People who winge and moan and CCTV cameras are obviously either, paraniod, have something to hide, or just plain stupid.
The irony here is that you, in all probability, have no clue why this statement alone provides all the reasoning for not having 24x7 cctv monitoring in public spaces. While I also have nothing to hide, what I am more concerned with is that my opposing cctv monitoring will have all the 'Right Thinking' groupthink types classifying me as a paranoid, criminal or a moron.
When I enter a private space I know that I know that I am likely to be monitored. This is the price I, and others, pay upon entering someone else's space. But I draw the line whenever I am in a public space. I also know that high crime areas benefit from the inclusion of cctv monitoring and I agree with the temporary installation of those devices in those areas. After all, there are only so many police available. But to monitor everywher all the time is much too close to a police state to be reasonable.
No, it won't have the same problem. The cpus that cause the problems are 400MHz/4MB E-cache and the 440MHz/8MB E-cache. (Other may be affected such as the 480s, but I am not sure.)
Those cpus have problems because the cache is, apparently, not ECC. Sun's 'fix' is to clear the cache every Nns. Not really a fix, but at least the box doesn't fall over.
Looking at the specs for the new X1s, the cache has been reduced to 256KB, and although it isn't stated, the cache is likely on the die.
If you are paying $16K a pop for a Netra T1 105, you are on crack. A T1 with a 360/1MB cache and 128MB is ~$3200. Make it a 440/2MB and the price climbs all the way up to ~$5200. Single discs, no cdrom.
The RAM _is_ silly expensive, but even maxed out with two drives and 512MB, we still get them for less than $7000.
Automatic Watches: Rolex, Bell & Ross. Those are still automatic. I'm sure there are others.
Slide rules: Airline pilots use a circular slide rule still. And as of 1994, USMC loadmasters used a version to determine aircraft center of balance.
I hate to be the bearer of bad news (well, not really...), but even at 4.7GB per disc, this technology isn't even close to replacing tapes.
I got a bit of a chuckle out of your statement 'Now stepping out of the business world' though. You see, your entire post _is_ about desktops.
Why is that? I think when you say 'tape' you mean Travan or some form of DDS. While desktop level tape drives are usually in the 4-12GB range, there are many forms of tape that have much more capability.
I am running a relatively low end jukebox with four DLT7000 drives. It is not unusual to get data rates to tape in the 10MB/sec range. Another reason that discs aren't quite ready to replace tapes is that, at least as far as jukeboxes are concerned, you generally don't write one file (or stream of files) to one tape. The backups are threaded onto multiple tapes at the same time.
Did I mention that DLTs are in the 80GB/tape range?
Bzzzzt! Sorry, wrong answer, but thanks for trying.
Although you are correct when you say 'wide lvd scsi arrays do not quite make it then'.
There is much more to enterprise class storage arrays than buzzword compliancy. While lvd scsi wipes the floor with IDE drives, FC-AL arrays wipe the floor with standard scsi.
Once you get into the multi TB range, there are issues involving disk read/write throughput and just plain the overhead with controlling that many discs.
As an example, an EMC FC4500 array has fully redundant CPUs, cache(4GB), power supplies, and controllers. If all of that fails, it has an onboard UPS which writes the contents of the disk cache to a small scsi disk so that all pending writes don't get flushed.
I'm not sure if sgi makes their own arrays, or re-badges like HP does (EMC), but a _small_ EMC fibre attached drive cabinet with ~500GB of disc costs in the range of $125,000.
I'm not sure that, as far as the Gartner Group predictions are concerned, 'end user' friendliness matters. Face it, what this report relates to are enterprise class installations.
Given that, you aren't going to toss a newbie in front of a full on production Unix environment that a large portion of your company is basing a revenue stream on.
Well, I suppose you could, but then again, I doubt that company would reap any benefits from said large installation.
In other words, if a company is using Unix, that company is much more likely to hire experienced admins. Or at least clued individuals.
Oh, you mean the same Linuxcare that was GOING to have an IPO in June, but decided not to because of "market conditions"?
If you would have taken, say 30 seconds or so, the time to check any of the brokerage houses out there, you would have seen that many dozens of privately funded companies chose to delay an IPO.
But it doesn't sound like a support structure I'd want to depend on for mission-critical systems.
What difference does it make if it is a privately owned company or a publicly traded company?
{snip] they expect to do BUSINESS with a BUSINESS.
To further your argument that 'Ask/.' is a good jumping off point for starting a discussion, I find that it is sometimes a good place for me to learn about things I am not currently involved with, but that are interesting nevertheless.
Sure, a search engine is generally where I start to find out about different solutions I might be looking to implement, but it is nice to hear discussion about various things as a 'Gee, that's cool!' discussion also.
This is true if you are counting on a single RAID array for the entire cluster.
For the cluster we are currently installing, there are two dual homed FC disk arrays with 14 disks in each cabinet. Internally the cabinets are grouped in two arrays of seven disks. Each array of disks has four GBICs for a total of 8 GBICs for pair of arrays. Then the arrays are mirrored.
Since each array is dual-homed, and each cabinet in the array has a pair of GBICs to the FC controller on the server, there would have to be a fairly comprehensive set of failures before the cluster lost communication with the disk array.
I suppose the truly paranoid would go ahead and spend $300K plus for something like an EMC, but clustering to one 'RAID' does not imply a single point of failure.
Let me tell you, $1995 is peanuts for failover. We are in the process of installing a cluster/HA failover solution onto a pair of Sun boxen here. The software alone costs $105,000. Heck, even the training Sun requires the local admin to have before the installation is $3495. True, the RH is likely to be very basic compared to what is available for the large enterprise boxen, but I expect that to change rapidly if RH can get the ball rolling.
HP-UX allows you to resize a partition with it's LVM without unmounting the filesystem (on 10.20, I haven't tried this with 11.).
At first it looks like this doesn't work as SAM pukes when it is attempted. The trick is to resize the volume using the command line.
I suppose I should qualify the above by saying that resize == grow. If I recall correctly, shrinking a volume is verboten. I don't have the documentation to back it up, but I also believe that the LVM included with HP-UX is actually a licensed Veritas VxFS.
Sun releases free copies of Solaris/x86, SCO shortly thereafter announces the free version of UnixWare.
If you are going to lambaste a company for something you perceive they did, you can, at a minimum, check your facts. If you did check, those facts, you would find that SCO has been selling 'free' SCO OSen for considerably longer than Sun. I myself purchased both OpenServer and UnixWare waay back in 1997 for the cost of the materials. They may have changed their license recently to be more in line with some of the other pseudo-OSS licenses, but at least they have been providing the software for quite a while.
This is one of those posts where I have to ask:' What's up with the piss poor moderation?' I se plenty of other posts along the same lines, but none of those were marked as troll.
At worst, I can see a possibility for redundant, but 'troll' is just inexcusable.
Using alternate words to make yourself feel better about stealing doens't mean that you haven't stolen.
If I copy someone else's work to make myself look better to my boss, or to get better grades in school, among other things, that is theft. Have I deprived the original author of the work of anything? Sure I have: fair compensation.
The next argument is that 'well, its just a few dollars from millions, so it isn't noticeable'. More crap, it is stil theft.
Someone care to post a short explanation on NUMA for the non-computer engineering types?
I don't know whether an SGI Origin 2000 is 'NUMA', but there was one of those beasts at my last orkplace. Alas, I wasn't allowed to play with it.
The one thing I recall clearly about that machine was the cabling between the nodes. The cabling was rather thick and labelled 'CrayLINK'.
I assume that this was some type of high speed/bandwidth inter-node communications channel?
Re:This is the wrong question
on
Linux Failover?
·
· Score: 4
why do you need dual-port NICs?
On Suns at least, the dual (well, quad) port NICs are used as a heartbeat signal between the active server and the failover box (when using FirstWatch).
True, you could use two separate NICs in each box to provide the same solution, but then you are using up three PCI slots since the heartbeat NICs do not carry any packets.
I am wondering why NICs with more than one port are so danged expensive though? I can see a bit of an increase in price, but there is no way these things should be $400 and up (last time I looked..)
I am curious as to whether (no pun intended...:)) or not you have ever done any testing to see if a distributed.net type enviornment would be useful for your type of work?
It seems to me that there are more than a few people who are willing to donate spare cpu cycles for various projects. At a minimum. you could concentrate on the client side binaries and not worry as mouch about hardware issues.
Yeah, but why do you assume that assembly line workers are less intelligent? You have what appears to be an amazingly low estimation of people's intelligence.
True, as a sysadmin, I know that there is a good percentage of lusers out there, but I don't think that percentage is any different across different work enviornments.
Hehe...the are likely _more_ lusers in a computing enviornment, but they just think they are smarter.
True, the cache is 'off-die', but it does run at cpu speed. This is why the Sparc modules have about 2 pounds worth of heatsink on them.
It is also why a 450MHz/8MB Ultra module costs $9500. (Note: I haven't looked in quite a while, so that price was when I last looked -Dec, 99-, it is likely less now, and I am sure big customers get deep discounts.....)
jmv is more correct than I in that it isn't strictly the cpu which is designed to 'handle' the load. I should have been more clear in that regard.
What I should have said is that both the cpu and the OS are designed to handle higher loads more efficiently. With Sparc cpus, task switching is aided by the Context registers built into the silicon. I am no hardware designer, so I couldn't say with any authority how x86 handles task switching.
So, to answer one of the above respondents, by load I mean how many tasks a cpu is required to switch between, and how efficiently both the cpu architecture and the OS can do that.
Don't think 'speed' when I say 'load'. Rather compare the ability of the machine to respond when the 'loadavg' starts approaching 1 or greater per cpu.
Slashdot Mirror
People who winge and moan and CCTV cameras are obviously either, paraniod, have something to hide, or just plain stupid.
The irony here is that you, in all probability, have no clue why this statement alone provides all the reasoning for not having 24x7 cctv monitoring in public spaces. While I also have nothing to hide, what I am more concerned with is that my opposing cctv monitoring will have all the 'Right Thinking' groupthink types classifying me as a paranoid, criminal or a moron.
When I enter a private space I know that I know that I am likely to be monitored. This is the price I, and others, pay upon entering someone else's space. But I draw the line whenever I am in a public space. I also know that high crime areas benefit from the inclusion of cctv monitoring and I agree with the temporary installation of those devices in those areas. After all, there are only so many police available. But to monitor everywher all the time is much too close to a police state to be reasonable.
No, it won't have the same problem. The cpus that cause the problems are 400MHz/4MB E-cache and the 440MHz/8MB E-cache. (Other may be affected such as the 480s, but I am not sure.)
Those cpus have problems because the cache is, apparently, not ECC. Sun's 'fix' is to clear the cache every Nns. Not really a fix, but at least the box doesn't fall over.
Looking at the specs for the new X1s, the cache has been reduced to 256KB, and although it isn't stated, the cache is likely on the die.
If you are paying $16K a pop for a Netra T1 105, you are on crack. A T1 with a 360/1MB cache and 128MB is ~$3200. Make it a 440/2MB and the price climbs all the way up to ~$5200. Single discs, no cdrom.
The RAM _is_ silly expensive, but even maxed out with two drives and 512MB, we still get them for less than $7000.
Automatic Watches: Rolex, Bell & Ross. Those are still automatic. I'm sure there are others.
Slide rules: Airline pilots use a circular slide rule still. And as of 1994, USMC loadmasters used a version to determine aircraft center of balance.
I hate to be the bearer of bad news (well, not really...), but even at 4.7GB per disc, this technology isn't even close to replacing tapes.
I got a bit of a chuckle out of your statement 'Now stepping out of the business world' though. You see, your entire post _is_ about desktops.
Why is that? I think when you say 'tape' you mean Travan or some form of DDS. While desktop level tape drives are usually in the 4-12GB range, there are many forms of tape that have much more capability.
I am running a relatively low end jukebox with four DLT7000 drives. It is not unusual to get data rates to tape in the 10MB/sec range. Another reason that discs aren't quite ready to replace tapes is that, at least as far as jukeboxes are concerned, you generally don't write one file (or stream of files) to one tape. The backups are threaded onto multiple tapes at the same time.
Did I mention that DLTs are in the 80GB/tape range?
Dang, even 'previewing' didn't help.....
Of course, the UPS doesn't write the data to the drive, it only powers the remaining cpu long enough for the cache to flush all data to the drive.
Bzzzzt! Sorry, wrong answer, but thanks for trying.
Although you are correct when you say 'wide lvd scsi arrays do not quite make it then'.
There is much more to enterprise class storage arrays than buzzword compliancy. While lvd scsi wipes the floor with IDE drives, FC-AL arrays wipe the floor with standard scsi.
Once you get into the multi TB range, there are issues involving disk read/write throughput and just plain the overhead with controlling that many discs.
As an example, an EMC FC4500 array has fully redundant CPUs, cache(4GB), power supplies, and controllers. If all of that fails, it has an onboard UPS which writes the contents of the disk cache to a small scsi disk so that all pending writes don't get flushed.
I'm not sure if sgi makes their own arrays, or re-badges like HP does (EMC), but a _small_ EMC fibre attached drive cabinet with ~500GB of disc costs in the range of $125,000.
I'm not sure that, as far as the Gartner Group predictions are concerned, 'end user' friendliness matters. Face it, what this report relates to are enterprise class installations.
Given that, you aren't going to toss a newbie in front of a full on production Unix environment that a large portion of your company is basing a revenue stream on.
Well, I suppose you could, but then again, I doubt that company would reap any benefits from said large installation.
In other words, if a company is using Unix, that company is much more likely to hire experienced admins. Or at least clued individuals.
Oh, you mean the same Linuxcare that was GOING to have an IPO in June, but decided not to because of "market conditions"?
If you would have taken, say 30 seconds or so, the time to check any of the brokerage houses out there, you would have seen that many dozens of privately funded companies chose to delay an IPO.
But it doesn't sound like a support structure I'd want to depend on for mission-critical systems.
What difference does it make if it is a privately owned company or a publicly traded company?
{snip] they expect to do BUSINESS with a BUSINESS.
Since when is a private business not a business?
To further your argument that 'Ask /.' is a good jumping off point for starting a discussion, I find that it is sometimes a good place for me to learn about things I am not currently involved with, but that are interesting nevertheless.
Sure, a search engine is generally where I start to find out about different solutions I might be looking to implement, but it is nice to hear discussion about various things as a 'Gee, that's cool!' discussion also.
SKU == Stock Keeping Unit, ie: part number.
This is true if you are counting on a single RAID array for the entire cluster.
For the cluster we are currently installing, there are two dual homed FC disk arrays with 14 disks in each cabinet. Internally the cabinets are grouped in two arrays of seven disks. Each array of disks has four GBICs for a total of 8 GBICs for pair of arrays. Then the arrays are mirrored.
Since each array is dual-homed, and each cabinet in the array has a pair of GBICs to the FC controller on the server, there would have to be a fairly comprehensive set of failures before the cluster lost communication with the disk array.
I suppose the truly paranoid would go ahead and spend $300K plus for something like an EMC, but clustering to one 'RAID' does not imply a single point of failure.
Let me tell you, $1995 is peanuts for failover.
We are in the process of installing a cluster/HA failover solution onto a pair of Sun boxen here. The software alone costs $105,000.
Heck, even the training Sun requires the local admin to have before the installation is $3495.
True, the RH is likely to be very basic compared to what is available for the large enterprise boxen, but I expect that to change rapidly if RH can get the ball rolling.
After digesting some Ritalin, try and discover the difference between 'sponsoring' and 'bundling'.
I have a BusLogic BT-948 UltraSCSI controller with an Intel 80186 processor.
HP-UX allows you to resize a partition with it's LVM without unmounting the filesystem (on 10.20, I haven't tried this with 11.).
At first it looks like this doesn't work as SAM pukes when it is attempted. The trick is to resize the volume using the command line.
I suppose I should qualify the above by saying that resize == grow. If I recall correctly, shrinking a volume is verboten. I don't have the documentation to back it up, but I also believe that the LVM included with HP-UX is actually a licensed Veritas VxFS.
Sun releases free copies of Solaris/x86, SCO shortly thereafter announces the free version of UnixWare.
If you are going to lambaste a company for something you perceive they did, you can, at a minimum, check your facts. If you did check, those facts, you would find that SCO has been selling 'free' SCO OSen for considerably longer than Sun.
I myself purchased both OpenServer and UnixWare waay back in 1997 for the cost of the materials.
They may have changed their license recently to be more in line with some of the other pseudo-OSS licenses, but at least they have been providing the software for quite a while.
This is one of those posts where I have to ask:' What's up with the piss poor moderation?' I se plenty of other posts along the same lines, but none of those were marked as troll.
At worst, I can see a possibility for redundant, but 'troll' is just inexcusable.
Sorry, just my two cents....
Using alternate words to make yourself feel better about stealing doens't mean that you haven't stolen.
If I copy someone else's work to make myself look better to my boss, or to get better grades in school, among other things, that is theft. Have I deprived the original author of the work of anything? Sure I have: fair compensation.
The next argument is that 'well, its just a few dollars from millions, so it isn't noticeable'. More crap, it is stil theft.
Someone care to post a short explanation on NUMA for the non-computer engineering types?
I don't know whether an SGI Origin 2000 is 'NUMA', but there was one of those beasts at my last orkplace. Alas, I wasn't allowed to play with it.
The one thing I recall clearly about that machine was the cabling between the nodes. The cabling was rather thick and labelled 'CrayLINK'.
I assume that this was some type of high speed/bandwidth inter-node communications channel?
why do you need dual-port NICs?
On Suns at least, the dual (well, quad) port NICs are used as a heartbeat signal between the active server and the failover box (when using FirstWatch).
True, you could use two separate NICs in each box to provide the same solution, but then you are using up three PCI slots since the heartbeat NICs do not carry any packets.
I am wondering why NICs with more than one port are so danged expensive though? I can see a bit of an increase in price, but there is no way these things should be $400 and up (last time I looked..)
I am curious as to whether (no pun intended...:)) or not you have ever done any testing to see if a distributed.net type enviornment would be useful for your type of work?
It seems to me that there are more than a few people who are willing to donate spare cpu cycles for various projects. At a minimum. you could concentrate on the client side binaries and not worry as mouch about hardware issues.
Yeah, but why do you assume that assembly line workers are less intelligent? You have what appears to be an amazingly low estimation of people's intelligence.
True, as a sysadmin, I know that there is a good percentage of lusers out there, but I don't think that percentage is any different across different work enviornments.
Hehe...the are likely _more_ lusers in a computing enviornment, but they just think they are smarter.
True, the cache is 'off-die', but it does run at cpu speed. This is why the Sparc modules have about 2 pounds worth of heatsink on them.
It is also why a 450MHz/8MB Ultra module costs $9500. (Note: I haven't looked in quite a while, so that price was when I last looked -Dec, 99-, it is likely less now, and I am sure big customers get deep discounts.....)
I'll try and kill two birds with one stone.
jmv is more correct than I in that it isn't strictly the cpu which is designed to 'handle' the load. I should have been more clear in that regard.
What I should have said is that both the cpu and the OS are designed to handle higher loads more efficiently. With Sparc cpus, task switching is aided by the Context registers built into the silicon. I am no hardware designer, so I couldn't say with any authority how x86 handles task switching.
So, to answer one of the above respondents, by load I mean how many tasks a cpu is required to switch between, and how efficiently both the cpu architecture and the OS can do that.
Don't think 'speed' when I say 'load'. Rather compare the ability of the machine to respond when the 'loadavg' starts approaching 1 or greater per cpu.