With high-density SSDs at roughly $2.50/GB, and high-density (8+ GB) server-class DIMMs at roughly $50/GB, there is still a huge window where SSDs are the right choice versus in-memory databases. For applications where the IOPS requriements exceed mechanical disk, and the data size isn't trivially small, SSDs are the logical choice if the factor of 100+ improvement in IOPS is enough. That includes the majority of IOPS-bound databases, I think.
When you add in the fact that SSDs are "drop-in" replacements for an existing architecture piece, and do not require any code or operational changes at all, the appeal is great. In-memory databases are, by and large, still niche products. You can run MySQL, PostgreSQL, Oracle, DB2, and MSSQL from SSDs simply by swapping out the mechanical drives.
I was talking about actually finding the optimal price-point, graphing price/performance, and finding the farthest point on the performance axis before the slope of the line significantly increases indicating that you're getting very little return on your extra money. That math is still very relevant. And sure, it's going to point you at a quad-core processor today. But not the fastest one.
It's funny. I do the same thing often the server space, but include total system costs, including chassis, memory, lifetime power & cooling costs, lifetime warranty and maintenance costs. These costs actually dominate the raw chip costs to such a degree that it almost always makes sense to buy the fastest processor you can and fully load the chassis with as much memory, networking, storage, etc. as you can. Assuming, of course, your workloads can actually use all those resources. Thanks to virtualization, this is pretty much always true these days.
I imagine this is why AMD has been so dominated by Intel in the server space in the last few years: a cheaper chip might make the TCO of a system 5% lower over 4 years. But if it is 30% slower, it's a bad deal.
Cameras don't prevent crime. They may help solve a crime after the fact, but I imagine a camera means little to a crackhead wearing a ski mask.
London is one example of this. Total "violence against the person" rates are largely unchanged in the last 5 years. Similar statistics can be seen in Chicago, USA, despite thousands of cameras added in "high crime" areas over the last 5 years. The cameras have been a bonanza for local contractors who install and maintain them, but that's about the only benefit.
To paraphrase Bruce Schneier, replacing smart people with dumb technology rarely results in increased security.
What I'd want is a way to have more control over the program. Maybe put it in a sandbox and trick it into thinking it's got full privileges even though it's really sandboxed so it won't crash or maybe just set advanced settings for that specific application to disallow it from writing to specific registry/files/network/other process' memory.
Which is... umm... pretty much exactly what Windows Vista, Windows 7, and Windows Server 2008 can do.
Your ISP pays a backbone provider for a connection.
Maybe. But not if your ISP is a tier-1 provider (like Sprint, Verizon, AT&T, Qwest). Or if your ISP has a crapload of private settlement-free peering arrangements that cover the traffic in question. This is far more common than people think. Looking at traceroutes, my ISP, for instance, seems to peer directly with Google, MSFT, Akamai, and Limelight. That's a huge chunk of traffic not crossing any "backbone". Tier-2 and smaller ISPs are happy to peer directly with these big boys for nothing, since it lowers their infrastructure costs and keeps traffic off connections for which they do pay.
Backbone providers pay each other to carry each other's data.
Not really, at least most of the time. What we think of as the Internet "backbone" - the collection of Tier-1 providers - do "settlement-free" peering with each other. They do this because it is cheaper to simply connect up and let the data flow than it is to meter and bill each other for any imbalances. It's an exclusive club that resists new players, but some upstarts have "joined" or come very close in recent years (XO and Cogent).
You better stay the hell out of Europe. Police there have vastly greater powers of search, seizure, and detention than they do in the USA. I particularly enjoyed being "monitored" by two jackbooted thugs with sub-machine guns while being "interviewed" by customs in Frankfort. And that was in the late 90s, pre-9/11.
Remember, Democrats are always wrong on every topic because they murder babies, and you don't want to trust a baby murderer, do you? The sad part is that I've heard more or less that specific argument in the recent past.
Actually, that's not such a terrible argument. If someone has such a fundamental disconnect with your own personal values that they promote something which you consider murder by natural law and common sense, it's very hard to trust them. You're simply too different to deal with one another.
I believe this is actually the same reason WWII ended for Japan in a mushroom cloud... we in the USA simply had no basis for understanding how the Japanese people at the time thought - their culture was too different. It was alien to us that a nation - including civilians - would effectively commit suicide to preserve traditions and a notion of honor that we just didn't understand. I fear the same divide exists between the western world and fundamentalist Islam - we simply can't grok them, nor they us.
Do you trust apt/yum/portage/whatever on your Linux/BSD distro of choice? Same thing... you trust that the developer's code-signing and key management policies are solid, and they won't dick you by releasing something really bad.
If you're not turning on automatic updates on Windows boxes (and even MacOS and Linux boxes), you might be part of the problem. Yes, you should have centralized patch review and deployment in place for all the machines you manage... but make sure it is all of them. All my company's servers and workstations have managed deployments, but I've configured Mom's laptop to get all updates ASAP straight from the vendor so I don't have to fuck with it and she remains malware-free. She hasn't had major breakage from an automatic update from any vendor in more than a decade (unless you count Adobe, whose best code is still horribly broken). There are probably 20-30 machines I "manage" informally at any given time, and I don't want to tackle patching them interactively, or deal with setting up my own WSUS or apt repository for them.
Whatever doesn't involve turning away half your potential customers.
Linux desktop users are generally not "potential customers" of Microsoft, and are an extremely tiny market compared with Windows and Apple desktop users. This thing is actually available to about 97% of web surfers.
running on a system that does not support multiple user accounts (well)
1996 called. They want their anti-Microsoft rant back. This hasn't been true since NT 3.5.1 was released. The NT series of the Windows operating system has always supported multiple users very well (I would say better than *nix-like systems because of the more robust ACL model). End-user applications, on the other hand, have in the past not supported multiple users well (e.g. sticking configuration in %WINDIR% or HKEY_LOCAL_MACHINE instead of per-user locations).
Meanwhile, if resolutions above 1080p become popular (as could be possible for people who have a large iMac (2560 x 1440) for example could use, YouTube's network will fall over.
No, it won't. Google has kilometers of 10 Gb/s fiber spanning the globe, and far better connectivity inside each of their datacenters. It's your ISP's network (via private peering), or that of any public peering points that will "fall over".
Modern workstation and server worloads are dominated by random access IO, where SSDs excel. Even a 10-drive 15K RAID array can only do about 1800 random IOPS, while a single Intel X-25M can do 30,000+ random reads and 3000+ random writes for $400. There really is no difference if you've used a good SSD on a day-to-day basis - I will never go back.
Transfer rate, which sites like TR and Anand always highlight, is a mostly useless metric for a disk subsystem. Transfer rate is essentially meaninful for one common operation: backup. Which is a stupid thing to optimize, since it happens in the middle of the night.
There used to be a gay bar in Chicago's "boys Town" neighborhood called "The Manhole". Never went in myself, mind you. Not that there's anything wrong with that.
Good luck getting a GPS receiver on the roof of an existing facility. Running conduit, grounding, waterproofing, etc. People say "hey, the Garmin GPS 18x costs only $100", but it will require ten months and $5K to install.
Greenpeace et. all. only begrudgingly accepted DDT as a anti-malaria measure since the mid-2000s, despite 30+ years of a massive resurgence in malaria deaths.
For 30 years they did everything they could to ban the use, manufacture, and even donation of DDT to third-world countries, despite staggering increases in malarial deaths. There was no practical substitute for DDT, yet they spent their money on a campaign for its eradication instead of research into safer alternatives. Evil.
Even if what you assert were factual, so what? Millions of (mostly poor and brown) human children died from malaria, which was a non-problem until DDT was banned. Fuck the birds, and fuck all the racist sociopatic Eco-terrorists who trade human lives for their camping trips.
I'd prefer the RIAA have their money and power, personally. But that's not my original point. They are all allowed to operate in a free society, like it or not.
Greenpeace is directly responsible for the deaths of millions of children from malaria, as they led the campaign for the ban of DDT without any scientific basis. They're far more evil than any for-profit entity.
Yes they are, in the way Greenpeace, the American Red Cross, or the FSF are "people". Corporations are simply groups of people acting together. Why should one group of people be allowed to pool their resources and influence, but not others?
And you could still use LVM to make a sane, fast, performant snapshot of it either way.
No, you can't.. Do make sane snapshots, you need application-layer support, especially for databases. Otherwise your snapshots are not clean, they are in a crash-recovery state as far as the application is concerned. This is particularly important for databases, distributed file-systems, and queue-based software (including mail stores).
The beauty of the Microsoft Volume ShadowCopy Service is that it is a stable API through wich applications, backup software, and storage systems can all communicate. Backup software says "I want a snapshot". All the applications who have registrerd themselves with VSS get their writes flushed into a consistent state, and say "I am ready for snapshot". Then the storage device (either a software filesystem like NTFS or a SAN/NAS) says "I am doing the snapshot". This level of coordination is simply not possible on Linux.
While I like the idea on a visceral level, the "only veterans can judge" thing could never work in practice. There would be far too much room for abuse and collusion, just like the "blue line of silence" shown by police officials towards internal corruption. This is the real world, and not Starship Troopers. A jury of randomly selected ordinary citizens is shown the evidence, and determine if a supposed crime was an accident, negligence, or willful action. That's the system, and it needs to be applied here.
"Anycast TCP" means advertising a connection-oriented TCP service via anycast routing. I did not make up the term, Google for "anycast TCP".
And as far as Gomez is concerned, I am not a customer nor an advocate, but they are probably the best-known distributed monitoring service out there. My point was that you need lots of long-term measurements from many vantage points to see if using anycast with TCP services is really as trouble-free as CacheFly et. all contend.
Slashdot Mirror
With high-density SSDs at roughly $2.50/GB, and high-density (8+ GB) server-class DIMMs at roughly $50/GB, there is still a huge window where SSDs are the right choice versus in-memory databases. For applications where the IOPS requriements exceed mechanical disk, and the data size isn't trivially small, SSDs are the logical choice if the factor of 100+ improvement in IOPS is enough. That includes the majority of IOPS-bound databases, I think.
When you add in the fact that SSDs are "drop-in" replacements for an existing architecture piece, and do not require any code or operational changes at all, the appeal is great. In-memory databases are, by and large, still niche products. You can run MySQL, PostgreSQL, Oracle, DB2, and MSSQL from SSDs simply by swapping out the mechanical drives.
I was talking about actually finding the optimal price-point, graphing price/performance, and finding the farthest point on the performance axis before the slope of the line significantly increases indicating that you're getting very little return on your extra money. That math is still very relevant. And sure, it's going to point you at a quad-core processor today. But not the fastest one.
It's funny. I do the same thing often the server space, but include total system costs, including chassis, memory, lifetime power & cooling costs, lifetime warranty and maintenance costs. These costs actually dominate the raw chip costs to such a degree that it almost always makes sense to buy the fastest processor you can and fully load the chassis with as much memory, networking, storage, etc. as you can. Assuming, of course, your workloads can actually use all those resources. Thanks to virtualization, this is pretty much always true these days.
I imagine this is why AMD has been so dominated by Intel in the server space in the last few years: a cheaper chip might make the TCO of a system 5% lower over 4 years. But if it is 30% slower, it's a bad deal.
Cameras don't prevent crime. They may help solve a crime after the fact, but I imagine a camera means little to a crackhead wearing a ski mask.
London is one example of this. Total "violence against the person" rates are largely unchanged in the last 5 years. Similar statistics can be seen in Chicago, USA, despite thousands of cameras added in "high crime" areas over the last 5 years. The cameras have been a bonanza for local contractors who install and maintain them, but that's about the only benefit.
To paraphrase Bruce Schneier, replacing smart people with dumb technology rarely results in increased security.
What I'd want is a way to have more control over the program. Maybe put it in a sandbox and trick it into thinking it's got full privileges even though it's really sandboxed so it won't crash or maybe just set advanced settings for that specific application to disallow it from writing to specific registry/files/network/other process' memory.
Which is... umm... pretty much exactly what Windows Vista, Windows 7, and Windows Server 2008 can do.
the shear popularity of the iPhone
Wow, there's an app for haircuts too?
Who is doing the taking in this scenario?
Chuck Norris, of course.
You, the customer pays the ISP for a connection.
Yeah, OK.
Your ISP pays a backbone provider for a connection.
Maybe. But not if your ISP is a tier-1 provider (like Sprint, Verizon, AT&T, Qwest). Or if your ISP has a crapload of private settlement-free peering arrangements that cover the traffic in question. This is far more common than people think. Looking at traceroutes, my ISP, for instance, seems to peer directly with Google, MSFT, Akamai, and Limelight. That's a huge chunk of traffic not crossing any "backbone". Tier-2 and smaller ISPs are happy to peer directly with these big boys for nothing, since it lowers their infrastructure costs and keeps traffic off connections for which they do pay.
Backbone providers pay each other to carry each other's data.
Not really, at least most of the time. What we think of as the Internet "backbone" - the collection of Tier-1 providers - do "settlement-free" peering with each other. They do this because it is cheaper to simply connect up and let the data flow than it is to meter and bill each other for any imbalances. It's an exclusive club that resists new players, but some upstarts have "joined" or come very close in recent years (XO and Cogent).
You better stay the hell out of Europe. Police there have vastly greater powers of search, seizure, and detention than they do in the USA. I particularly enjoyed being "monitored" by two jackbooted thugs with sub-machine guns while being "interviewed" by customs in Frankfort. And that was in the late 90s, pre-9/11.
Remember, Democrats are always wrong on every topic because they murder babies, and you don't want to trust a baby murderer, do you? The sad part is that I've heard more or less that specific argument in the recent past.
Actually, that's not such a terrible argument. If someone has such a fundamental disconnect with your own personal values that they promote something which you consider murder by natural law and common sense, it's very hard to trust them. You're simply too different to deal with one another.
I believe this is actually the same reason WWII ended for Japan in a mushroom cloud... we in the USA simply had no basis for understanding how the Japanese people at the time thought - their culture was too different. It was alien to us that a nation - including civilians - would effectively commit suicide to preserve traditions and a notion of honor that we just didn't understand. I fear the same divide exists between the western world and fundamentalist Islam - we simply can't grok them, nor they us.
Do you trust apt/yum/portage/whatever on your Linux/BSD distro of choice? Same thing... you trust that the developer's code-signing and key management policies are solid, and they won't dick you by releasing something really bad.
If you're not turning on automatic updates on Windows boxes (and even MacOS and Linux boxes), you might be part of the problem. Yes, you should have centralized patch review and deployment in place for all the machines you manage... but make sure it is all of them. All my company's servers and workstations have managed deployments, but I've configured Mom's laptop to get all updates ASAP straight from the vendor so I don't have to fuck with it and she remains malware-free. She hasn't had major breakage from an automatic update from any vendor in more than a decade (unless you count Adobe, whose best code is still horribly broken). There are probably 20-30 machines I "manage" informally at any given time, and I don't want to tackle patching them interactively, or deal with setting up my own WSUS or apt repository for them.
Whatever doesn't involve turning away half your potential customers.
Linux desktop users are generally not "potential customers" of Microsoft, and are an extremely tiny market compared with Windows and Apple desktop users. This thing is actually available to about 97% of web surfers.
running on a system that does not support multiple user accounts (well)
1996 called. They want their anti-Microsoft rant back. This hasn't been true since NT 3.5.1 was released. The NT series of the Windows operating system has always supported multiple users very well (I would say better than *nix-like systems because of the more robust ACL model). End-user applications, on the other hand, have in the past not supported multiple users well (e.g. sticking configuration in %WINDIR% or HKEY_LOCAL_MACHINE instead of per-user locations) .
Meanwhile, if resolutions above 1080p become popular (as could be possible for people who have a large iMac (2560 x 1440) for example could use, YouTube's network will fall over.
No, it won't. Google has kilometers of 10 Gb/s fiber spanning the globe, and far better connectivity inside each of their datacenters. It's your ISP's network (via private peering), or that of any public peering points that will "fall over".
Modern workstation and server worloads are dominated by random access IO, where SSDs excel. Even a 10-drive 15K RAID array can only do about 1800 random IOPS, while a single Intel X-25M can do 30,000+ random reads and 3000+ random writes for $400. There really is no difference if you've used a good SSD on a day-to-day basis - I will never go back.
Transfer rate, which sites like TR and Anand always highlight, is a mostly useless metric for a disk subsystem. Transfer rate is essentially meaninful for one common operation: backup. Which is a stupid thing to optimize, since it happens in the middle of the night.
There used to be a gay bar in Chicago's "boys Town" neighborhood called "The Manhole". Never went in myself, mind you. Not that there's anything wrong with that.
Good luck getting a GPS receiver on the roof of an existing facility. Running conduit, grounding, waterproofing, etc. People say "hey, the Garmin GPS 18x costs only $100", but it will require ten months and $5K to install.
Greenpeace et. all. only begrudgingly accepted DDT as a anti-malaria measure since the mid-2000s, despite 30+ years of a massive resurgence in malaria deaths.
For 30 years they did everything they could to ban the use, manufacture, and even donation of DDT to third-world countries, despite staggering increases in malarial deaths. There was no practical substitute for DDT, yet they spent their money on a campaign for its eradication instead of research into safer alternatives. Evil.
Even if what you assert were factual, so what? Millions of (mostly poor and brown) human children died from malaria, which was a non-problem until DDT was banned. Fuck the birds, and fuck all the racist sociopatic Eco-terrorists who trade human lives for their camping trips.
Yes, Greenpeace is a sociopath.
I'd prefer the RIAA have their money and power, personally. But that's not my original point. They are all allowed to operate in a free society, like it or not.
Greenpeace is directly responsible for the deaths of millions of children from malaria, as they led the campaign for the ban of DDT without any scientific basis. They're far more evil than any for-profit entity.
"Improving society" is a very subjective thing.
Yes they are, in the way Greenpeace, the American Red Cross, or the FSF are "people". Corporations are simply groups of people acting together. Why should one group of people be allowed to pool their resources and influence, but not others?
They look cooler than little round holes or slits, and probably use a little less metal, meaning more can be recycled after stamping.
And you could still use LVM to make a sane, fast, performant snapshot of it either way.
No, you can't.. Do make sane snapshots, you need application-layer support, especially for databases. Otherwise your snapshots are not clean, they are in a crash-recovery state as far as the application is concerned. This is particularly important for databases, distributed file-systems, and queue-based software (including mail stores).
The beauty of the Microsoft Volume ShadowCopy Service is that it is a stable API through wich applications, backup software, and storage systems can all communicate. Backup software says "I want a snapshot". All the applications who have registrerd themselves with VSS get their writes flushed into a consistent state, and say "I am ready for snapshot". Then the storage device (either a software filesystem like NTFS or a SAN/NAS) says "I am doing the snapshot". This level of coordination is simply not possible on Linux.
While I like the idea on a visceral level, the "only veterans can judge" thing could never work in practice. There would be far too much room for abuse and collusion, just like the "blue line of silence" shown by police officials towards internal corruption. This is the real world, and not Starship Troopers. A jury of randomly selected ordinary citizens is shown the evidence, and determine if a supposed crime was an accident, negligence, or willful action. That's the system, and it needs to be applied here.
"Anycast TCP" means advertising a connection-oriented TCP service via anycast routing. I did not make up the term, Google for "anycast TCP".
And as far as Gomez is concerned, I am not a customer nor an advocate, but they are probably the best-known distributed monitoring service out there. My point was that you need lots of long-term measurements from many vantage points to see if using anycast with TCP services is really as trouble-free as CacheFly et. all contend.