
Are we sure about that? My understanding is that the server should be rejecting the message as undeliverable during the initial SMTP conversation, not sending any bounce messages afterwards.
This would mean that $SPAM_SERVER says "Hello, $LEGITIMATE_SERVER, please deliver this message to $RECIPIENT from $FAKE_SENDER" and $LEGITIMATE_SERVER says "Sorry, I don't know who $RECIPIENT is."
$LEGITIMATE_SERVER never sends a bounce message to anyone. It just says "no such address". What are the odds that a botnet spamming mail server is doing to waste CPU cycles returning undeliverable notices back up the chain that it already knows is fake? What would be the profit in that?
Actually, you should never use NTP in a virtual machine. Virtual machines perceive inconsistent clock ticks from their virtual CPU, which can confuse the holy hell out of NTP as it is constantly trying to predict the clock drift based on ticks.
At least, that's what some VMWare engineers told me at a conference once. But it was consistent with some problems I'd seen with NTP clients in VMs having problems keeping the clock synced.
As for "Where do I stop this virtualization thingie?" You stop it when it makes sense to do so. You probably shouldn't do it for an NTP server, but you should still pick the server you decide to add the NTP service to carefully.