"I've searched several websites (including Tom's Hardware), and I've googled, but can't seem to locate any commercially available AMD MBs supporting more than 4 sticks of RAM, or 4 gigs."
I'm actually building a system right now. Dual Opteron 224. The Tyan Thunder server boards are your best bet. This is the one going in my system:
http://www.tyan.com/products/html/thunderk8we.ht ml
It has 8 DIMM slots and supports 16GB of memory.
You just aren't going to find any i386 archs to support more than that right now. We have a few x345 servers from IBM that support 32GB but we run with memory mirroring for a total of 16GB rather than the full 32GB.
If you REALLY need more memory than that you're going to have to jump to pSeries/POWER5 line. You might be able to find a used POWER4 on ebay.
Even then I think our p570's only support up to 32GB using 2GB memory modules PER chassis/CEC/building block.
I'm not sure what your application need for the memory is but if you really need 64GB, you may have to buy a couple of servers and use something like memcached to get the full benefit. There is support for memcached in almost every language out there.
I don't like either of the parties but back in November or some such I posted the exact same thing. Democrats bitch and bitch about uncounted votes and I warned that the same rules they want are fair game for the "other" side.
I just keep warning all of the republicans I know that this path they are going down now (catering to the extreme right) is going to bite them in the end when the same laws they pushed through are used against them.
It never ceases to amaze me when politicians are so shortsighted that they pass x law to keep appease y crowd and keep the job for a few more years. Then 10 years later x law is used to oppress y crowd and everyone bitches. I'm waiting for the day that donating to the Christian Coalition is deemed sponsoring terrorism because they gave money to the Eric Rudolph Anti-abortion foundation which blew up a clinic with 30 people inside.
I swear to $DIETY that New Hampshire has is right when the state legislature is only allowed to meet for a very brief time period (I think 2 weeks?).
Can't fuck up much in two weeks. Then again politicians never let me down in the sheer mass of things they can fuck up on a daily basis.
And wonder why in the hell there needs to be a freaking panel on opensource software in any government.
I've read the articles left and right about hinderances to government implementing non-COTS environments but in the end it's just software!
I don't remember these kinds of panels and hububaloo about implementing Windows here and there. Was there an implementation panel to provide service and support for Windows when it was brought into the Australian government?
I'm all for open source as I like getting a paycheck but some of these "program" and "panels" and committees strike me as another sign of government waste.
Where's the benifit of using an opensource solution when it takes a panel to advise to implement? It's just annoying that people want to make this into something that requires an army of consultants and panelists to do when it's really about just buying software.
Here's the full story. We were building our new TSM server. It's connected via fiber to a 3581 tape library. When we did the install from the media shipped by Redhat, it contained 2.4.21-z. The IBM Tape driver was built specifically for 2.4.21-y which was an older version number. We had to run the newer version because (A) the older wasn't available and (B) the driver for our HBA didn't work properly with the older one anyway.
Checking RHN and contacting Redhat proved that we could not get the older kernel version ANYWHERE. We basically ran on the new version after manually extracting the IBM rpm file and installing it.
We ran that way until IBM got in sync with our kernel which was about 6 months. Now IBM is mirroring the current kernel releases really closely.
You've obviously never installed the IBMTape driver then.
This fucking bastard actually checked the sublevel of the redhat kernel.
So basically, while I might be running:
2.4.21 on RHAS 3.0
IBM would check for the freaking sub-release number so I had to be running: 2.4.21-27.0.2.ELsmp
I couldn't run: 2.4.21-9.ELsmp
Oddly enough when we had no choice but to use a newer kernel version and IBM hadn't updated the tape driver, we were able to manually extract everything from the RPM (The goddamned RPM refused to install unless it saw -27.0.2.ELsmp for example) and the driver worked fine.
If your curious why we had to upgrade, Fiber Channel HBAs.
The only place they are getting creamed in the datacenter is places that DON'T run commercial products under WBEL/CentOS.
I would love to see some corporate employee call up IBM or Oracle and say "I can't my database server working." "Are you running RHEL or Suse?" "Neither" "Call us back when you are."
Sure you can lie and say that you're running RHEL2.1 but the first time you submit an egather report to IBM and it says you're running WBEL, they'll tell you to call back when you are.
The only thing Redhat has to worry about is when a vendor OFFICIALLY supports WBEL/CentOS at the same level as RHEL/Suse.
Snort has a flexresp package that will update quote a few vendor firewalls and routers to block machines that are sending suspicious traffic.
And by "in a sense" I mean your time is money and the time it takes to really tune your snort rules properly is a big investment. You don't want to have your CEO come bitching at you because his windows update set off a NOOP x86 shellcode attach on the snort filters;)
While I would agree, there is a big enough problem with Clueless Admins blocking ALL ICMP and breaking things for the rest of the world.
Let me put it here for future generations. If you want to block ICMP, go ahead but leave the following enabled and the rest of the fucking internet will thank you:
echo echo-reply Packet too big TTL Exceeded Unreach
Go ahead and block the fragments and all others but leave those allowed so that things don't break.
If you really get snarky go ahead and block echo/echo-reply but the other three are musts to be allowed.
n general I think 1GB is good for 512MB systems, 1.5GB is good for 1GB systems, and 2GB is good for 2GB systems.
There is no frelling way I'm going to set a swap space to 2GB. I'm sorry. I made this mistake with a production server and I've been paying for it ever since.
Sure the general rule used to be double your physical memory but that rule just doesn't fly anymore.
We've got a few RHEL servers that were installed with 2GB of memory. I couldn't bring myself to create a 4GB swap space so I set it to 2GB. It was the single worst choice I've ever made.
There is no way in hell I want to swap out a full 2GB of memory. If my system needs to swap out a full 2GB of something, I've got other issues. There is no way you're going to be able to fit that back in when it wants to go from swap to RAM so something else is going to get paged out and the cycle continues.
I've contented myself to set a max of 768MB no matter how much memory I have. One of my DB2 servers has 16GB of RAM. There is no way I'm creating a 32GB swap vol much less a 16GB one.
Actually I was talking about third-party code. I have no problem using pgsql and php;)
I was thinking more along the lines of jrandom php package that seems really cool but has decided to use mysql_* throughout the code instead of something like adodb or peardb.
Actually I was talking about third-party code. I have no problem using pgsql and php;)
In fact I've got about 5 internal applications running that use it right now.
I was thinking more along the lines of jrandom php package that seems really cool but has decided to use mysql_* throughout the code instead of something like adodb or peardb.
When you can take php.net into the bathroom or mark all over the pages, call me. I use php.net almost exlusively but alot of these types of books give a better explaination than the website for specific functions.
I've actually been planning on migrating our internal servers to 5 but I really need to revisit my code on alot of things.
I used the simpler OO in version 4 everywhere so I want to make sure I don't break anything. My favorite libraries - adodb and smarty - have all been ported to 5 so I don't see any problems there.
The biggest problem is support in some third-party apps we are looking at using. That and the fact that they don't freakin support postgresql which is where we're header with our internal databases for this type of stuff.
You know as well as I do you aren't forced to use GPL or even LGPL software in your commercial product. You are perfectly free to write everything yourself or choose something from the BSD license camp.
I love how different people define freedom these days. It means one thing to the Bush Administration and another thing to John Gilmore.
If you want my personal opinion, which you don't, BSD licenses view free in terms of free to do whatever you want with it. the GPL licenses view free as in the code must remain free in terms of it must be available for all people at all times. The GPL folks figure the way to do this is by drafting the terms under which you make use of the code.
Freedom minded people might also want to appreciate someone ELSE'S freedom to distribute his works under whatever license, agreement and policy they wish.
It's a double-standard pure and simple. I'm not saying that the majority of slashdotters are downloading movies and music against copyright law unlike the GP but I will say this.
The same fucking power and ruleset that the GPL is using is the same fucking power and ruleset that copyright holders are using.
The GPL is a copyright. Copyleft is just a cutsie term that the FSF attached to it. I love the GPL. I love libre software. I appreciate the FSF working WITHIN the law just as the Nature Conservancy works with contract law to preserve greenspace.
The question of someone's vigilante method of serving notices (I can't read the original site to determine if he is operating as counsel for the FSF) is a different issue. The pure and simple fact is that people who download or distribute music against the original terms of copyright are just as shitty as people who download or distribute GPL code against the terms of the GPL.
Actually you misunderstand the GPL in a way that many do.
They do not HAVE to provide a download link to the source. They don't have to provide the source on a website. They don't have to provide ANYTHING unless you ask for it at which point they can charge you a reasonable fee for media and shipping.
Why not go pSeries. You can get a nice 2 way p5 for around $4k US. It won't run Windows but I don't think you'll have an issue with SAP on it. It can run AIX or Linux. IBM *tells* me that a p4 is about the same as 2.5 Xeons. Those numbers don't really mean much to me but the fact that it's 64bit does. And I have a few choices of OS's and LPARs to boot.
See the following comment. I was way off base. I was including everything I purchased on that order and not just the parts for her machine. I bought one of those samsung monitors for myself as well.
I don't know how well this will format but here it is. I admit now that I was way off the cost. I was actually including some components that I bought myself on the same order. I just went back and looked at the final invoice. I added up JUST her parts and it came out to around $1900.
Since you called bullshit, do I have to pick up the cards? hehe.
Monarch Computer
80232 Barebone AMD Athlon 64( 939 ) Pre-Tested and Setup System 1 $18.00 $18.00 100080 LIAN LI PC-V1000 ( BLACK ) ALUMINUM QUIET TOWER 1 $189.00 $189.00 100311 Antec True Power 480W Power Supply Gold Plated Connectors 1 $91.00 $91.00 110228 MSI K8N Neo2 Platinum Socket 939 for AMD® Athlon 64FX/64 processor Supports 3500+, 3800+ Athlon 64FX 53, or higher CPU 1 $139.00 $139.00 120421 AMD Athlon 64 3000 939 Pin ADA3000DIK4BI 1 $149.00 $149.00 130150 Thermaltake (A1772) AMD Opteron & Athlon 64 Heatsink and fan. AMD Recomended 1 $35.00 $35.00 800018 Thermal Grease, Shin-Etsu G675, apply combo(cools cpu better) 1 $14.00 $14.00 140803 CORSAIR TWINX1024-3200C2PT DDR 400 1GB (2 pcs 512) 1 $199.00 $199.00 800036 MB-CPU COMBO 3 Y WARRANTYw/30 DAY FRIEGHT RETURN PREPAID 1 $29.95 $29.95 230122 Logitech® Cordless MX Duo Cordless Enhanced Function Keys Mac Multimedia Control Optical PC PS/2 Scroll wheel USB Zero Degree Tilt 1 $59.00 $59.00 250234 Logitech X-530 5.1 Speaker with Subwoofer -RETAIL 970114-0403 1 $58.00 $58.00 160941 SAMSUNG DVD+-RW 16x DUAL LAYER BULK Black TS-H552B/WBCH WITH SOFTWARE 1 $74.00 $74.00 170108 MITSUMI FLOPPY,7 in 1, BLACK,CARD,READER 1 $49.00 $49.00
NewEgg
22-145-072 HD 160GB|HITC SATA 8MB 14R9463 % 2 $93.06 $186.12 14-150-080 VGA XFX|GF 6600 GT 128MB PVT43AND 1 $240.50 $240.50
Well mind you that this includes speakers, high-end video card, lcd monitor, raid 1 in hardware.
I built a machine that will be future proof for the foreseeable future. I wanted it to last as long as possible. I feel really good about the whole process.
I didn't buy a dell the past christmas for my wife.
I went round and round trying to find a good vendor to provide an AMD64 machine for my wife to do her video and photo stuff on and I actually looked at Dell. I searched all over and found that they didn't sell a single AMD64 machine.
So I walked away.
I ended up spending around 3k for a machine with Monarch/NewEgg and did some of the assembly myself but my wife now has a computer that I won't have to upgrade for quite some time. That includes the 3 year warranty from Monarch.
I understand that Dell keeps costs inline by pulling an assembly line approach but this is going to bite them in the ass even more as people start demanding x86_64 and Dell can't come to market. EMT64 just doesn't cut it in my mind.
Hell even IBM is starting to introduce AMD64 in it's xSeries line.
You know, as a side note, when we we're spec'ing out our p570 boxen and we we're concerned with redundancy (do we go 2 lpar on an 8-way 570 or two 4-way 570?). I was grilling a guy about all the points of failure along the way (this is a huge financial database for all of our stores). We finally got down to the hypervisor level.
I asked him what happens if the hypervisor fails.
He didn't have an answer. He seemed flabergasted that I would ask that. When he called back after consulting the engineers, he said that indeed, if the hypervisor failed, all LPARS would be unavailable.
He told me though that the Hypervisor NEVER failed.
I'll find out down the road when we get our pSeries install finished.
We decided to go 2 LPAR for failover and bought a separate 4 way as a ultimate backup.
Anyone ever seen a hypervisor fail? How ugly is it?
Slashdot Mirror
"I've searched several websites (including Tom's Hardware), and I've googled, but can't seem to locate any commercially available AMD MBs supporting more than 4 sticks of RAM, or 4 gigs."
t ml
I'm actually building a system right now. Dual Opteron 224. The Tyan Thunder server boards are your best bet. This is the one going in my system:
http://www.tyan.com/products/html/thunderk8we.h
It has 8 DIMM slots and supports 16GB of memory.
You just aren't going to find any i386 archs to support more than that right now. We have a few x345 servers from IBM that support 32GB but we run with memory mirroring for a total of 16GB rather than the full 32GB.
If you REALLY need more memory than that you're going to have to jump to pSeries/POWER5 line. You might be able to find a used POWER4 on ebay.
Even then I think our p570's only support up to 32GB using 2GB memory modules PER chassis/CEC/building block.
I'm not sure what your application need for the memory is but if you really need 64GB, you may have to buy a couple of servers and use something like memcached to get the full benefit. There is support for memcached in almost every language out there.
I don't like either of the parties but back in November or some such I posted the exact same thing. Democrats bitch and bitch about uncounted votes and I warned that the same rules they want are fair game for the "other" side.
I just keep warning all of the republicans I know that this path they are going down now (catering to the extreme right) is going to bite them in the end when the same laws they pushed through are used against them.
It never ceases to amaze me when politicians are so shortsighted that they pass x law to keep appease y crowd and keep the job for a few more years. Then 10 years later x law is used to oppress y crowd and everyone bitches. I'm waiting for the day that donating to the Christian Coalition is deemed sponsoring terrorism because they gave money to the Eric Rudolph Anti-abortion foundation which blew up a clinic with 30 people inside.
I swear to $DIETY that New Hampshire has is right when the state legislature is only allowed to meet for a very brief time period (I think 2 weeks?).
Can't fuck up much in two weeks. Then again politicians never let me down in the sheer mass of things they can fuck up on a daily basis.
And wonder why in the hell there needs to be a freaking panel on opensource software in any government.
I've read the articles left and right about hinderances to government implementing non-COTS environments but in the end it's just software!
I don't remember these kinds of panels and hububaloo about implementing Windows here and there. Was there an implementation panel to provide service and support for Windows when it was brought into the Australian government?
I'm all for open source as I like getting a paycheck but some of these "program" and "panels" and committees strike me as another sign of government waste.
Where's the benifit of using an opensource solution when it takes a panel to advise to implement? It's just annoying that people want to make this into something that requires an army of consultants and panelists to do when it's really about just buying software.
Actually I'm not running CentOS or WBEL but RHEL.
Here's the full story. We were building our new TSM server. It's connected via fiber to a 3581 tape library. When we did the install from the media shipped by Redhat, it contained 2.4.21-z. The IBM Tape driver was built specifically for 2.4.21-y which was an older version number. We had to run the newer version because (A) the older wasn't available and (B) the driver for our HBA didn't work properly with the older one anyway.
Checking RHN and contacting Redhat proved that we could not get the older kernel version ANYWHERE. We basically ran on the new version after manually extracting the IBM rpm file and installing it.
We ran that way until IBM got in sync with our kernel which was about 6 months. Now IBM is mirroring the current kernel releases really closely.
You've obviously never installed the IBMTape driver then.
This fucking bastard actually checked the sublevel of the redhat kernel.
So basically, while I might be running:
2.4.21 on RHAS 3.0
IBM would check for the freaking sub-release number so I had to be running:
2.4.21-27.0.2.ELsmp
I couldn't run:
2.4.21-9.ELsmp
Oddly enough when we had no choice but to use a newer kernel version and IBM hadn't updated the tape driver, we were able to manually extract everything from the RPM (The goddamned RPM refused to install unless it saw -27.0.2.ELsmp for example) and the driver worked fine.
If your curious why we had to upgrade, Fiber Channel HBAs.
The only place they are getting creamed in the datacenter is places that DON'T run commercial products under WBEL/CentOS.
I would love to see some corporate employee call up IBM or Oracle and say "I can't my database server working." "Are you running RHEL or Suse?" "Neither" "Call us back when you are."
Sure you can lie and say that you're running RHEL2.1 but the first time you submit an egather report to IBM and it says you're running WBEL, they'll tell you to call back when you are.
The only thing Redhat has to worry about is when a vendor OFFICIALLY supports WBEL/CentOS at the same level as RHEL/Suse.
Sure it's cheap in a sense.
;)
Snort has a flexresp package that will update quote a few vendor firewalls and routers to block machines that are sending suspicious traffic.
And by "in a sense" I mean your time is money and the time it takes to really tune your snort rules properly is a big investment. You don't want to have your CEO come bitching at you because his windows update set off a NOOP x86 shellcode attach on the snort filters
While I would agree, there is a big enough problem with Clueless Admins blocking ALL ICMP and breaking things for the rest of the world.
Let me put it here for future generations. If you want to block ICMP, go ahead but leave the following enabled and the rest of the fucking internet will thank you:
echo
echo-reply
Packet too big
TTL Exceeded
Unreach
Go ahead and block the fragments and all others but leave those allowed so that things don't break.
If you really get snarky go ahead and block echo/echo-reply but the other three are musts to be allowed.
It could have been a public IP address but that doesn't mean it was publicly available. Think MPLS.
n general I think 1GB is good for 512MB systems, 1.5GB is good for 1GB systems, and 2GB is good for 2GB systems.
There is no frelling way I'm going to set a swap space to 2GB. I'm sorry. I made this mistake with a production server and I've been paying for it ever since.
Sure the general rule used to be double your physical memory but that rule just doesn't fly anymore.
We've got a few RHEL servers that were installed with 2GB of memory. I couldn't bring myself to create a 4GB swap space so I set it to 2GB. It was the single worst choice I've ever made.
There is no way in hell I want to swap out a full 2GB of memory. If my system needs to swap out a full 2GB of something, I've got other issues. There is no way you're going to be able to fit that back in when it wants to go from swap to RAM so something else is going to get paged out and the cycle continues.
I've contented myself to set a max of 768MB no matter how much memory I have. One of my DB2 servers has 16GB of RAM. There is no way I'm creating a 32GB swap vol much less a 16GB one.
Actually I was talking about third-party code. I have no problem using pgsql and php ;)
I was thinking more along the lines of jrandom php package that seems really cool but has decided to use mysql_* throughout the code instead of something like adodb or peardb.
Actually I was talking about third-party code. I have no problem using pgsql and php ;)
In fact I've got about 5 internal applications running that use it right now.
I was thinking more along the lines of jrandom php package that seems really cool but has decided to use mysql_* throughout the code instead of something like adodb or peardb.
When you can take php.net into the bathroom or mark all over the pages, call me. I use php.net almost exlusively but alot of these types of books give a better explaination than the website for specific functions.
I've actually been planning on migrating our internal servers to 5 but I really need to revisit my code on alot of things.
I used the simpler OO in version 4 everywhere so I want to make sure I don't break anything. My favorite libraries - adodb and smarty - have all been ported to 5 so I don't see any problems there.
The biggest problem is support in some third-party apps we are looking at using. That and the fact that they don't freakin support postgresql which is where we're header with our internal databases for this type of stuff.
I'll feed the troll.
You know as well as I do you aren't forced to use GPL or even LGPL software in your commercial product. You are perfectly free to write everything yourself or choose something from the BSD license camp.
I love how different people define freedom these days. It means one thing to the Bush Administration and another thing to John Gilmore.
If you want my personal opinion, which you don't, BSD licenses view free in terms of free to do whatever you want with it. the GPL licenses view free as in the code must remain free in terms of it must be available for all people at all times. The GPL folks figure the way to do this is by drafting the terms under which you make use of the code.
Freedom minded people might also want to appreciate someone ELSE'S freedom to distribute his works under whatever license, agreement and policy they wish.
It's a double-standard pure and simple. I'm not saying that the majority of slashdotters are downloading movies and music against copyright law unlike the GP but I will say this.
The same fucking power and ruleset that the GPL is using is the same fucking power and ruleset that copyright holders are using.
The GPL is a copyright. Copyleft is just a cutsie term that the FSF attached to it. I love the GPL. I love libre software. I appreciate the FSF working WITHIN the law just as the Nature Conservancy works with contract law to preserve greenspace.
The question of someone's vigilante method of serving notices (I can't read the original site to determine if he is operating as counsel for the FSF) is a different issue. The pure and simple fact is that people who download or distribute music against the original terms of copyright are just as shitty as people who download or distribute GPL code against the terms of the GPL.
Actually you misunderstand the GPL in a way that many do.
They do not HAVE to provide a download link to the source. They don't have to provide the source on a website. They don't have to provide ANYTHING unless you ask for it at which point they can charge you a reasonable fee for media and shipping.
You have to understand the Mozilla mindset on development.
Mozilla suite is the reference platform. Pure and simple. It was intended for people to spin off thier own projects.
Firefox, Thunderbird and Sunbird are all spinoffs from the mozilla code base. Sunbird was actually the result of a bunch of work done by OEone, IIRC.
Why not go pSeries. You can get a nice 2 way p5 for around $4k US. It won't run Windows but I don't think you'll have an issue with SAP on it. It can run AIX or Linux. IBM *tells* me that a p4 is about the same as 2.5 Xeons. Those numbers don't really mean much to me but the fact that it's 64bit does. And I have a few choices of OS's and LPARs to boot.
See the following comment. I was way off base. I was including everything I purchased on that order and not just the parts for her machine. I bought one of those samsung monitors for myself as well.
i d= 11797633
http://it.slashdot.org/comments.pl?sid=140748&c
Shit I just realized you were talking about the guy who answered me. My bad. But it did make me go back and look to see how much I spent.
I don't know how well this will format but here it is. I admit now that I was way off the cost. I was actually including some components that I bought myself on the same order. I just went back and looked at the final invoice. I added up JUST her parts and it came out to around $1900.
,CARD,READER 1 $49.00 $49.00
Since you called bullshit, do I have to pick up the cards? hehe.
Monarch Computer
80232 Barebone AMD Athlon 64( 939 ) Pre-Tested and Setup System 1 $18.00 $18.00
100080 LIAN LI PC-V1000 ( BLACK ) ALUMINUM QUIET TOWER 1 $189.00 $189.00
100311 Antec True Power 480W Power Supply Gold Plated Connectors 1 $91.00 $91.00
110228 MSI K8N Neo2 Platinum Socket 939 for AMD® Athlon 64FX/64 processor Supports 3500+, 3800+ Athlon 64FX 53, or higher CPU 1 $139.00 $139.00
120421 AMD Athlon 64 3000 939 Pin ADA3000DIK4BI 1 $149.00 $149.00
130150 Thermaltake (A1772) AMD Opteron & Athlon 64 Heatsink and fan. AMD Recomended 1 $35.00 $35.00
800018 Thermal Grease, Shin-Etsu G675, apply combo(cools cpu better) 1 $14.00 $14.00
140803 CORSAIR TWINX1024-3200C2PT DDR 400 1GB (2 pcs 512) 1 $199.00 $199.00
800036 MB-CPU COMBO 3 Y WARRANTYw/30 DAY FRIEGHT RETURN PREPAID 1 $29.95 $29.95
230122 Logitech® Cordless MX Duo Cordless Enhanced Function Keys Mac Multimedia Control Optical PC PS/2 Scroll wheel USB Zero Degree Tilt 1 $59.00 $59.00
250234 Logitech X-530 5.1 Speaker with Subwoofer -RETAIL 970114-0403 1 $58.00 $58.00
160941 SAMSUNG DVD+-RW 16x DUAL LAYER BULK Black TS-H552B/WBCH WITH SOFTWARE 1 $74.00 $74.00
170108 MITSUMI FLOPPY,7 in 1, BLACK
NewEgg
22-145-072 HD 160GB|HITC SATA 8MB 14R9463 % 2 $93.06 $186.12
14-150-080 VGA XFX|GF 6600 GT 128MB PVT43AND 1 $240.50 $240.50
Techdepot
SAMSUNG 17'' 710N-Black LCD Monitor 1 $329.99 *
Well mind you that this includes speakers, high-end video card, lcd monitor, raid 1 in hardware.
I built a machine that will be future proof for the foreseeable future. I wanted it to last as long as possible. I feel really good about the whole process.
I didn't buy a dell the past christmas for my wife.
I went round and round trying to find a good vendor to provide an AMD64 machine for my wife to do her video and photo stuff on and I actually looked at Dell. I searched all over and found that they didn't sell a single AMD64 machine.
So I walked away.
I ended up spending around 3k for a machine with Monarch/NewEgg and did some of the assembly myself but my wife now has a computer that I won't have to upgrade for quite some time. That includes the 3 year warranty from Monarch.
I understand that Dell keeps costs inline by pulling an assembly line approach but this is going to bite them in the ass even more as people start demanding x86_64 and Dell can't come to market. EMT64 just doesn't cut it in my mind.
Hell even IBM is starting to introduce AMD64 in it's xSeries line.
You know, as a side note, when we we're spec'ing out our p570 boxen and we we're concerned with redundancy (do we go 2 lpar on an 8-way 570 or two 4-way 570?). I was grilling a guy about all the points of failure along the way (this is a huge financial database for all of our stores). We finally got down to the hypervisor level.
I asked him what happens if the hypervisor fails.
He didn't have an answer. He seemed flabergasted that I would ask that. When he called back after consulting the engineers, he said that indeed, if the hypervisor failed, all LPARS would be unavailable.
He told me though that the Hypervisor NEVER failed.
I'll find out down the road when we get our pSeries install finished.
We decided to go 2 LPAR for failover and bought a separate 4 way as a ultimate backup.
Anyone ever seen a hypervisor fail? How ugly is it?