We had a single JBOD made by Nstor, who were bought by Xyratex. Xyratex in theory serviced it, but we had to rip the interface board out and send it away for several weeks, no advance or on-site replacement, and when it came back, it was still broken. This array actually had two disks per tray, which meant that for online servicing one had to know that in advance (not always obvious when one is remote) and lay out volumes very carefully. Mind you this wasn't a Xyratex product per se, but the way they handled it didn't impress me.
The Linux md driver is meh, but at least it doesn't require compiling metadevice layout into the kernel. It's there and it works, but like SVM/SDS/ODS from 15 years ago it's dated and limited, and mdadmin et al are clumsy at best.
"Hardware RAID" means different things to different people. Two main divisions:
o HBA RAID. Claimed advantage in that mirror/parity writes don't clog up the host channel. Another is that on systems with anachronistic BIOS stupidity, booting from a mirror when the primary fails can be difficult or often impossible to set up. Big disadvantages:
- No model that I've yet seen does 3-way mirrors. None. Disks fail, and more than once I've had a mirror fail while replacing the other side. It can take time to get bad disks replaced, and if remote local hands yank the good side of a mirror pair instead of the bad, you're screwed.
- SPoF. Volumes can't span HBA's, and yes, HBA's do go bad.
- Crummy monitoring. Some vendors supply CLI's for some OS's, but ongoing support is a uncertain, and the interfaces are always downright horrid. raidctl, for example, on my systems behaves in at least three distinct ways.
- Want to use one for the boot device? Be prepared to hit a multi-key sequence during a split-second at the BIOS level after POST. These often assume that you're sitting in front of a desktop, and can enter Alt-keys or function keys to invoke. Using a remote serial console? Sorry, you're screwed.
o Chassis RAID: SPoF unless 2-3 identical arrays are mirrored in software. No monitoring to speak of. If you get a serial console, it may need a proprietary adapter cable, null-modem, *and* DB9-RJ45 adapter, as many mistake a 1/8" round headphone-style jack for a serial connector.
Linux storage is miserable, still stuck in the mindset of a desktop user with 1-2 drives. The sda/sdb/sdc/etc. naming convention is meant to echo legacy MS-DOS drive letters, and obscures vital information about which drives are in what slot of what array. cXtXdX FTW. When adding an HBA or disks, existing disks can even suffer name changes. ZFS / btrfs is *desperately* needed, but there is no indication of viability anytime soon.
HP assumes that *everything* is a desktop machine plugged into a tube and keyboard, even rackmount systems. They nickle/dime everything they possibly can, including functionality on their HBA's and service processors. Yes, one has to give them ~$150 to unlock features on a FRIGGIN HBA. The third strike: their tech support is offshored, with all the worst features that brings: techs who call during their working hours (not yours), who you can't understand anyway. Techs who don't read the ticket you enter and give non-sensical, non-sequitor answers that are clearly cut/pasted from a list.
A few years ago I was stuck with ISDN for a while at home, which would have sucked for downloading Lion. But even then, I was 10 minutes away from 3-4 coffee places that would have sufficed, and 12 away from a library with excellent speed.
but people using Linux get an inferior experience
This is the Linux way, of course. Having recently been forced to start deploying RHEL instead of Solaris, I'm constantly running into shit that's either just stupid, or 15 years out of date.
Therefore there are about 7000/5 = 1400 NetBSD users
Am I the only one who sees the sophistry here? Just in case:
o Unspecifed method for counting *BSD posts
o Usenet itself has been irrelevant for years, due both to the flood of spammers and migration to web-based forums
I argue that *BSD has been irrelevant for years. What innovation, or even keeping up with the times, has it done since the 1980's?
In the 1995-1997 timeframe I was stuck with BSD/OS thanks to a decision that preceded me, made by a network engineer of all people. It was a fscking nightmare. The preceding decision maker had asserted that it would be inexpensive by virtue of running on peecee hardware that we could maintain ourselves. In the end the countless hours and karma lost dicking with both the hardware and software were soul-devouring. A few low points:
o Rob Kolstad, who believed that home directories -- yes, home directories -- belong in/mnt
o Systems were extremely prone to wedging, requiring physical intervention
o Systems were especially unstable with >256MB of physmem, and BSDi wouldn't even look at their kernel dumps because they didn't have a machine with more than that themselves to load them into, and refused to get/upgrade one.
o No matter how hard I tried, and no matter what well-regarded mobo I sourced, I could not for the life of me get two SCSI HBA's to co-exist in a single system.
o No software striping/mirroring. In an OS that cost a grand. At one point someone emailed me some old *BSD striping driver that I could hack into the kernel, but it was slow, required that the set of metadevices be compiled and hardcoded into the kernel, including component devices and their sizes.
o Getting stuff to compile and run without dumping immediate core was a nightmare, even worse than Ultrix
I eventually won out against the unrelenting nightmares and installed Solaris on the same systems. They ran faster and stayed up.
This is much like doing business with HP. Want to actually *use* iLO? Cough up $200+ for the "advanced" license key. Want your Smart Array HBA to do RAID60? Cough up more and enter a license key *onto the frickin' HBA*.
In 1987 or 1986 I was told by a friend that Hurd was going to be released RSN. That never happened but he went on to "work" for the FSF for a few years, denying bugs with Stallmanesque fervor.
I can go to Redbox, Steve Swasey, you little asshole and get almost 3 DVDs a week!
If want a hundred titles to choose from instead of thousands, then sure, go for it.
Reading the site, the speeds available smell like DSL, and indeed DSL is mentioned in the price list, so I suspect that this is FTTN with DSL local loops. Notably PSTN service is required to order. I'm sure in their area given what's available the service is reasonable for the price, but when talking about FTTH without PSTN, it's not entirely relevant.
I was in a similar situation just a few years ago -- my place at the time was within DSL range of the local site, but Verizon refused into drop a mini DSLAMM into it. Comcast stopped a couple of miles away and wouldn't even return my calls to talk about extending to me. I ended up with ISDN again for a while since it was at least better than analog dialup, and satellite service is useless for interactive work. Later my employer sprung for a DS1, which required a couple of repeaters to be placed and quite a bit of prodding. Trees and terrain didn't allow use of the single wireless provider in the area without a 50' tower. Cell signals were scarce and flaky.
Back in the mid 1980's I spoke with someone who ran what is now called a "data center" for a university. She had previously worked for a bank, and among other tasks she dealt with a line printer spewing some sort of reports or other material -- in all caps. She poke around and found a way to print mixed case instead. She presented this to her boss, who responded "Why would anyone want to do that?". She quit the same day.
Constantly! and it happens even without over-tightening. Other issues with DB cables (including the 50 pin ones used for SCSI):
o Gripping of the cable to the chassis connector is dominated by the hood on the former's connector clamping onto the shell of the latter, much like the way a pump's heel cup is shaped to press against the wearer's heel. This doesn't make for a reliable physical connection.
o The pins in the male DB connectors always have a big of wiggle to them, and are prone to being pushed back into the body of the connector if they bend even a little. This is a serious PITA.
Even the flimsy clips on RJ45 jacks are better than DB connectors, and the SFF-8086 connectors for SAS beat the hell out of them.
Slashdot Mirror
or, admit that anti-aliasing is a euphemism for blur, and rarely improves the appearance of text.
I'm not aware of a jurisdiction that outlaws laptops.
Hardware RAID indeed requires a software mirroring layer on top of multiple HBA's/arrays to be really trustworthy.
We had a single JBOD made by Nstor, who were bought by Xyratex. Xyratex in theory serviced it, but we had to rip the interface board out and send it away for several weeks, no advance or on-site replacement, and when it came back, it was still broken. This array actually had two disks per tray, which meant that for online servicing one had to know that in advance (not always obvious when one is remote) and lay out volumes very carefully. Mind you this wasn't a Xyratex product per se, but the way they handled it didn't impress me. The Linux md driver is meh, but at least it doesn't require compiling metadevice layout into the kernel. It's there and it works, but like SVM/SDS/ODS from 15 years ago it's dated and limited, and mdadmin et al are clumsy at best. "Hardware RAID" means different things to different people. Two main divisions: o HBA RAID. Claimed advantage in that mirror/parity writes don't clog up the host channel. Another is that on systems with anachronistic BIOS stupidity, booting from a mirror when the primary fails can be difficult or often impossible to set up. Big disadvantages: - No model that I've yet seen does 3-way mirrors. None. Disks fail, and more than once I've had a mirror fail while replacing the other side. It can take time to get bad disks replaced, and if remote local hands yank the good side of a mirror pair instead of the bad, you're screwed. - SPoF. Volumes can't span HBA's, and yes, HBA's do go bad. - Crummy monitoring. Some vendors supply CLI's for some OS's, but ongoing support is a uncertain, and the interfaces are always downright horrid. raidctl, for example, on my systems behaves in at least three distinct ways. - Want to use one for the boot device? Be prepared to hit a multi-key sequence during a split-second at the BIOS level after POST. These often assume that you're sitting in front of a desktop, and can enter Alt-keys or function keys to invoke. Using a remote serial console? Sorry, you're screwed. o Chassis RAID: SPoF unless 2-3 identical arrays are mirrored in software. No monitoring to speak of. If you get a serial console, it may need a proprietary adapter cable, null-modem, *and* DB9-RJ45 adapter, as many mistake a 1/8" round headphone-style jack for a serial connector. Linux storage is miserable, still stuck in the mindset of a desktop user with 1-2 drives. The sda/sdb/sdc/etc. naming convention is meant to echo legacy MS-DOS drive letters, and obscures vital information about which drives are in what slot of what array. cXtXdX FTW. When adding an HBA or disks, existing disks can even suffer name changes. ZFS / btrfs is *desperately* needed, but there is no indication of viability anytime soon.
HP assumes that *everything* is a desktop machine plugged into a tube and keyboard, even rackmount systems. They nickle/dime everything they possibly can, including functionality on their HBA's and service processors. Yes, one has to give them ~$150 to unlock features on a FRIGGIN HBA. The third strike: their tech support is offshored, with all the worst features that brings: techs who call during their working hours (not yours), who you can't understand anyway. Techs who don't read the ticket you enter and give non-sensical, non-sequitor answers that are clearly cut/pasted from a list.
A few years ago I was stuck with ISDN for a while at home, which would have sucked for downloading Lion. But even then, I was 10 minutes away from 3-4 coffee places that would have sufficed, and 12 away from a library with excellent speed.
but people using Linux get an inferior experience This is the Linux way, of course. Having recently been forced to start deploying RHEL instead of Solaris, I'm constantly running into shit that's either just stupid, or 15 years out of date.
Therefore there are about 7000/5 = 1400 NetBSD users /mnt
o Systems were extremely prone to wedging, requiring physical intervention
o Systems were especially unstable with >256MB of physmem, and BSDi wouldn't even look at their kernel dumps because they didn't have a machine with more than that themselves to load them into, and refused to get/upgrade one.
o No matter how hard I tried, and no matter what well-regarded mobo I sourced, I could not for the life of me get two SCSI HBA's to co-exist in a single system.
o No software striping/mirroring. In an OS that cost a grand. At one point someone emailed me some old *BSD striping driver that I could hack into the kernel, but it was slow, required that the set of metadevices be compiled and hardcoded into the kernel, including component devices and their sizes.
o Getting stuff to compile and run without dumping immediate core was a nightmare, even worse than Ultrix
I eventually won out against the unrelenting nightmares and installed Solaris on the same systems. They ran faster and stayed up.
Am I the only one who sees the sophistry here? Just in case: o Unspecifed method for counting *BSD posts o Usenet itself has been irrelevant for years, due both to the flood of spammers and migration to web-based forums I argue that *BSD has been irrelevant for years. What innovation, or even keeping up with the times, has it done since the 1980's? In the 1995-1997 timeframe I was stuck with BSD/OS thanks to a decision that preceded me, made by a network engineer of all people. It was a fscking nightmare. The preceding decision maker had asserted that it would be inexpensive by virtue of running on peecee hardware that we could maintain ourselves. In the end the countless hours and karma lost dicking with both the hardware and software were soul-devouring. A few low points: o Rob Kolstad, who believed that home directories -- yes, home directories -- belong in
*All* vivisectionists and torturers are sick. I'm disappointed to read the above, as I'd hoped that some human scum had received some payback.
By and large when I trust other people, they fsck me over, so I can't really blame the gubberment for feeling the same way.
Clearly Europeans aren't big fans of tentacle rape pr0n.
This is much like doing business with HP. Want to actually *use* iLO? Cough up $200+ for the "advanced" license key. Want your Smart Array HBA to do RAID60? Cough up more and enter a license key *onto the frickin' HBA*.
In 1987 or 1986 I was told by a friend that Hurd was going to be released RSN. That never happened but he went on to "work" for the FSF for a few years, denying bugs with Stallmanesque fervor.
I can go to Redbox, Steve Swasey, you little asshole and get almost 3 DVDs a week! If want a hundred titles to choose from instead of thousands, then sure, go for it.
You must be a carnie.
"Criteria" is plural. You want the singular, "criterion". I really must be the only person who understands this.
Thank you for not making me the only one to see this.
Reading the site, the speeds available smell like DSL, and indeed DSL is mentioned in the price list, so I suspect that this is FTTN with DSL local loops. Notably PSTN service is required to order. I'm sure in their area given what's available the service is reasonable for the price, but when talking about FTTH without PSTN, it's not entirely relevant.
I was in a similar situation just a few years ago -- my place at the time was within DSL range of the local site, but Verizon refused into drop a mini DSLAMM into it. Comcast stopped a couple of miles away and wouldn't even return my calls to talk about extending to me. I ended up with ISDN again for a while since it was at least better than analog dialup, and satellite service is useless for interactive work. Later my employer sprung for a DS1, which required a couple of repeaters to be placed and quite a bit of prodding. Trees and terrain didn't allow use of the single wireless provider in the area without a 50' tower. Cell signals were scarce and flaky.
Back in the mid 1980's I spoke with someone who ran what is now called a "data center" for a university. She had previously worked for a bank, and among other tasks she dealt with a line printer spewing some sort of reports or other material -- in all caps. She poke around and found a way to print mixed case instead. She presented this to her boss, who responded "Why would anyone want to do that?". She quit the same day.
Right on! Just like scanning a CD with a laser isn't playing it. Oh, wait ...
This is the guy who pushed NutraSweet through the FDA, after all.
Constantly! and it happens even without over-tightening. Other issues with DB cables (including the 50 pin ones used for SCSI): o Gripping of the cable to the chassis connector is dominated by the hood on the former's connector clamping onto the shell of the latter, much like the way a pump's heel cup is shaped to press against the wearer's heel. This doesn't make for a reliable physical connection. o The pins in the male DB connectors always have a big of wiggle to them, and are prone to being pushed back into the body of the connector if they bend even a little. This is a serious PITA. Even the flimsy clips on RJ45 jacks are better than DB connectors, and the SFF-8086 connectors for SAS beat the hell out of them.
Got any more obscure acronyms to add to make this post even less readable?
If you don't like a given new feature, then DON'T USE IT. How hard is that?