Since you're such a low UID I'll bother answering your question.
Thank you, this is one of the few valid answers to my primary question which is of actual experience with clustered file systems.
I had already thrown out OCFS2 and GFS2 as possible candidates, but that was irrelevant to my reply. Also currently I am unaware of any non-proprietary hardware or software RAID (mdadm in particular) that supports active/active or active/passive on a shared backplane at any RAID level other than 1 or 0 (i.e. DRBD) and rather expensive and not yet released Areca external RAID controllers. Also I'm looking for whitebox OSS solutions.
Hmmm.....I'll have to look more deeply in to ZFS as I keep hearing it thrown out there. I should have probably qualified my statement as "thereby supporting live guest migrations". The non-sequitur was basically a hint thrown in to suggest what I meant by high-availability for those less likely to catch or understand the subtle distinction of what high-availability typically means. Just like most other things in life, key requirements may be more basic than what I've described, but if the car salesman throws in the air freshener in free with the car, I'll take it as long as it doesn't stink.
http://wiki.lustre.org/index.php/Lustre_2.0_Features lists filesystem replication as a benefit of Luster 2.0 back in November 2009. I won't be running any RAID since my requirement isn't really to reduce number of disks used by relying on parity. One or more replication partners/mirrors will handle that function. Rsync won't work for the aforementioned clustered virtualization needs.
Yeah, the complication is why I'm leaning more towards GlusterFS, yet so far Lustre is more proven. Unless I get some useful anecdotal experience here I'll probably model out all three solutions with VMs and do my own comparisons and performance analysis. Maybe I'll even post my experiences and results here afterwards.
Thank you, this is one of the few valid answers to my primary question which is of actual experience with clustered file systems. I don't think most of the responders got the clue that I'm looking for a solution that will hopefully scale over a decade's worth of time. Investing in separate backup hardware doesn't work because it won't scale seamlessly either. I also don't think many of these people realize that snapshots are effective backups and with distributed and clustered filesystems you can ensure your snapshots are stored redundantly without overwrite corruption. Although RAID is not backup, backup hardware is really old hat. And when my paranoia really sinks in, I can transport those snapshots to alternative media to balance the MTFR of platter-based technologies when the hardware matches my budget. And lastly I think they missed the mentioning of virtualization so they're probably not thinking much about HA and qcow2 base imaging. I'll likely be upgrading to a Super Micro 2U Twin with QDR Infiniband which none of the mentioned solutions have support for in comparison to Lustre and GlusterFS.
It's difficult to believe nobody here has discussed what availability http://en.wikipedia.org/wiki/High_availability#Percentage_calculation your boss wants for your systems. Likely different systems need different levels of service. Perhaps you only need 98% uptime for most of your systems and 99.9% for some others. Can your internal team ensure a particular system is back up in 8-9 hours? Perhaps a particularly critical system needs 99.99% uptime or better. In this situation it is unthinkable to not have external support available at a minute's notice. Now you have to look at what kind of SLA Redhat support can give you. Do they have a band of service where you can get on the line with an actual support person in less than an hour? You really need to know your reasons for each system rather than just setting a carte-blanch policy across all your servers, otherwise you're just paying a tax for having a running computer. And if you ask your boss if he wants 98%, 99.9%, or 99.99% availability and (s)he says "yes" with that blank look in their eye that shows they really don't comprehend the technical implication of each guarantee then don't even bother trying to handle this battle; you'll get nowhere.
It's typically used in conjunction with ebXML http://en.wikipedia.org/wiki/EbXML in the healthcare and other market segments that are heavily regulated.
Use GlusterFS http://www.gluster.org/ for redundancy spanned across one or more JBOD machines for a much easier hardware and data upgrade path. Use oVirt for easy set up http://www.gluster.com/community/documentation/index.php/GlusterFS_oVirt_Setup_Guide. Mount GlusterFS directly to your clients or export via iSCSI target, fibrechannel target, FCoE, NBD, or traditional NFS for a more advanced shared storage solution. And you can still run more of a NAS type setup with CIFS, WebDAV, or the like.
Something nobody seems to be pointing out is that maybe Apple and other American companies need to understand the Chinese culture and Chinese laws and hire better Chinese lawyers to prevent or be able to litigate against these kinds of maneuvers in China.
Another idea, how about American companies just make a part of their manufacturing deal with Chinese manufacturing companies for that company X to effectively buy all IP rights for said product to be manufactured in only China. If Chinese manufacturing company X doesn't make an even more cut-rate offer to get the business, company Y will get it instead and make tons of money not only manufacturing for the US, but by also making local knockoffs legitimately.
Is it just me or does the graph do a steep and then gentle curve in the actuals and then take a linear nose dive in his projections? I'm not a math genius but that totally looks like somebody making up their own agendas and skewing the evidence to support it.
I'm going to hold off on buying these because everybody knows Monster Cables are the best. Their sweet gold-plated impedance really accentuates the harmonics of my digital bits, giving my data soft warm tones and the largest acoustical threshold range that guarantees that my ones are as oney as they can be and my zeros actually stop the measurements in my voltmeter because all the electrons are at a complete standstill. I mean seriously Apple, $50? You're practically admitting that these cables are just junk.
This is unfortunate considering what lulzsec is currently doing for the [door and lock] job market. These attacks are getting incompetent [locksmiths] fired and making companies go out and look for competent [locksmiths] to hire in their place. Also, it is forcing them to actually invest money in their [door and key] infrastructure instead of just slapping some [keyholes] together and letting some clowns straight out of a [high school] mill [install] them. People need to realize that this is a net good thing because if a 19 year old with no formal education is ripping [doors] owned by multi-billion dollar international corporations then the [hobos] have already been there. A company would not even know about the [hobo's] intrusion much less publicize it once they found out so what lulzsec is doing is shining the light on how poorly these companies that hold your data are run.
Not having to have a native mp3 decoder codec on your machine. Also being javascript it is theoretically compatible across all platforms that the browser supports, particular those that may not have native mp3 decoder codecs. The HTML5 standard isn't attempting to establish a standard for codecs. Not saying it's worthwhile or anything, just pointing out potential benefits as you requested.
Seriously, all the engineers are turning to other industries like software development because companies farm out work for pennies on the dollar to India.
I don't have any mod points available. Somebody mod this up +Insightful. Recreating original sound is extremely difficult and has a more limited interest in providing equivalent emulation.
Nobody really mentions this, but if ever I download something whether legitimate or not, I rarely if ever see my seed ratio exceed 10:1. So even if I let 10 people fully download a movie from me, the real loss of distribution is only $200 assuming the inflated $20 a box office ticket. Even if we take a potential loss of 100:1, that's still only $2000 and I highly doubt anyone ever really seeds to this amount. Even a $2,000 fine is incentive enough to get someone to stop torrenting copyrighted works and doesn't use ridiculous Dr. Evil math.
I second this. GNS3 rocks, though you'll need to find the Cisco IOS firmware. Typically you can find these with some google searching for a specific IOS firmware name. You also might want to checkout PEMU for PIX emulation. Unfortunately it hogs up the CPU so you'll want to use a CPU limiter on it, but it works really well. Settings up network bridges, particularly in Windows through loopback adapters, can be a bit challenging, but I was able to take an old multi-cpu Dell 2560 with a decent amount of RAM and Windows XP, run GNS3 and PEMU on it, and connect it to a trunk port on our switch (after restricting the VLANs of course) and we were able to use it to interview potential engineers. We had them configure routers, switches, and firewalls to our specification including seting up NATs to our external VLAN to demonstrate they knew how to do simple things like map things to a public IP, etc. We even modeled and tested a redundant ATM bridge infrastructure and got the ATM configuration just right even before the physical equipment came in to our offices for an upcoming project since none of our network engineers had really done it before. To be able to lab that out we would have had to drop a ton of money just to verify we were capable of doing the project correctly and we were able to finalize the production configuration even before the hardware was drop shipped.
It's been a while for me, but they had also gotten Dynamips (included in GNS3) running Cisco ASA code, but it was restricted to 7.X. I'm not really sure if they overcame the issues extracting the ADSM firmware with 8.4 or not. Fortunately for me I went from being a star network admin to being a star programmer and I rarely touch the stuff anymore.
Slashdot Mirror
Somebody reverse Carmack's Reverse and put it on a tshirt!
Since you're such a low UID I'll bother answering your question.
Thank you, this is one of the few valid answers to my primary question which is of actual experience with clustered file systems.
I had already thrown out OCFS2 and GFS2 as possible candidates, but that was irrelevant to my reply. Also currently I am unaware of any non-proprietary hardware or software RAID (mdadm in particular) that supports active/active or active/passive on a shared backplane at any RAID level other than 1 or 0 (i.e. DRBD) and rather expensive and not yet released Areca external RAID controllers. Also I'm looking for whitebox OSS solutions.
Hmmm.....I'll have to look more deeply in to ZFS as I keep hearing it thrown out there. I should have probably qualified my statement as "thereby supporting live guest migrations". The non-sequitur was basically a hint thrown in to suggest what I meant by high-availability for those less likely to catch or understand the subtle distinction of what high-availability typically means. Just like most other things in life, key requirements may be more basic than what I've described, but if the car salesman throws in the air freshener in free with the car, I'll take it as long as it doesn't stink.
http://wiki.lustre.org/index.php/Lustre_2.0_Features lists filesystem replication as a benefit of Luster 2.0 back in November 2009. I won't be running any RAID since my requirement isn't really to reduce number of disks used by relying on parity. One or more replication partners/mirrors will handle that function. Rsync won't work for the aforementioned clustered virtualization needs.
Yeah, the complication is why I'm leaning more towards GlusterFS, yet so far Lustre is more proven. Unless I get some useful anecdotal experience here I'll probably model out all three solutions with VMs and do my own comparisons and performance analysis. Maybe I'll even post my experiences and results here afterwards.
Not a bad suggestion and more helpful than most. Thanks for the input!
Thank you, this is one of the few valid answers to my primary question which is of actual experience with clustered file systems. I don't think most of the responders got the clue that I'm looking for a solution that will hopefully scale over a decade's worth of time. Investing in separate backup hardware doesn't work because it won't scale seamlessly either. I also don't think many of these people realize that snapshots are effective backups and with distributed and clustered filesystems you can ensure your snapshots are stored redundantly without overwrite corruption. Although RAID is not backup, backup hardware is really old hat. And when my paranoia really sinks in, I can transport those snapshots to alternative media to balance the MTFR of platter-based technologies when the hardware matches my budget. And lastly I think they missed the mentioning of virtualization so they're probably not thinking much about HA and qcow2 base imaging. I'll likely be upgrading to a Super Micro 2U Twin with QDR Infiniband which none of the mentioned solutions have support for in comparison to Lustre and GlusterFS.
It's difficult to believe nobody here has discussed what availability http://en.wikipedia.org/wiki/High_availability#Percentage_calculation your boss wants for your systems. Likely different systems need different levels of service. Perhaps you only need 98% uptime for most of your systems and 99.9% for some others. Can your internal team ensure a particular system is back up in 8-9 hours? Perhaps a particularly critical system needs 99.99% uptime or better. In this situation it is unthinkable to not have external support available at a minute's notice. Now you have to look at what kind of SLA Redhat support can give you. Do they have a band of service where you can get on the line with an actual support person in less than an hour? You really need to know your reasons for each system rather than just setting a carte-blanch policy across all your servers, otherwise you're just paying a tax for having a running computer. And if you ask your boss if he wants 98%, 99.9%, or 99.99% availability and (s)he says "yes" with that blank look in their eye that shows they really don't comprehend the technical implication of each guarantee then don't even bother trying to handle this battle; you'll get nowhere.
It's typically used in conjunction with ebXML http://en.wikipedia.org/wiki/EbXML in the healthcare and other market segments that are heavily regulated.
Use GlusterFS http://www.gluster.org/ for redundancy spanned across one or more JBOD machines for a much easier hardware and data upgrade path. Use oVirt for easy set up http://www.gluster.com/community/documentation/index.php/GlusterFS_oVirt_Setup_Guide. Mount GlusterFS directly to your clients or export via iSCSI target, fibrechannel target, FCoE, NBD, or traditional NFS for a more advanced shared storage solution. And you can still run more of a NAS type setup with CIFS, WebDAV, or the like.
Something nobody seems to be pointing out is that maybe Apple and other American companies need to understand the Chinese culture and Chinese laws and hire better Chinese lawyers to prevent or be able to litigate against these kinds of maneuvers in China.
Another idea, how about American companies just make a part of their manufacturing deal with Chinese manufacturing companies for that company X to effectively buy all IP rights for said product to be manufactured in only China. If Chinese manufacturing company X doesn't make an even more cut-rate offer to get the business, company Y will get it instead and make tons of money not only manufacturing for the US, but by also making local knockoffs legitimately.
back when AT&T was a very fat and happy, and heavily regulated, monopoly.
What do you mean was?
Is it just me or does the graph do a steep and then gentle curve in the actuals and then take a linear nose dive in his projections? I'm not a math genius but that totally looks like somebody making up their own agendas and skewing the evidence to support it.
I'm going to hold off on buying these because everybody knows Monster Cables are the best. Their sweet gold-plated impedance really accentuates the harmonics of my digital bits, giving my data soft warm tones and the largest acoustical threshold range that guarantees that my ones are as oney as they can be and my zeros actually stop the measurements in my voltmeter because all the electrons are at a complete standstill. I mean seriously Apple, $50? You're practically admitting that these cables are just junk.
HBGary has finally enacted their revenge with the most heinous rootkit ever conceived by man. BWAHAHAHAHAHAHAHAH!!!!!!!!
This is unfortunate considering what lulzsec is currently doing for the [door and lock] job market. These attacks are getting incompetent [locksmiths] fired and making companies go out and look for competent [locksmiths] to hire in their place. Also, it is forcing them to actually invest money in their [door and key] infrastructure instead of just slapping some [keyholes] together and letting some clowns straight out of a [high school] mill [install] them. People need to realize that this is a net good thing because if a 19 year old with no formal education is ripping [doors] owned by multi-billion dollar international corporations then the [hobos] have already been there. A company would not even know about the [hobo's] intrusion much less publicize it once they found out so what lulzsec is doing is shining the light on how poorly these companies that hold your data are run.
Not having to have a native mp3 decoder codec on your machine. Also being javascript it is theoretically compatible across all platforms that the browser supports, particular those that may not have native mp3 decoder codecs. The HTML5 standard isn't attempting to establish a standard for codecs. Not saying it's worthwhile or anything, just pointing out potential benefits as you requested.
If Itanium is dead, then why does Intel have all this architectural investment?
Think of bitcoin like being out on the outer rim. Always keep a gun at your hip and always be prepared to get jacked.
Seriously, all the engineers are turning to other industries like software development because companies farm out work for pennies on the dollar to India.
TFA: "Quadriplegic gamers now have around a dozen different actions they can work with their mouth."
Damn, I need a quadriplegic gamer consultant to provide training for my wife.
I don't have any mod points available. Somebody mod this up +Insightful. Recreating original sound is extremely difficult and has a more limited interest in providing equivalent emulation.
Nobody really mentions this, but if ever I download something whether legitimate or not, I rarely if ever see my seed ratio exceed 10:1. So even if I let 10 people fully download a movie from me, the real loss of distribution is only $200 assuming the inflated $20 a box office ticket. Even if we take a potential loss of 100:1, that's still only $2000 and I highly doubt anyone ever really seeds to this amount. Even a $2,000 fine is incentive enough to get someone to stop torrenting copyrighted works and doesn't use ridiculous Dr. Evil math.
I second this. GNS3 rocks, though you'll need to find the Cisco IOS firmware. Typically you can find these with some google searching for a specific IOS firmware name. You also might want to checkout PEMU for PIX emulation. Unfortunately it hogs up the CPU so you'll want to use a CPU limiter on it, but it works really well. Settings up network bridges, particularly in Windows through loopback adapters, can be a bit challenging, but I was able to take an old multi-cpu Dell 2560 with a decent amount of RAM and Windows XP, run GNS3 and PEMU on it, and connect it to a trunk port on our switch (after restricting the VLANs of course) and we were able to use it to interview potential engineers. We had them configure routers, switches, and firewalls to our specification including seting up NATs to our external VLAN to demonstrate they knew how to do simple things like map things to a public IP, etc. We even modeled and tested a redundant ATM bridge infrastructure and got the ATM configuration just right even before the physical equipment came in to our offices for an upcoming project since none of our network engineers had really done it before. To be able to lab that out we would have had to drop a ton of money just to verify we were capable of doing the project correctly and we were able to finalize the production configuration even before the hardware was drop shipped.
It's been a while for me, but they had also gotten Dynamips (included in GNS3) running Cisco ASA code, but it was restricted to 7.X. I'm not really sure if they overcame the issues extracting the ADSM firmware with 8.4 or not. Fortunately for me I went from being a star network admin to being a star programmer and I rarely touch the stuff anymore.
Michael Jackson what?