I can't see any of the things you are talking about being done in a centralized government datacenter. For all the paper pushing and website presentation stuff that the government is going to be doing the cloud is a fine solution. I will conceded that there are probably plenty of legacy applications that will need to stay on mainframes and Unix boxes but what percentage of the states datacenter needs are those? Like most things cloud computing is a tool and the job of IT professionals is to find the right tool for the job.
Watt/IOP they crush HDD, Watt/GB the opposite is true. I use my SAN for a heck of a lot more than just database so I need a much more balanced approach. I have VM's, email, bulk file storage, content management, and various drives from application servers all mounted in the same array. If you're big enough to have arrays dedicated to just database then for sure SSD's are the future for that niche, but I doubt that's more than 25% of the SAN market.
Except with MS or Google you can rent space in multiple disparate datacenters which is better than running your own single large one. The problem with cloud computing is that it assumes every application is horizontally scalable and available on open systems platforms (IE Windows or Linux).
and realize that some projects are too wide for web applications.
Like what exactly? The biggest applications I have experience with are all n-tier web fronted applications. Our ERP, ECM, BI, and CRM* are all web based and so are almost every app in those product spaces. The only apps that I am aware of that are large and not necessarily improved by using a web frontend are greenscreen apps like SABRE access or something like state DMV systems and you can generally make those snappy enough with modern web technologies to be acceptable.
*:
ERP= Enterprise Resource Planning (JD Edwards, SAP, MS Dynamix, Peoplesoft)
ECM= Enterprise Content Management (Livelink, Documentum, Sharepoint)
BI= Business Intelligence (Oracle BI, Cognos, Hyperion, Business Objects)
CRM= Customer Relationship management (salesforce.com, MS CRM, Siebel)
I bet one drive could saturate the PERC 6i, I know it can saturate the HP P400 no problem. In fact I got about 2x the 4k random write IOPS when I used it in a workstation with Intel ICH as I did when it was connected to the P400.
The x-25e is great, and I use it in a few situations, but at 8x the cost per GB of 15k FC I'm not moving to it wholesale. It's true that for $10K I could get as many IOPS as my $200K EVA, but it would only have the storage of a single drive in the array.
Basically every colo datacenter is metered. Standard is 2x20A circuits (16A max draw) per rack. If you want 30A or quad 20A circuits in your rack you generally pay a hefty overage charge. At our DR colo provider they require metered PDU's. I have this capping capability with my HP servers which means I can fill a rack right up to the edge but make sure I don't overload the circuit by keeping my peak usage closer to average. You also need to use staggered startup on your servers if you want to play it that close to your power cap.
Those are insanely expensive and only store enough power for a few MW load for a handful of minutes. In theory they are better than a UPS in the long run due to not needing to replace batteries every few years but they aren't going to shift much load from offpeak to peak.
I'm worried about the heat from the leaking laser damaging the inside of the jacket and making the fiberoptics not work anymore. The spot temperature where say 20W of light dissipates in 65nm of surface area has got to be huge.
In the last 3 years I have replaced 2 GBIC's in our 6509, 5 in our 4500's, 6 in a couple different fiberchannel switches from 2 vendors, 3 in disk shelves in our SAN, and replaced a couple HBA's that were probably not working due to laser death. In that same time I have replaced 3 fans across ~150 servers and none in all my networking gear.
Loss=heat, I doubt the cladding would stand up very well to that kind of loss.
In my experience solid state lasers aren't very reliable even at fairly low power. In networking gear GBIC's/SFP's are by far the least reliable components, dying far more often then even mechanical components like fans and probably on par with enterprise HDD's.
Servers=static
Printers=static
network gear=static
random network devices=static
In any non-trivial network you will have a significant percentage of your IP space utilized by static devices. Then you get into tracking multiple sites and their associated network information and it starts to get fairly complicated. We're small enough with a couple dozen sites and a dozen or so subnets at our corporate campus that we use a multi-tabbed excel document with the first tab being a table of contents.
How? Smash the glass and physically or chemically separate the Uranium. The only reason they do glassification is to keep the material from becoming soluble in water that might one day enter the site.
That doesn't debunk anything, using 5,500 MWhr's to enrich the fuel to produce 8,760,000 MWhr's sounds like a huge freaking reduction to me! As to their claim that there's only 50 years of uranium, that's so far off from reality that I don't know how to respond. Oh and the majority of the costs of opening a new plant are government imposed so it's kind of hard to compare them to anything free-market add to that the fact that fossil fuel prices don't currently have their externality costs priced into the consumption price and making a comparison is almost impossible.
Nah, the on-site storage just ups the chance of leakage into the environment. Yucca is a fine place to keep the waste until we get our head out of the sand and accept breeder designs. The material is going to be transported eventually since it's highly unlikely the will build new breeder reactors on the same sites as the existing facilities (lack of sufficient thermal sinks is the main design concern).
We have a system for ascertaining guilt or innocence, it's called trial by jury. Also there should be a limit to the length of punishment for a single act of simple theft, and it's probably less than 14 years.
Except you aren't found guilty of contempt, you are ruled in contempt by a judge. That is a very important distinction, and one that should not exist. If you are to be put in jail for more than 30 days it should require a jury of your peers to find you guilty beyond a reasonable doubt.
Prometheus can patent their particular test for a given attribute but patenting the process of testing for that attribute is classically a no-no. Basically your unique invention can be patented but not the idea behind it. Then another clever person takes your idea and uses it to make a better/faster/cheaper test, that is how patents spur the progress of science and the useful arts. If they don't like the limitations of patent protection then they can attempt to keep the invention a trade secret.
Exactly, recently a Cleveland Browns football player got 30 days for vehicular manslaughter (he got trashed and killed someone) and this guy gets 14 years for not having money to give to his ex, that's just a travesty of justice pure and simple.
It's seriously scary that this guy was basically in debtors prison for 14 years! Wasn't it a bit obvious after say 6-12 months that the guy either didn't have the money or wasn't going to ever hand it over? I can't believe that the original judge thought it was fair and ok to keep this guy in jail for over a decade on the assumption (not proof) that he was lying.
Oh, it's definitely not HP, they show a nice shiny c-class enclosure in their presentation and if you look in this thread I posted the current code for iLo/OA from HP.
I doubt Opera is buying enough servers to matter. They show a few racks of blade enclosures, big whoop. The expense of having someone QA all of the software on the server to make sure it works with the customers software would wipe out any profit from the sale. And let's say that they DID do the tests and found the problems, what then? Are you really going to redevelop and retest the software for an entire server line just to win a small contract? I very highly doubt it. If opera were big enough to keep a factory busy for a couple shifts it might be a different story, but the fact is they are small fish and the costs associated with winning the contract would never work out. I have much more problem with the general state of things in the industry that led to the problem then I do with the lack of testing on the part of the vendor.
Slashdot Mirror
I can't see any of the things you are talking about being done in a centralized government datacenter. For all the paper pushing and website presentation stuff that the government is going to be doing the cloud is a fine solution. I will conceded that there are probably plenty of legacy applications that will need to stay on mainframes and Unix boxes but what percentage of the states datacenter needs are those? Like most things cloud computing is a tool and the job of IT professionals is to find the right tool for the job.
Watt/IOP they crush HDD, Watt/GB the opposite is true. I use my SAN for a heck of a lot more than just database so I need a much more balanced approach. I have VM's, email, bulk file storage, content management, and various drives from application servers all mounted in the same array. If you're big enough to have arrays dedicated to just database then for sure SSD's are the future for that niche, but I doubt that's more than 25% of the SAN market.
Except with MS or Google you can rent space in multiple disparate datacenters which is better than running your own single large one. The problem with cloud computing is that it assumes every application is horizontally scalable and available on open systems platforms (IE Windows or Linux).
and realize that some projects are too wide for web applications.
Like what exactly? The biggest applications I have experience with are all n-tier web fronted applications. Our ERP, ECM, BI, and CRM* are all web based and so are almost every app in those product spaces. The only apps that I am aware of that are large and not necessarily improved by using a web frontend are greenscreen apps like SABRE access or something like state DMV systems and you can generally make those snappy enough with modern web technologies to be acceptable.
*:
ERP= Enterprise Resource Planning (JD Edwards, SAP, MS Dynamix, Peoplesoft)
ECM= Enterprise Content Management (Livelink, Documentum, Sharepoint)
BI= Business Intelligence (Oracle BI, Cognos, Hyperion, Business Objects)
CRM= Customer Relationship management (salesforce.com, MS CRM, Siebel)
I bet one drive could saturate the PERC 6i, I know it can saturate the HP P400 no problem. In fact I got about 2x the 4k random write IOPS when I used it in a workstation with Intel ICH as I did when it was connected to the P400.
The x-25e is great, and I use it in a few situations, but at 8x the cost per GB of 15k FC I'm not moving to it wholesale. It's true that for $10K I could get as many IOPS as my $200K EVA, but it would only have the storage of a single drive in the array.
Basically every colo datacenter is metered. Standard is 2x20A circuits (16A max draw) per rack. If you want 30A or quad 20A circuits in your rack you generally pay a hefty overage charge. At our DR colo provider they require metered PDU's. I have this capping capability with my HP servers which means I can fill a rack right up to the edge but make sure I don't overload the circuit by keeping my peak usage closer to average. You also need to use staggered startup on your servers if you want to play it that close to your power cap.
Those are insanely expensive and only store enough power for a few MW load for a handful of minutes. In theory they are better than a UPS in the long run due to not needing to replace batteries every few years but they aren't going to shift much load from offpeak to peak.
I'm worried about the heat from the leaking laser damaging the inside of the jacket and making the fiberoptics not work anymore. The spot temperature where say 20W of light dissipates in 65nm of surface area has got to be huge.
In the last 3 years I have replaced 2 GBIC's in our 6509, 5 in our 4500's, 6 in a couple different fiberchannel switches from 2 vendors, 3 in disk shelves in our SAN, and replaced a couple HBA's that were probably not working due to laser death. In that same time I have replaced 3 fans across ~150 servers and none in all my networking gear.
Loss=heat, I doubt the cladding would stand up very well to that kind of loss. In my experience solid state lasers aren't very reliable even at fairly low power. In networking gear GBIC's/SFP's are by far the least reliable components, dying far more often then even mechanical components like fans and probably on par with enterprise HDD's.
Bingo, KISS. It's also why I like static routes in a small to midsized environment, there's no additional moving parts to potentially break.
Landesk has Mac and Linux clients, which is one of the big things that makes it stand out in it's market segment.
Servers=static
Printers=static
network gear=static
random network devices=static
In any non-trivial network you will have a significant percentage of your IP space utilized by static devices. Then you get into tracking multiple sites and their associated network information and it starts to get fairly complicated. We're small enough with a couple dozen sites and a dozen or so subnets at our corporate campus that we use a multi-tabbed excel document with the first tab being a table of contents.
How? Smash the glass and physically or chemically separate the Uranium. The only reason they do glassification is to keep the material from becoming soluble in water that might one day enter the site.
That doesn't debunk anything, using 5,500 MWhr's to enrich the fuel to produce 8,760,000 MWhr's sounds like a huge freaking reduction to me! As to their claim that there's only 50 years of uranium, that's so far off from reality that I don't know how to respond. Oh and the majority of the costs of opening a new plant are government imposed so it's kind of hard to compare them to anything free-market add to that the fact that fossil fuel prices don't currently have their externality costs priced into the consumption price and making a comparison is almost impossible.
Nah, the on-site storage just ups the chance of leakage into the environment. Yucca is a fine place to keep the waste until we get our head out of the sand and accept breeder designs. The material is going to be transported eventually since it's highly unlikely the will build new breeder reactors on the same sites as the existing facilities (lack of sufficient thermal sinks is the main design concern).
We have a system for ascertaining guilt or innocence, it's called trial by jury. Also there should be a limit to the length of punishment for a single act of simple theft, and it's probably less than 14 years.
She would be jailed if found guilty
Except you aren't found guilty of contempt, you are ruled in contempt by a judge. That is a very important distinction, and one that should not exist. If you are to be put in jail for more than 30 days it should require a jury of your peers to find you guilty beyond a reasonable doubt.
Prometheus can patent their particular test for a given attribute but patenting the process of testing for that attribute is classically a no-no. Basically your unique invention can be patented but not the idea behind it. Then another clever person takes your idea and uses it to make a better/faster/cheaper test, that is how patents spur the progress of science and the useful arts. If they don't like the limitations of patent protection then they can attempt to keep the invention a trade secret.
Exactly, recently a Cleveland Browns football player got 30 days for vehicular manslaughter (he got trashed and killed someone) and this guy gets 14 years for not having money to give to his ex, that's just a travesty of justice pure and simple.
It's seriously scary that this guy was basically in debtors prison for 14 years! Wasn't it a bit obvious after say 6-12 months that the guy either didn't have the money or wasn't going to ever hand it over? I can't believe that the original judge thought it was fair and ok to keep this guy in jail for over a decade on the assumption (not proof) that he was lying.
I meant standard for that vendor, IE the approved code load for the onboard admin processor.
Oh, it's definitely not HP, they show a nice shiny c-class enclosure in their presentation and if you look in this thread I posted the current code for iLo/OA from HP.
I doubt Opera is buying enough servers to matter. They show a few racks of blade enclosures, big whoop. The expense of having someone QA all of the software on the server to make sure it works with the customers software would wipe out any profit from the sale. And let's say that they DID do the tests and found the problems, what then? Are you really going to redevelop and retest the software for an entire server line just to win a small contract? I very highly doubt it. If opera were big enough to keep a factory busy for a couple shifts it might be a different story, but the fact is they are small fish and the costs associated with winning the contract would never work out. I have much more problem with the general state of things in the industry that led to the problem then I do with the lack of testing on the part of the vendor.