Compressed air is the key though... not the lack of tap water or soap. Many places use nitrogen instead of compressed air, but either way you have a very clean, dry airstream to clean it.
Enter "Pete the Pirate". He's using the bandwidth in full and he won't fit in that normal distribution. The nice normal distribution turns skewed to the right, everyone gets worse response times and less bandwidth on average. The solution? Sell everyone guaranteed 10M/512k or what? Most of the people don't want to pay 60 times as much as they do because they don't have the need for guaranteed bandwidth. ISDN was about fixed bandwidth and it sucked. Nobody needed that bandwidth that much and therefore the costs were significantly higher than with ADSL technologies.
Solution: Transfer based billing. I think the sender should pay for the bandwidth as it is with the web sites as well. Your incoming traffic requires also outgoing traffic and you attach the interest of the company (build as little infrastructure as economically feasible) with the interest of the client (use that infrastructure as little as economically feasible).
The problem with that logic is that the statistical average of all users is pushed up by "Peter." He might not fit into the old distribution, but he is a part of the new one. As Quincy, Robert, Sam and Tom all begin to have similar usage patterns, the average usage begins to fit more closely Peter's usage.
The ISP needs to adjust their models to reflect these changes over time.
Personally, I would prefer for an ISP to tier levels of service and commit to a contention ratio they can afford. If a user exceeds the preset contention ratio for their package over a 7 or 30 day period, they are bumped into the next tier after a warning. Start out with a 512k, 1% contention which should be adequate for most users (ends up at 1.5G/month), then go to a 1.024M, 2% (6G/month), 2M, 5% (30GB/month), 6M, 10%...
Tie the sense of value (bandwidth) into the true cost (transfer), and give the ISP the incentive to improve over time as well as give the customer an incentive to buy into a higher package. If internet TV takes off (for example), over time a market is created for improvements...
As did the terrorists of the past. It's a better strategic location.
The problem is ultimately not with a little knife or any other necessity people bring with them on the plane, it is the process overall. There simply isn't, nor was there ever, a good way to secure an airplane while making it an attractive means of travel. About the only way would be to put passengers in individual cages and knock them out.
Maybe a better way to approach airport security is to handle it at check-in rather than at a screening station. Make it easy for people to check and retrieve their bags, and quickly get them into the sanitized environment. But so much for the sanitized environment, since First Class has metal knives again.
Aaaah, the joys of being a 100k flyer. We actually get the nice folks... mostly. I did make the mistake of going in a non-premier line a few weeks back, and they actually had the gall to take my toiletries and leatherman. Don't know how many times I had gotten through with the same bag before...
Not NYC, but there was a major power failure in the Bay Area when construction workers accidently earthed the entire grid. They threw a grounding switch before disconnecting power lines from the grid for regular maintenance work. The entire region went down. We figured it wasn't just our office when workers from the other office blocks started pouring out of their offces onto the streets.
Are you referring to the outage in ~1997 or a more recent incident? The irony of your comment in this context is that it was specifically the 66kV grid tie between substations that made a simple mistake (elsewhere in the system) a catastrophic outage. If my memory serves me, one of the transmission lines coming up the Peninsula was down, and the worker grounded a different line.
The problem with these ties is that they become extremely complicated to operate if they are anything more than two point links, and it takes a gridded structure to provide the reliability improvement.
One approach for creating an incentive for peer review is the (theoretical) value of the information presented in the patent. If someone comes up with a great idea, you have a chance to develop a business around it and pay them license fees.
The military security clearance process requires non-constitutional forfeitures of rights. They can't make you do that until after you are an employee.
Doing work for any financial institution will require a background check. Just to access some data centers I have been fingerprinted. The fingerprinting generally seems to be more regional, being more prevalent in the northeast US and less on the west coast.
A quick perusal of Apple's site suggests otherwise. Refer to This link for the following accolades of Free Software:
The power and simplicity of Mac OS X Server begin with a UNIX-based foundation built around the Mach microkernel and the latest advances from the open source BSD community.
nstead of developing proprietary technologies, Apple has embraced the best open source projects, such as Apache, Samba, OpenLDAP, Kerberos, Postfix, Jabber and SpamAssassin.
Granted, Apple wants to sell their software. They do however rely on Open Source software to help create a total solution, which is about the only way for them to compete with MS. The only "cut" on Linux I saw was a reference to the complexities of Linux... which in my book is a fair enough comment relative to their target audience.
After the bubble burst, I traveled for two years, given similar situations.
To this day, I still get teased for going into the interview for the position I eventually accepted with dreadlocks, suit, and my general discomfort with wearing shoes. (Funny what flip flops on the beach for a couple years does to your tolerance of shoes.) While all the interviews I had originally set up might have had the impression that "What the hell was I thinking?", I got offers for seven of the eight positions I interviewed for.
I also do a pretty good job of protecting my online identity.
It is a little scary how partial information could obscure the real picture, but when you go in for an interview, most of that is a non-issue. It's important to know what might show up going in, though.
Maybe I should just count myself lucky for being named John Smith.
The best approach is really to have a separate EPO for each of the two redundant UPS systems. That way, when the electrician is modifying the EPO, it is almost impossible to have them shut down both UPSs.
I have actually heard of a UK data center with two EPO buttons where both are required to trigger the system, but it when down when polarity was reversed to the system.
The most classic stories are about people with boxes over the button that have built in squealer alarms. Woman hits box with purse or something, and to shut off the noise she lifts the cover and pushes the button. That made the room quiet...
I don't think that's the real reason. Retailers are already on the hook for any "slippage" that occurs in their inventory. Why should the studio get involved since they're compensated regardless?
I think you are missing the opportunity there. The novel approach for the system would be that the studio activates the disk on sale only-- the retailer carries the bulk product, and can trash it if it does not sell. Maybe the retailer is charged closer to cost of production for the disks themselves, and the lion's share on activation, when they actually get the money from the transaction. Lowers the store owner's overhead, and gives the studios an opportunity to place margin pressure on them.
The other thought is that this could offer a competitive delivery method to "burn on demand", and keep the market alive for brick and mortar stores.
As for the theft prevention-- the store owner has a vested interest in making sure his "activation" device is only used for real sales, since he will pay out cash for any other transactions. (Opportunities for tying in the cash register to the activation mechanism could also help secure the system.) Makes it harder to fence un-activated DVDs.
It isn't about the King wanting to send people to jail, it is the Coup leaders of the interim military government wanting to restore pride in the Monarchy. The monarchy is what gives them the authority to depose the former government... a government that was trying to usurp power and prestige from the traditional powers and beliefs.
If you have a real complaint that is mishandled by the support tree, writing the CEO is the easiest way to get things resolved. It's worked for me with United Airlines and several other computer vendors.
I just wish it worked with elected officials. Senator Feinstein would actually read the letters rather than just having a flunky pick a form letter that matches the general topic, but does nothing to address my stated concerns. Barbara Boxer does a much better job of actually getting a response out that makes it sound like she at least understands the issues.
(And of course, in all cases it is an assistant that is actually reading and responding to the letters, but I am sure that some type of summary of that type of action is provided to them periodically.)
I still miss the "salvage" from Netware-- the ability to restore any revision to a file as disk space permits. Just hacking rm doesn't fix someone overwriting a file.
As for security, you could disable salvage for sensitive volumes or directories, or have firm policy based wipes of deleted files on a scheduled basis. Often times, Salvage was most useful when a problem was discovered 20 minutes after it occurred.
It sounds like ZFS will do a better job of allowing snapshots to support something like this, but I'm disappointed it isn't in the kernel now.
3) Apple risks pissing off the movie studios that offer video on iTunes stores. (AFAIK, only Disney so far.) People expect to be able to rip CDs, so that's OK. But if people aren't expecting to rip DVDs, why let them? It would cannibalise sales from iTunes Video Store.
Actually, the key for Apple is to create a market where people have an incentive to put movies into iTunes in the first place. Once people put in movies they already own, they have a library and future purchases are more likely to be made online. Same held true for music.
The real challenge will be the DVD rental business. It becomes much harder to tell if people own a DVD or are just renting it. I'm not sure the studios would be ok with people ripping rented movies. This might be the pill they have to swallow though to allow penetration of downloads.
Personally, I can't see there being a big market for purchasing movies online for download at current DVD price points. Most people only watch most of their movies once or twice-- the big exceptions are kid's movies and cult classics. The studios would need to swallow the idea that people pay a slight premium over "renting" to "own" the content... but it has to be a very small premium.
I really hope Apple can take advantage of the situation. The appleTV is kind of an odd ball without that type of functionality.
This becomes even more irrelevant if you can make fair-use duplicates of the DVD itself.
If your method of producing fair-use duplicates is through software on those 3.5" floppies (which are still DRM protected), then you are SOL. This is the problem with "perfect DRM." It can't anticipate the future in codec, needs, or physical media.
And to say that they can't achieve 99.9% availability for power is entirely false. Rackspace, who hosts my companies website, has far better than 99.9% uptime on their infrastructure. That's what redundancy and backup generators are for.
Kaiser isn't Rackspace. Kaiser isn't a bank, either. Kaiser's primary data center related to Health Connect is/was in a sad state-- systems loaded to 90% or more of capacity, eliminating whatever redundancy once existed. Major problems with utility electrical supply, generator systems, and UPS alike. They decided to expand their ~20-year old UPS system rather than replace it! This alone caused major issues to their reliability.
Redundant servers are great, but when they are in the same building, supplied from the same PDU, and often in the same rack, there isn't a whole lot of real redundancy provided. That is when the core infrastructure really determines the net system reliability.
As for excluding scheduled maintenance from availability calculations, that is hogwash. If a business unit does not have the services required, the system is down and that counts against availability. It might be a low severity incident, but it is an incident none the less.
What's wrong with N+1 redundancy? While I agree it requires an attentive response from Facilities to replace the failed unit quickly, my experience with facilities departments is that, if there is greater redundancy, they tend to rely on it to stretch out response and replacement intervals.
N+1 Redundancy is appropriate for systems where individual pieces of equipment require significant maintenance, like a pump or chiller. The redundancy is provided for the purposes of concurrent maintainability. N+1 Redundancy does not do much for improving overall reliability.
To improve reliability, you must introduce fault tolerance. Since an N+1 system always has single points of failure, it cannot tolerate a complete range of faults to a system.
In your example, redundant CRAC units are useful in allowing maintenance of a single unit, or supporting a failure of a single unit. Maintenance wasn't "stupid"; they reasonably counted on the system redundancy. Unfortunately, the scenario you describe should require escalation by maintenance to identify hightened risk for business units. Some contingency planning, like move-n-cool's could have been provisioned if deemed necessary.
The point I originally was trying to make was that when you have just IT people looking at a problem, they often fail to realize operational issues. IBM has often made this mistake with their consulting clients. They are just interested in selling more hardware and consulting services.
In fairness, they had some huge growing pains with the project that simply were not properly planned out. They outgrew all of their data centers at once, bought a new one that they thought they could just move right into... and had to spend $100M to upgrade it. That forced them to stay in the facilities that were bursting at the seams longer, and work through some extremely high risk activities.
In theory, 99.7% uptime is a system outage of 6 hours once a quarter for maintenance. It might take them a couple years to get there, but they have some great people on board now to at least reduce the facility contribution significantly. Unfortunately, it isn't likely that they can keep 99.9% availability for input power to the IT systems.
I think in 2006 they had 2-3 major incidents in each of their two data centers. Power might have only been out for around 15 minutes each, but system recovery took well over 8 hours for many of the problems. There were also several planned shutdowns for construction activities, which accounted for another 4-5 6 hour scheduled system outages....And all the while, IBM is telling them they just need an N+1 UPS system...
Looks like things are getting better now; they have brought in several IT and facilities people from the banking world where there is a better sense of uptime. Unfortunately, the fundamental decision early on in system architecture will make it hard for them to get even 3 9's from Health Connect. They did open a new data center in Napa, but I understand that is an N+1 UPS as well.
Slashdot Mirror
Compressed air is the key though... not the lack of tap water or soap. Many places use nitrogen instead of compressed air, but either way you have a very clean, dry airstream to clean it.
Or for that matter the Optimus Maximus.
(Seriously... are they making it yet?)
The problem with that logic is that the statistical average of all users is pushed up by "Peter." He might not fit into the old distribution, but he is a part of the new one. As Quincy, Robert, Sam and Tom all begin to have similar usage patterns, the average usage begins to fit more closely Peter's usage.
The ISP needs to adjust their models to reflect these changes over time.
Personally, I would prefer for an ISP to tier levels of service and commit to a contention ratio they can afford. If a user exceeds the preset contention ratio for their package over a 7 or 30 day period, they are bumped into the next tier after a warning. Start out with a 512k, 1% contention which should be adequate for most users (ends up at 1.5G/month), then go to a 1.024M, 2% (6G/month), 2M, 5% (30GB/month), 6M, 10%...
Tie the sense of value (bandwidth) into the true cost (transfer), and give the ISP the incentive to improve over time as well as give the customer an incentive to buy into a higher package. If internet TV takes off (for example), over time a market is created for improvements...
Network segmentation, anyone?
As did the terrorists of the past. It's a better strategic location.
The problem is ultimately not with a little knife or any other necessity people bring with them on the plane, it is the process overall. There simply isn't, nor was there ever, a good way to secure an airplane while making it an attractive means of travel. About the only way would be to put passengers in individual cages and knock them out.
Maybe a better way to approach airport security is to handle it at check-in rather than at a screening station. Make it easy for people to check and retrieve their bags, and quickly get them into the sanitized environment. But so much for the sanitized environment, since First Class has metal knives again.
Aaaah, the joys of being a 100k flyer. We actually get the nice folks... mostly. I did make the mistake of going in a non-premier line a few weeks back, and they actually had the gall to take my toiletries and leatherman. Don't know how many times I had gotten through with the same bag before...
Are you referring to the outage in ~1997 or a more recent incident? The irony of your comment in this context is that it was specifically the 66kV grid tie between substations that made a simple mistake (elsewhere in the system) a catastrophic outage. If my memory serves me, one of the transmission lines coming up the Peninsula was down, and the worker grounded a different line.
The problem with these ties is that they become extremely complicated to operate if they are anything more than two point links, and it takes a gridded structure to provide the reliability improvement.
One approach for creating an incentive for peer review is the (theoretical) value of the information presented in the patent. If someone comes up with a great idea, you have a chance to develop a business around it and pay them license fees.
The military security clearance process requires non-constitutional forfeitures of rights. They can't make you do that until after you are an employee.
Doing work for any financial institution will require a background check. Just to access some data centers I have been fingerprinted. The fingerprinting generally seems to be more regional, being more prevalent in the northeast US and less on the west coast.
Metricom was the first stock I ever owned. I still feel a sense of geek pride when I see them.
I wish this technology had kept moving... lots of opportunity for mesh networks to help with local and mobile network access.
After the bubble burst, I traveled for two years, given similar situations.
To this day, I still get teased for going into the interview for the position I eventually accepted with dreadlocks, suit, and my general discomfort with wearing shoes. (Funny what flip flops on the beach for a couple years does to your tolerance of shoes.) While all the interviews I had originally set up might have had the impression that "What the hell was I thinking?", I got offers for seven of the eight positions I interviewed for.
I also do a pretty good job of protecting my online identity.
It is a little scary how partial information could obscure the real picture, but when you go in for an interview, most of that is a non-issue. It's important to know what might show up going in, though.
Maybe I should just count myself lucky for being named John Smith.
The best approach is really to have a separate EPO for each of the two redundant UPS systems. That way, when the electrician is modifying the EPO, it is almost impossible to have them shut down both UPSs.
I have actually heard of a UK data center with two EPO buttons where both are required to trigger the system, but it when down when polarity was reversed to the system.
The most classic stories are about people with boxes over the button that have built in squealer alarms. Woman hits box with purse or something, and to shut off the noise she lifts the cover and pushes the button. That made the room quiet...
I think you are missing the opportunity there. The novel approach for the system would be that the studio activates the disk on sale only-- the retailer carries the bulk product, and can trash it if it does not sell. Maybe the retailer is charged closer to cost of production for the disks themselves, and the lion's share on activation, when they actually get the money from the transaction. Lowers the store owner's overhead, and gives the studios an opportunity to place margin pressure on them.
The other thought is that this could offer a competitive delivery method to "burn on demand", and keep the market alive for brick and mortar stores.
As for the theft prevention-- the store owner has a vested interest in making sure his "activation" device is only used for real sales, since he will pay out cash for any other transactions. (Opportunities for tying in the cash register to the activation mechanism could also help secure the system.) Makes it harder to fence un-activated DVDs.
Not to be a smart ass, but have you heard of 2-story homes?
With odd roof shapes, even a ranch home can have a limited amount of south-facing roof.
It isn't about the King wanting to send people to jail, it is the Coup leaders of the interim military government wanting to restore pride in the Monarchy. The monarchy is what gives them the authority to depose the former government... a government that was trying to usurp power and prestige from the traditional powers and beliefs.
The missing part is that different networks must operate together as one. Not required for power water, etc.
If you have a real complaint that is mishandled by the support tree, writing the CEO is the easiest way to get things resolved. It's worked for me with United Airlines and several other computer vendors.
I just wish it worked with elected officials. Senator Feinstein would actually read the letters rather than just having a flunky pick a form letter that matches the general topic, but does nothing to address my stated concerns. Barbara Boxer does a much better job of actually getting a response out that makes it sound like she at least understands the issues.
(And of course, in all cases it is an assistant that is actually reading and responding to the letters, but I am sure that some type of summary of that type of action is provided to them periodically.)
I still miss the "salvage" from Netware-- the ability to restore any revision to a file as disk space permits. Just hacking rm doesn't fix someone overwriting a file.
As for security, you could disable salvage for sensitive volumes or directories, or have firm policy based wipes of deleted files on a scheduled basis. Often times, Salvage was most useful when a problem was discovered 20 minutes after it occurred.
It sounds like ZFS will do a better job of allowing snapshots to support something like this, but I'm disappointed it isn't in the kernel now.
Actually, the key for Apple is to create a market where people have an incentive to put movies into iTunes in the first place. Once people put in movies they already own, they have a library and future purchases are more likely to be made online. Same held true for music.
The real challenge will be the DVD rental business. It becomes much harder to tell if people own a DVD or are just renting it. I'm not sure the studios would be ok with people ripping rented movies. This might be the pill they have to swallow though to allow penetration of downloads.
Personally, I can't see there being a big market for purchasing movies online for download at current DVD price points. Most people only watch most of their movies once or twice-- the big exceptions are kid's movies and cult classics. The studios would need to swallow the idea that people pay a slight premium over "renting" to "own" the content... but it has to be a very small premium.
I really hope Apple can take advantage of the situation. The appleTV is kind of an odd ball without that type of functionality.
If your method of producing fair-use duplicates is through software on those 3.5" floppies (which are still DRM protected), then you are SOL. This is the problem with "perfect DRM." It can't anticipate the future in codec, needs, or physical media.
Kaiser isn't Rackspace. Kaiser isn't a bank, either. Kaiser's primary data center related to Health Connect is/was in a sad state-- systems loaded to 90% or more of capacity, eliminating whatever redundancy once existed. Major problems with utility electrical supply, generator systems, and UPS alike. They decided to expand their ~20-year old UPS system rather than replace it! This alone caused major issues to their reliability.
Redundant servers are great, but when they are in the same building, supplied from the same PDU, and often in the same rack, there isn't a whole lot of real redundancy provided. That is when the core infrastructure really determines the net system reliability.
As for excluding scheduled maintenance from availability calculations, that is hogwash. If a business unit does not have the services required, the system is down and that counts against availability. It might be a low severity incident, but it is an incident none the less.
N+1 Redundancy is appropriate for systems where individual pieces of equipment require significant maintenance, like a pump or chiller. The redundancy is provided for the purposes of concurrent maintainability. N+1 Redundancy does not do much for improving overall reliability.
To improve reliability, you must introduce fault tolerance. Since an N+1 system always has single points of failure, it cannot tolerate a complete range of faults to a system.
In your example, redundant CRAC units are useful in allowing maintenance of a single unit, or supporting a failure of a single unit. Maintenance wasn't "stupid"; they reasonably counted on the system redundancy. Unfortunately, the scenario you describe should require escalation by maintenance to identify hightened risk for business units. Some contingency planning, like move-n-cool's could have been provisioned if deemed necessary.
The point I originally was trying to make was that when you have just IT people looking at a problem, they often fail to realize operational issues. IBM has often made this mistake with their consulting clients. They are just interested in selling more hardware and consulting services.
In fairness, they had some huge growing pains with the project that simply were not properly planned out. They outgrew all of their data centers at once, bought a new one that they thought they could just move right into... and had to spend $100M to upgrade it. That forced them to stay in the facilities that were bursting at the seams longer, and work through some extremely high risk activities.
In theory, 99.7% uptime is a system outage of 6 hours once a quarter for maintenance. It might take them a couple years to get there, but they have some great people on board now to at least reduce the facility contribution significantly. Unfortunately, it isn't likely that they can keep 99.9% availability for input power to the IT systems.
I think in 2006 they had 2-3 major incidents in each of their two data centers. Power might have only been out for around 15 minutes each, but system recovery took well over 8 hours for many of the problems. There were also several planned shutdowns for construction activities, which accounted for another 4-5 6 hour scheduled system outages. ...And all the while, IBM is telling them they just need an N+1 UPS system...
Looks like things are getting better now; they have brought in several IT and facilities people from the banking world where there is a better sense of uptime. Unfortunately, the fundamental decision early on in system architecture will make it hard for them to get even 3 9's from Health Connect. They did open a new data center in Napa, but I understand that is an N+1 UPS as well.