My experience has been that RH support primarily gets one offshored robots.
Me: Foo is broken
Them: We cannot be fixing that because that would be FORKING TEH_DISTRIBUTION
More than once the offshored frontline droids have given me answers that were demonstrably wrong and even dangerous.
A few times with repeated insistence I've been passed along to an engineer on-continent who was able to at least understand the problem.
We're looking at CentOS to avoid the licensing altogether for some systems, but I wonder about the CentOS lag: there have been a couple of times when something broke, notably the kernel, and RedHat got a fix out reasonably quickly. With CentOS, would I have to wait months for such fixes to filter down, so that I could make my systems bootable again? This would be slightly less of a concern if there were any way to revert updates. I really miss Solaris' Live Upgrade.
Also, transit isn't the only cost. Serious routers can easily add up to half a million bux -- I've seen the PO's. Want your provider to be highly available? Need two of those then. And space to run them. Lots of power. Big-ass UPS systems and/or redundant diesel generators on the roof. Talented people to architect and maintain the network. Hell, buildings sometimes even charge monthly fees per *crossconnect*.
Now enter Joe Six Pack who throws up some random "server". He has no concept of "bandwidth" or how much will get used over the course of a month, but you can bet he'll be pissed and tie up CSO resources (which also cost) when he gets charged overage.
Trust me, ISP's are hardly raking it in.
Umm, look up what they culture the stuff in. Plenty of killing went into it. Wake me up when they culture the stuff in something that isn't calf blood.
"Cultured meat is now grown in medium with fetal calf serum, a supplement made from blood collected at slaughterhouses; scientists have yet to find an alternative that doesn't involve dead animals"
My first thought was to think of today's "supercomputers" which are rack upon rack of individual systems clustered, not badass huge SIMD iron like they used to be.
How is it that people *still* don't get this?
You're making a false assumption re "the bandwidth that you've paid for". Residential services are structured -- both pricing and infrastructure -- for typical residential usage patterns. A nominal, say, 10/1 residential connection is for 10/1 to the provider, not 10/1 transit 24x7. Read your contract. Residential end-user customers predominately pull traffic rather than push it, and mostly don't saturate the pipe 24x7. The access link is shared and the contracts tend to make that very clear. Residential customers also by and large don't need or get a static IP address, but I'm sure that the CSO load would be high for someone entering an A record for the address their server happens to get, then whining when it changes. A customer running a server is also going to be way more sensitive to outages.
Business class and leased-line/private circuit services cost more for multiple reasons.
The degree to which that organization counts as a charity is debatable, eg. it's heavy with overcompensated millionaire management who do stuff like $10,000 airline tickets.
The averages in the article's table really don't span a huge range, not even 2:1. Maybe Latvia ranks highly because only two people there have connectivity, and both are 14Mbit.
The accursed things were called thongs or tongs when I was a kid growing up near Pittsburgh. I never heard the term "flip-flop" until college, from a Floridian.
When I visited Wales an English git in our group remarked "I'm dying for a fag". The Welshman I was visiting at one point asked me if I'd like a can. He mean a can of pop, which I didn't initially understand.
No kidding. Sun's x86 hardware kicked ass. Unlike Dell and even HP, Sun actually engineered their x86 servers. They didn't just slap cheap-ass commodity components around the CPU.
The x4NNN systems for sure, we bought scores of them. The earlier models less so. And since you wrote x86 not x64, go way back to the Roadrunners in the late 1980's. x4NNN the serial console works out of the box and can be used to bootstrap the system. HP does that now too, but none of the other big names AFAICT.
And even with the Oracle markup, after my customer had to go out and buy licences for software to manage HP servers - OOOOOPS! (iLO software, etc), the HP boxes turned out to be more expensive than the equivalent Oracle (nee Sun) servers. At least Oracle doesn't charge extra for things like that.
Yep, charging extra for iLO Advanced crippleware is the same nickle & dime crap that's been HP practice for decades. FWIW ISTR that Sun/LSI like HP also charge extra for certain features on their RAID HBA's.
Sun was always something of a hassle to deal with, especially the contract people, but when Oracle took over it degraded substantially. I needed to buy an x4270m2 and it took me something like two months to find someone who was *allowed* to quote it for me, and even then I had to write up a description of why I wanted to buy it so that it could be submitted to Oracle for *permission* to quote.
Not long after that certain happenings forced us to take used HP G6 systems instead of buying anything new - iLO 2 is kind of a dog on them, etc. With the current G8 systems, though, HP has a variety of things right. HBA's have FBWC instead of short-lived batteries, the serial console works out of the box, iLO 4 lets one do virtual media via HTTP... Volume discounts and other strategies help with the extra cost of iLO Advanced and SAAP2.
HP's servers also came with cheap off-brand FC HBAs that wouldn't play nice on the customer's SAN. Good God, crappy FC hardware that can't interoperate with other vendor's equipment was solved by QLogic et al a fucking decade ago.
I'll have to defer to you on FC since we've mostly been able to avoid it. I do know though that HP's switches demand HP-branded transceivers. Luckily I don't have any of those to contend with, and HP's NIC's seem to work fine with third-party modules -- $329 for a 10GE SFP+ vs the $2200+ HP wants.
Lately I've been looking at SuperMicro-based gear as a potential HP alternative, and it's like stepping back in time. 1U systems only have 8 disk bays, no 10 bay option. Only 3Gb/s SAS onboard, need a mezzanine card for 6Gb/s. Only one PCI slot, two if you use the mezzanine slot for a riser. Only two onboard GE NICs. Serial console doesn't work out of the box. No HTTP media redirection on the generic service processor. Firmware updates mostly as bootable.ISO images (see above under "no HTTP media redirection". Feh.
The 1976 Ford Elite and 1977 Mercury Monarch I grew up with sure weren't made to last.
But wrt the Voyagers -- how confident are we that the measuring instruments are still accurate?
Mine works great for my purpose -- playing files over a network mount, something that many competitors (yes, Apple TV, I'm looking at you) can't do. AFP was flaky so I switched to SMB, which helped quite a bit. The only app we use is Netflix -- most of the others seem to basically be interfaces to blogs.
My BB plays the common formats that one finds on the net -- unlike Roku who have some bug up their ass re divx/xvid avi.
At a street worksite last week I saw two guys doing nothing but using a spray bottle and a rag to wipe down pylons.
Hard to believe that union abuse wasn't involved.
I am a consultant, which means that every single one of my clients has decided to outsource some or all of their work. However because I am domestic, that makes it OK?
That depends -- do you have only a single tool, which is a buggy and expensive commercial product sold by your brother's company? Do you slap that into every environment you're hired to work on, then leave the client hosed when you leave? Yes, at one employer I saw another department spend on something called "Tuxedo" stuffed onto a bunch of underpowered Mac II's, then the consultant left with mad $ and nobody in the company had a clue what to do with the stuff.
At a different employer before that one, a consultant came to me one day wanting a DB25 gender changer. Curious WTF this was about I followed him back to the accounting office and found that he was trying to hook some goofy proprietary network into their AT-class systems using the parallel connectors. We kicked the guy out and put ethernet cards into the systems after chastising the accounting droids for not having come to us in the first place.
Re:At least they're not rolling their own.
on
The DNA Data Deluge
·
· Score: 1
I would think that sequence data would reduce especially well with directed compression and especially deduplication.
Slashdot Mirror
My experience has been that RH support primarily gets one offshored robots. Me: Foo is broken Them: We cannot be fixing that because that would be FORKING TEH_DISTRIBUTION More than once the offshored frontline droids have given me answers that were demonstrably wrong and even dangerous. A few times with repeated insistence I've been passed along to an engineer on-continent who was able to at least understand the problem. We're looking at CentOS to avoid the licensing altogether for some systems, but I wonder about the CentOS lag: there have been a couple of times when something broke, notably the kernel, and RedHat got a fix out reasonably quickly. With CentOS, would I have to wait months for such fixes to filter down, so that I could make my systems bootable again? This would be slightly less of a concern if there were any way to revert updates. I really miss Solaris' Live Upgrade.
Actually, marriage is when teh_sex dries up :-/
Someone I used to work with lives in Germany, and when he makes waffles his kids' friends wonder WTF they are.
The first thing that struck me is that "orbit" != "rotation"
Also, transit isn't the only cost. Serious routers can easily add up to half a million bux -- I've seen the PO's. Want your provider to be highly available? Need two of those then. And space to run them. Lots of power. Big-ass UPS systems and/or redundant diesel generators on the roof. Talented people to architect and maintain the network. Hell, buildings sometimes even charge monthly fees per *crossconnect*. Now enter Joe Six Pack who throws up some random "server". He has no concept of "bandwidth" or how much will get used over the course of a month, but you can bet he'll be pissed and tie up CSO resources (which also cost) when he gets charged overage. Trust me, ISP's are hardly raking it in.
lol, how did you get modded insightful for something patently untrue?
You must be new here.
Umm, look up what they culture the stuff in. Plenty of killing went into it. Wake me up when they culture the stuff in something that isn't calf blood.
Ob. Yuggoth reference. Somebody had to say it.
"Cultured meat is now grown in medium with fetal calf serum, a supplement made from blood collected at slaughterhouses; scientists have yet to find an alternative that doesn't involve dead animals"
My first thought was to think of today's "supercomputers" which are rack upon rack of individual systems clustered, not badass huge SIMD iron like they used to be.
yep, just like each previous story here on the SAME THING
It saddens me that this is the only post where "bacterium" is used properly in the singular context, rather than the plural "bacteria".
How is it that people *still* don't get this? You're making a false assumption re "the bandwidth that you've paid for". Residential services are structured -- both pricing and infrastructure -- for typical residential usage patterns. A nominal, say, 10/1 residential connection is for 10/1 to the provider, not 10/1 transit 24x7. Read your contract. Residential end-user customers predominately pull traffic rather than push it, and mostly don't saturate the pipe 24x7. The access link is shared and the contracts tend to make that very clear. Residential customers also by and large don't need or get a static IP address, but I'm sure that the CSO load would be high for someone entering an A record for the address their server happens to get, then whining when it changes. A customer running a server is also going to be way more sensitive to outages. Business class and leased-line/private circuit services cost more for multiple reasons.
No, it doesn't. It might however AFFECT them.
Perhaps they want useful software to run on the things, instead of mostly malware.
1) ZFS snapshots 2) CrashPlan
The degree to which that organization counts as a charity is debatable, eg. it's heavy with overcompensated millionaire management who do stuff like $10,000 airline tickets.
The averages in the article's table really don't span a huge range, not even 2:1. Maybe Latvia ranks highly because only two people there have connectivity, and both are 14Mbit.
The accursed things were called thongs or tongs when I was a kid growing up near Pittsburgh. I never heard the term "flip-flop" until college, from a Floridian. When I visited Wales an English git in our group remarked "I'm dying for a fag". The Welshman I was visiting at one point asked me if I'd like a can. He mean a can of pop, which I didn't initially understand.
No kidding. Sun's x86 hardware kicked ass. Unlike Dell and even HP, Sun actually engineered their x86 servers. They didn't just slap cheap-ass commodity components around the CPU.
The x4NNN systems for sure, we bought scores of them. The earlier models less so. And since you wrote x86 not x64, go way back to the Roadrunners in the late 1980's. x4NNN the serial console works out of the box and can be used to bootstrap the system. HP does that now too, but none of the other big names AFAICT.
And even with the Oracle markup, after my customer had to go out and buy licences for software to manage HP servers - OOOOOPS! (iLO software, etc), the HP boxes turned out to be more expensive than the equivalent Oracle (nee Sun) servers. At least Oracle doesn't charge extra for things like that.
Yep, charging extra for iLO Advanced crippleware is the same nickle & dime crap that's been HP practice for decades. FWIW ISTR that Sun/LSI like HP also charge extra for certain features on their RAID HBA's. Sun was always something of a hassle to deal with, especially the contract people, but when Oracle took over it degraded substantially. I needed to buy an x4270m2 and it took me something like two months to find someone who was *allowed* to quote it for me, and even then I had to write up a description of why I wanted to buy it so that it could be submitted to Oracle for *permission* to quote. Not long after that certain happenings forced us to take used HP G6 systems instead of buying anything new - iLO 2 is kind of a dog on them, etc. With the current G8 systems, though, HP has a variety of things right. HBA's have FBWC instead of short-lived batteries, the serial console works out of the box, iLO 4 lets one do virtual media via HTTP ... Volume discounts and other strategies help with the extra cost of iLO Advanced and SAAP2.
HP's servers also came with cheap off-brand FC HBAs that wouldn't play nice on the customer's SAN. Good God, crappy FC hardware that can't interoperate with other vendor's equipment was solved by QLogic et al a fucking decade ago.
I'll have to defer to you on FC since we've mostly been able to avoid it. I do know though that HP's switches demand HP-branded transceivers. Luckily I don't have any of those to contend with, and HP's NIC's seem to work fine with third-party modules -- $329 for a 10GE SFP+ vs the $2200+ HP wants. Lately I've been looking at SuperMicro-based gear as a potential HP alternative, and it's like stepping back in time. 1U systems only have 8 disk bays, no 10 bay option. Only 3Gb/s SAS onboard, need a mezzanine card for 6Gb/s. Only one PCI slot, two if you use the mezzanine slot for a riser. Only two onboard GE NICs. Serial console doesn't work out of the box. No HTTP media redirection on the generic service processor. Firmware updates mostly as bootable .ISO images (see above under "no HTTP media redirection". Feh.
The 1976 Ford Elite and 1977 Mercury Monarch I grew up with sure weren't made to last. But wrt the Voyagers -- how confident are we that the measuring instruments are still accurate?
Mine works great for my purpose -- playing files over a network mount, something that many competitors (yes, Apple TV, I'm looking at you) can't do. AFP was flaky so I switched to SMB, which helped quite a bit. The only app we use is Netflix -- most of the others seem to basically be interfaces to blogs. My BB plays the common formats that one finds on the net -- unlike Roku who have some bug up their ass re divx/xvid avi.
At a street worksite last week I saw two guys doing nothing but using a spray bottle and a rag to wipe down pylons. Hard to believe that union abuse wasn't involved.
I am a consultant, which means that every single one of my clients has decided to outsource some or all of their work. However because I am domestic, that makes it OK?
That depends -- do you have only a single tool, which is a buggy and expensive commercial product sold by your brother's company? Do you slap that into every environment you're hired to work on, then leave the client hosed when you leave? Yes, at one employer I saw another department spend on something called "Tuxedo" stuffed onto a bunch of underpowered Mac II's, then the consultant left with mad $ and nobody in the company had a clue what to do with the stuff. At a different employer before that one, a consultant came to me one day wanting a DB25 gender changer. Curious WTF this was about I followed him back to the accounting office and found that he was trying to hook some goofy proprietary network into their AT-class systems using the parallel connectors. We kicked the guy out and put ethernet cards into the systems after chastising the accounting droids for not having come to us in the first place.
I would think that sequence data would reduce especially well with directed compression and especially deduplication.