Can you plug 36 Ethernet adapters to handle the same aggregated bandwidth as a cluster does?
The 15K has up to 72 PCI slots on 18 channels, according to the Sun website. They say the overall maximum I/O bandwidth is 21.6 GB/sec, which is more than adequate for a full complement of Gigabit Ethernet adapters. However, I'd bet a few slots are needed for FibreChannel to a storage array, so perhaps only a maximum of 60 or so Ethernet adapters could be installed.
After seeing this Slashdot article, I went and read the RAC whitepaper at Oracle's website. Their cluster architecture is more compelling than I anticipated. It literally uses the cluster nodes just for CPU/RAM and uses a shared storage model (e.g., FibreChannel SAN).
Clustering still does not allow for ultra-cheap hardware, however, because ECC on RAM and busses is critical for basic integrity. The $2,500 per node numbers floating around above are very optimistic, where $4,000 to $8,000 per node looks more likely (based on PenguinComputing's site and Sun's on-line store). Opteron-based servers are a shoe-in for an Oracle cluster. The latest Sun Fire V210 servers are also worth considering, because each one comes with four built-in Gigbabit adapters and an optional SSL accelerator for client connections.
Lisa: But I don't see any tigers around, do you? Homer: Lisa, I want to buy your rock.
Additionally, given the immense inertia of the government, could the Patriot Act even have an effect by now? My guess is that any successful intercepts of terrorist plans recently are still done the same way they would have been done five or ten years ago.
A good example of the inertia would be the Department of Homeland Security. They are progressing towards their goals, but I wouldn't be suprised if another decade goes by before any changes have really become effective. There are just too many people, too many departments, too many systems, etc.
Like I said in my other reply, you completely flew by the point I was making and began foaming your myopic zealot-speak.
Again, my point is that there are so many costs involved that the cost of the hardware itself isn't that big of a deal. Unless, of course, a business is so unsucessful that even buying donuts for the break room requires issuing bonds.
a 64 node cluster of linux boxes at $2500 each is still a savings of $1.74 million over a 36 way Sunfire 15K
The Sun Fire 15K can maintain 43.2 GB/sec bandwidth connecting all the CPUs (see here). In a cluster, a small multiple of 0.128 GB/sec for Gigabit Ethernet is about all you'd get.
Regardless, you missed my point, because I explicitly mentioned Sun's low-end servers (2 CPUs). Even though these do cost more per node than the PCs, the cost difference is not nearly as dramatic as you were trying to make it out to be.
Also, don't forget that storage costs are basically constant between the systems. The only real variable is the cost for each CPU-RAM bundle. Hardly groundshaking.
I recommend putting trays of popcorn on top of each rack-mount server. You can spot which ones are too hot almost immediately and get some needed brain food during the subsequent troubleshooting. The unbeatable feature of this monitoring method is the use of high-reliability redundant data paths. Visual, aural, and olfactory transmission methods are all employed to their fullest advantage.
I wonder if Oracle is still motivated by per-CPU licensing. Even if the hardware and OS are $2,500 per node, will Oracle still be another ten grand per node? And it would probably be wise, for responsiveness/throughput, to have two CPUs in each node. This can get expensive very quickly.
Add the fact that you still need a paid Oracle DBA to keep everything running smoothly, and I'd bet the real savings of using a Linux cluster is relatively small. Add annual support for Linux itself, and the relative savings are even smaller.
I think the people who feel they got burned by Sun or IBM or H-Paq on hardware costs naively caved into their salespeople and want to steer the blame somewhere else. Two or three dual-CPU Sun Fire servers, for example, can push enough throughput for most applications I can imagine for any but the largest businesses. How many Googles and NYSEs are there in the grand scheme of things?
SPARC/Power/PA-RISC/MIPS vs. x86/Linux is, in reality, a pretty small part of the big picture, IMO. All this versus Opteron, however, is probably a different story (AMD really knows how to stir the pot, don't they!).
But who is responsible for broke code? Surely not the consultants.
You should read a commercial software EULA some time, because you might be in for a suprise. I'd bet you have about as much recourse for broke code with Oracle as you do with MySQL.
The people who claim that an angel from Oracle decends down to magically fix their problems probably aren't aware that their boss just signed away another $10,000 in support fees. Running a fully supported Oracle server is expensive!
Take a look at how MS SQL Server performs clustering sometime. Everything (and I mean EVERYTHING) is performed via triggers and tsql.
Does MS SQL put those triggers into your applications database tables or in some sort of separate system table? If clustering in MS SQL requires mucking up the application's own data structures, then MS SQL must be worse than I already thought.
5. Voluminous reprints of public domain and easily accessible information.
Linux Unleashed was the first (and last) "big" tech book I've bought. That book turned out to be a simple reprint of the man pages that led me to look for a book in the first place. I regretted buying that book so much that it really helped me set my priorities for later purchases.
Instead of trying to find magical software, why not hire a really good management staff and reward them appropriately? I'd bet the cost is actually lower and doesn't reek of proprietary lock-in, five-year upgrade cycles, added technical support staff, and perpetual confusion over what the system is actually supposed to do.
I say proprietary lock-in above, because I think the ability of an Open Source project to organize and capture proprietary business processes is debatable. Open Source tends to have an outside perspective with respect to commercial business.
And, seriously, what "small to medium business" business processes can't be handed by e-mail and simple internal websites. This can even be done without Oracle (the horror!) or.NET/J2EE (think of the children!).
Are you familiar with the term "drag coefficient [business2.com]" with a slightly revised definition?
I glanced at that article, and the fact that anyone can think in those terms is sickening. Why should any industry punish those people who don't want to be enslaven to it? Is this a return to 19th century industrialization?
A person could spend seventy hours a week just keeping up with the artificial barrage of IT buzzwords. It isn't worth it. If a company can't train its employees on the buzzwords it feels are important, then that company doesn't have its priorities straight.
Agreed. However, has the drop in income for these people been matched with drops in the cost of living? I think a lot of people are seeing the simple dream of a house, a backyard, a car, and a dog evaporating before them.
Outside of a downtown, should families really be raised in two-bedroom apartments with thin walls? Should they be denied owning real estate (often the only stable investment available)? Should we drop our standards so low that living in tenements and corporate housing are again acceptable?
Our standards of living are the only real tangible measure of progress and success we have. If it continues to get harder for a single income to support a family and a modest home, then I say we are truly regressing, and progress is a myth.
For anyone who will whine, "look at the third world, whine, whine, we have it great," I respond, "we could easily end up just like them if we let housing and medical costs continue unchecked." Standards of living are more fragile than you might think.
Do the requirements necessary in the manufacture LEDs and LED componants out-weight the requirements for standard bulbs over 10 years?
Have faith in the free market. From a cost-payoff point of view, the only thing you should care about is the price you see on the shelf at Wal-Mart. Built into this price are all the costs associated with manufacturing, packaging, distributing, and selling the LED.
From an environmental point-of-view, I really don't know. The fact that LEDs are so durable and compact might mean a lower long-term impact, simply because the volume of manufacturing waste and consumer waste will be lower. Whether LEDs are more polluting per manufactured unit of weight, we'd probably never know due to trade secrets.
E.g., the main central ceiling light in your RV should probably remain fluorescent, but nook and cranny lights can certainly go LED.
Quite insightful:), but another idea would be to set up LEDs in a grid on the cieling. Perhaps each LED could be in its own little canister recess. The coolness factor would outweigh any inefficiencies, then.
...as soon as the LEDs get a scosh cheaper in a normal 12 vdc config for area lighting I will switch to them instead.
You might not want to be hasty in switching from fluorescent to LED. In the several Slashdot articles about LEDs recently, the consensus is that fluorescents are actually more effecient for area lighting. E.g., the main central ceiling light in your RV should probably remain fluorescent, but nook and cranny lights can certainly go LED. Combining both technologies would probably give you the biggest bang for the watt.
Yup. While bounded arrays help remove certain classes of bugs, for example, the "null pointer exception" stepped in exactly where "segmentation fault" left.
The complexity of programming has been pretty constant since FORTRAN was adopted over assembler. "Object-oriented" removed some complexity but added its own, so the net gain is small. Maybe "aspects" will finally save us...maybe not.
Automated tools like lint are an invaluable part of any software project and should be used at various points in a project's lifecycle. However, the bugs they find tend to be shallow (typecast problems, immediate memory violations, etc.). This is a certain improvement in software quality, but even Java programs can have side-effects from class to class that twist the mind of even the best programmer.
"Finding bugs is easy" makes sense in the context of my backyard but definitely not in programming.
If it is so hard to get a IT job with say 8 years of experience, imagine getting another where you have zero years.
There are some non-technical non-IT jobs that are well-served by people with technical backgrounds, such as consulting or writing. However, it may be the case that genuine engineering or science might be more appropriate than IT for these positions.
Perhaps retail or carpentry or plumbing or something is the best bet.
Retail would be fairly easy to get into, but professional carpentry and plumbing tend to be well-organized around trade unions and are also regulated. In some states where the unions don't have a foot-hold, I suppose anyone can take on these jobs, however (I recommend the south-east U.S. for low-quality statisfaction:)
...they use repeatable scientific methods to test the products.
This is a pretty strong statement. Do you really think their methods would stand up to a peer review?
They basically take a few products, look them over, poke at them, collect some "data", and, then, publish it. I agree they sometimes do an adequate job in evaluating something, but to claim they produce real science is laughable.
You can't tell us we're wrong, because Consumer Reports said so, and expect to be believed by any reasonable human being.
From what I've seen, Consumer Reports' "scientific methods" are often quite subjective. I read a tire review a while ago that left me baffled, but, then, I realized their results were true given their choice of measurement technique. Consumer Reports tends to project their opinions into their methods rather than their writing, which can actually mislead the reader.
I generally accept that Consumer Reports does good reporting when they have huge aggregate datasets for things like cars and appliances but question their reporting when they do their own home-brew experiments on a small number of products. The home-brew experiments really just end up being like the product reviews in the Sunday newspaper (unless there are blatant pros or cons to something, the review is worthless due to the author's bias).
Slashdot Mirror
Can you plug 36 Ethernet adapters to handle the same aggregated bandwidth as a cluster does?
The 15K has up to 72 PCI slots on 18 channels, according to the Sun website. They say the overall maximum I/O bandwidth is 21.6 GB/sec, which is more than adequate for a full complement of Gigabit Ethernet adapters. However, I'd bet a few slots are needed for FibreChannel to a storage array, so perhaps only a maximum of 60 or so Ethernet adapters could be installed.
After seeing this Slashdot article, I went and read the RAC whitepaper at Oracle's website. Their cluster architecture is more compelling than I anticipated. It literally uses the cluster nodes just for CPU/RAM and uses a shared storage model (e.g., FibreChannel SAN).
Clustering still does not allow for ultra-cheap hardware, however, because ECC on RAM and busses is critical for basic integrity. The $2,500 per node numbers floating around above are very optimistic, where $4,000 to $8,000 per node looks more likely (based on PenguinComputing's site and Sun's on-line store). Opteron-based servers are a shoe-in for an Oracle cluster. The latest Sun Fire V210 servers are also worth considering, because each one comes with four built-in Gigbabit adapters and an optional SSL accelerator for client connections.
The tinfoil hat crowd may think the NSA/CIA/FBI is monitoring all of their phone and computer communications but, really, there just isn't the....
ICARUS HAS FOUND YOU!!!! RUN WHILE YOU CAN!!!!
Lisa: But I don't see any tigers around, do you?
Homer: Lisa, I want to buy your rock.
Additionally, given the immense inertia of the government, could the Patriot Act even have an effect by now? My guess is that any successful intercepts of terrorist plans recently are still done the same way they would have been done five or ten years ago.
A good example of the inertia would be the Department of Homeland Security. They are progressing towards their goals, but I wouldn't be suprised if another decade goes by before any changes have really become effective. There are just too many people, too many departments, too many systems, etc.
...at half price, for 4 times as many cpus.
Like I said in my other reply, you completely flew by the point I was making and began foaming your myopic zealot-speak.
Again, my point is that there are so many costs involved that the cost of the hardware itself isn't that big of a deal. Unless, of course, a business is so unsucessful that even buying donuts for the break room requires issuing bonds.
a 64 node cluster of linux boxes at $2500 each is still a savings of $1.74 million over a
36 way Sunfire 15K
The Sun Fire 15K can maintain 43.2 GB/sec bandwidth connecting all the CPUs (see here). In a cluster, a small multiple of 0.128 GB/sec for Gigabit Ethernet is about all you'd get.
Regardless, you missed my point, because I explicitly mentioned Sun's low-end servers (2 CPUs). Even though these do cost more per node than the PCs, the cost difference is not nearly as dramatic as you were trying to make it out to be.
Also, don't forget that storage costs are basically constant between the systems. The only real variable is the cost for each CPU-RAM bundle. Hardly groundshaking.
I recommend putting trays of popcorn on top of each rack-mount server. You can spot which ones are too hot almost immediately and get some needed brain food during the subsequent troubleshooting. The unbeatable feature of this monitoring method is the use of high-reliability redundant data paths. Visual, aural, and olfactory transmission methods are all employed to their fullest advantage.
...with up to 64 nodes in the cluster.
I wonder if Oracle is still motivated by per-CPU licensing. Even if the hardware and OS are $2,500 per node, will Oracle still be another ten grand per node? And it would probably be wise, for responsiveness/throughput, to have two CPUs in each node. This can get expensive very quickly.
Add the fact that you still need a paid Oracle DBA to keep everything running smoothly, and I'd bet the real savings of using a Linux cluster is relatively small. Add annual support for Linux itself, and the relative savings are even smaller.
I think the people who feel they got burned by Sun or IBM or H-Paq on hardware costs naively caved into their salespeople and want to steer the blame somewhere else. Two or three dual-CPU Sun Fire servers, for example, can push enough throughput for most applications I can imagine for any but the largest businesses. How many Googles and NYSEs are there in the grand scheme of things?
SPARC/Power/PA-RISC/MIPS vs. x86/Linux is, in reality, a pretty small part of the big picture, IMO. All this versus Opteron, however, is probably a different story (AMD really knows how to stir the pot, don't they!).
But who is responsible for broke code? Surely not the consultants.
You should read a commercial software EULA some time, because you might be in for a suprise. I'd bet you have about as much recourse for broke code with Oracle as you do with MySQL.
The people who claim that an angel from Oracle decends down to magically fix their problems probably aren't aware that their boss just signed away another $10,000 in support fees. Running a fully supported Oracle server is expensive!
Take a look at how MS SQL Server performs clustering sometime. Everything (and I mean EVERYTHING) is performed via triggers and tsql.
Does MS SQL put those triggers into your applications database tables or in some sort of separate system table? If clustering in MS SQL requires mucking up the application's own data structures, then MS SQL must be worse than I already thought.
.NET is the future [of Microsoft], but it's not here yet.
I suspect this premise is true for all values of t.
"Visual Studio .NET 2003."
.NET 2003 Home Edition 7.0.
I'm holding out for Visual Studio Advanced Professional Server
Those big books are hugely padded by: ...
5. Voluminous reprints of public domain and easily accessible information.
Linux Unleashed was the first (and last) "big" tech book I've bought. That book turned out to be a simple reprint of the man pages that led me to look for a book in the first place. I regretted buying that book so much that it really helped me set my priorities for later purchases.
Instead of trying to find magical software, why not hire a really good management staff and reward them appropriately? I'd bet the cost is actually lower and doesn't reek of proprietary lock-in, five-year upgrade cycles, added technical support staff, and perpetual confusion over what the system is actually supposed to do.
.NET/J2EE (think of the children!).
I say proprietary lock-in above, because I think the ability of an Open Source project to organize and capture proprietary business processes is debatable. Open Source tends to have an outside perspective with respect to commercial business.
And, seriously, what "small to medium business" business processes can't be handed by e-mail and simple internal websites. This can even be done without Oracle (the horror!) or
Are you familiar with the term "drag coefficient [business2.com]" with a slightly revised definition?
I glanced at that article, and the fact that anyone can think in those terms is sickening. Why should any industry punish those people who don't want to be enslaven to it? Is this a return to 19th century industrialization?
A person could spend seventy hours a week just keeping up with the artificial barrage of IT buzzwords. It isn't worth it. If a company can't train its employees on the buzzwords it feels are important, then that company doesn't have its priorities straight.
Its capitalism at is best!
Agreed. However, has the drop in income for these people been matched with drops in the cost of living? I think a lot of people are seeing the simple dream of a house, a backyard, a car, and a dog evaporating before them.
Outside of a downtown, should families really be raised in two-bedroom apartments with thin walls? Should they be denied owning real estate (often the only stable investment available)? Should we drop our standards so low that living in tenements and corporate housing are again acceptable?
Our standards of living are the only real tangible measure of progress and success we have. If it continues to get harder for a single income to support a family and a modest home, then I say we are truly regressing, and progress is a myth.
For anyone who will whine, "look at the third world, whine, whine, we have it great," I respond, "we could easily end up just like them if we let housing and medical costs continue unchecked." Standards of living are more fragile than you might think.
Do the requirements necessary in the manufacture LEDs and LED componants out-weight the requirements for standard bulbs over 10 years?
Have faith in the free market. From a cost-payoff point of view, the only thing you should care about is the price you see on the shelf at Wal-Mart. Built into this price are all the costs associated with manufacturing, packaging, distributing, and selling the LED.
From an environmental point-of-view, I really don't know. The fact that LEDs are so durable and compact might mean a lower long-term impact, simply because the volume of manufacturing waste and consumer waste will be lower. Whether LEDs are more polluting per manufactured unit of weight, we'd probably never know due to trade secrets.
E.g., the main central ceiling light in your RV should probably remain fluorescent, but nook and cranny lights can certainly go LED.
:), but another idea would be to set up LEDs in a grid on the cieling. Perhaps each LED could be in its own little canister recess. The coolness factor would outweigh any inefficiencies, then.
Quite insightful
...as soon as the LEDs get a scosh cheaper in a normal 12 vdc config for area lighting I will switch to them instead.
You might not want to be hasty in switching from fluorescent to LED. In the several Slashdot articles about LEDs recently, the consensus is that fluorescents are actually more effecient for area lighting. E.g., the main central ceiling light in your RV should probably remain fluorescent, but nook and cranny lights can certainly go LED. Combining both technologies would probably give you the biggest bang for the watt.
"||gazm"
What a perfect name for a logic-based language or boolean calculator!
Bugs in Java code?
Yup. While bounded arrays help remove certain classes of bugs, for example, the "null pointer exception" stepped in exactly where "segmentation fault" left.
The complexity of programming has been pretty constant since FORTRAN was adopted over assembler. "Object-oriented" removed some complexity but added its own, so the net gain is small. Maybe "aspects" will finally save us...maybe not.
Why not call it "Lint"?
Automated tools like lint are an invaluable part of any software project and should be used at various points in a project's lifecycle. However, the bugs they find tend to be shallow (typecast problems, immediate memory violations, etc.). This is a certain improvement in software quality, but even Java programs can have side-effects from class to class that twist the mind of even the best programmer.
"Finding bugs is easy" makes sense in the context of my backyard but definitely not in programming.
Oil Fields can feed all of Iraq; it's the company's meal ticket.
Halliburton's?
If it is so hard to get a IT job with say 8 years of experience, imagine getting another where you have zero years.
There are some non-technical non-IT jobs that are well-served by people with technical backgrounds, such as consulting or writing. However, it may be the case that genuine engineering or science might be more appropriate than IT for these positions.
Perhaps retail or carpentry or plumbing or something is the best bet.
Retail would be fairly easy to get into, but professional carpentry and plumbing tend to be well-organized around trade unions and are also regulated. In some states where the unions don't have a foot-hold, I suppose anyone can take on these jobs, however (I recommend the south-east U.S. for low-quality statisfaction:)
...they use repeatable scientific methods to test the products.
This is a pretty strong statement. Do you really think their methods would stand up to a peer review?
They basically take a few products, look them over, poke at them, collect some "data", and, then, publish it. I agree they sometimes do an adequate job in evaluating something, but to claim they produce real science is laughable.
You can't tell us we're wrong, because Consumer Reports said so, and expect to be believed by any reasonable human being.
From what I've seen, Consumer Reports' "scientific methods" are often quite subjective. I read a tire review a while ago that left me baffled, but, then, I realized their results were true given their choice of measurement technique. Consumer Reports tends to project their opinions into their methods rather than their writing, which can actually mislead the reader.
I generally accept that Consumer Reports does good reporting when they have huge aggregate datasets for things like cars and appliances but question their reporting when they do their own home-brew experiments on a small number of products. The home-brew experiments really just end up being like the product reviews in the Sunday newspaper (unless there are blatant pros or cons to something, the review is worthless due to the author's bias).