I spent a considerable amount of time trying to figure out where you were
going with that, and eventually just decided to run with it anyway.
But yeah, it seems as if HP is deliberately self-destructing. They're the #1 PC maker with 20% of the market, and they're getting out of that business.
Shutting down the WebOS unit I can understand; no one seems to want those devices anymore. Spinning off the PC business seems as if they like what happened to Sun and they're trying to get to the same place as quickly as possible.
But yeah, it seems as if HP is deliberately self-destructing. They're the #1 PC maker with 20% of the market, and they're getting out of that business.
Shutting down the WebOS unit I can understand; no one seems to want those devices anymore. Spinning off the PC business seems as if they like what happened to Sun and they're trying to get to the same place as quickly as possible.
yea, seems as if they should get carly fiorina back and continue manufacturing printer ink.
as for webos... one would think that a company like HP could do better in buying a promising technology and developing it to some point where everybody wants it...
as the register comments, they might get a chance to license it, as maybe androids flag sinks with the motorola merger...
but they currently seem to be rather able to fork it instead of buying a license of another os...
I guess they're following their aquired companies compaq & dec into the hoods of insignificance.
I guess one could argue that intel managed to successfully evaporate all better alternatives to unix workstation & server processors with its itanic, and eventualy take them all out of business...
- alpha / dec dead.
- pa/risc HP; dead.
- mips well sort of dead man walking, trying to become the next ARM since a decade now... promising fork of the technology in china
- powerpc Ok, we've got IBM, who didn't buy on itanic in first place,
- Sparc who might survive within sun/oracle
now there is no ex-intel / microsoft employee to blame for their decision to move "into the cloud" like over at nokia, though the stories seem somewhat similar...
Aug 19 2011 5:00pm from Ford II @uncnsrd
I just assumed you were making your usual non-sense. :-)
All I can say is it definitely made sense to me at the time, and now it totally doesn't ;)
dothebart: in all fairness, the successor to PA-RISC is IA-64 (Itanium).
It isn't dead -- it simply failed to take its intended place as the successor
to x86. At this point it must simply be considered HP's entry in the high
end proprietary systems world; its brethren are SPARC and POWER.
Whether that's "enough" to keep it afloat is another story altogether. There are some workloads that will seemingly never migrate to the volume architecture (x86-64). I don't really have an opinion as to whether that's a good thing or a bad thing.
So it seems HP wants to become another IBM or Oracle. I think the next couple of years are going to serve as a painful reminder of how much they relied on the businesses they're dissolving or spinning off. A "focus on software" would seem to appear easy if HP is looking at companies like Oracle and Microsoft and IBM and saying "we're big like them; we can do that." It isn't that easy though. They could jump start it by acquiring SAP, but even with that head start, they'd need to have top management with flawless execution -- something that HP hasn't had much of lately.
Whether that's "enough" to keep it afloat is another story altogether. There are some workloads that will seemingly never migrate to the volume architecture (x86-64). I don't really have an opinion as to whether that's a good thing or a bad thing.
So it seems HP wants to become another IBM or Oracle. I think the next couple of years are going to serve as a painful reminder of how much they relied on the businesses they're dissolving or spinning off. A "focus on software" would seem to appear easy if HP is looking at companies like Oracle and Microsoft and IBM and saying "we're big like them; we can do that." It isn't that easy though. They could jump start it by acquiring SAP, but even with that head start, they'd need to have top management with flawless execution -- something that HP hasn't had much of lately.
(Itanium). It isn't dead -- it simply failed to take its intended place
Right, it's resting.
start it by acquiring SAP, but even with that head start, they'd need
to have top management with flawless execution -- something that HP
hasn't had much of lately.
Hey, they've always got printers. Paperless orifice be damned. PC load letter?
Perhaps I might add that DEC went out of business!
(Itanium). It isn't dead -- it simply failed to take its intendedplace
Right, it's resting.
When I have time, I'm going to port OS/2 to the itanium, then you guys will know true beauty.
True beauty was SPARC, during the era when Sun was a company run by engineers.
"We had joy, we had fun, we had Netscape on the Sun..."
(Even in its very first generation, x86 assembler was *never* elegant.)
"We had joy, we had fun, we had Netscape on the Sun..."
(Even in its very first generation, x86 assembler was *never* elegant.)
Perhaps I might add that DEC went out of business!
Zing! Nailed it this time :)
Perhaps that's the answer. All of the technology of three different companis, and HP goes strategic with the ones that suck the most. So it's obvious that this is a race to the bottom for them.
I wonder if there's a technology in there somewhere that's just an albatross around the neck of whatever company owns it. If that's the case, they should just go out and buy whatever's left of WordPerfect so they have the complete stack.
Divesting the unprofitable parts of a large, corporate conglomerate is probably a lot like divesting the bad assets of an involvent bank. You could split it into a "bad bank" and a "good bank", give the bad bank all the bad assets and a disproportionate share of the liabilities... but that's equivalent to defaulting on your debts, and the investors will sue.
No doubt they've got a few albatrosses in there somewhere... I'm guessing that thosse are going to have to be handled as asset sales if they can find a willing buyer who can work with the smaller margins of the PC business (like Lenovo was for IBM)
True beauty was SPARC, during the era when Sun was a company run by
Too bad about that crazy rotating register file which requires all those interconnects on the chip... maybe a nice architecture for the assembly programmer (who?) but a tough one to implement in silicon with decent performance.
I guess this is as good a place as any to chronicle the rebuild of my server
here at the Big Blue X, which is host to the various virtual machines I run
... including a site called uncensored.citadel.org (you're soaking in it).
Having been toasted by software RAID failures twice now, I'm simply not going to run RAID at all this time. And I don't have the money to spend on a hardware RAID controller.
I've been wanting to switch my server from CentOS to ProxMox VE [http://proxmox.org/products/proxmox-ve] which is, quite simply, the best way to deploy open source virtualization, hands down. I've touted the merits of this distribution before. It's based on a Debian build with the kernel optimized for KVM and OpenVZ deployments, and it's got a really nice web interface. Supports clustering, live migration, shared storage ... the works.
The question for me was how to begin taking full backups of the VM images without shutting them down. I settled on a practice that is fairly common among system administrators once you begin to think of the virtual machines as "database-like" applications. In my new deployment, the virtual disks (.qcow2 files) reside in a filesystem that lives on an LVM2 logical volume. I have written a script that will perform a snapshot of the logical volume, mount the snapshot read-only, rsync it to the backup destination, then unmount and destroy the snapshot.
After a couple of test runs I can confirm that I still *do* get the benefit of the rsync speedup, even though we are copying virtual machine images instead of files within the guest OS.
Once things settle in a bit I'll probably have it sync to both on-site and off-site backups.
Having been toasted by software RAID failures twice now, I'm simply not going to run RAID at all this time. And I don't have the money to spend on a hardware RAID controller.
I've been wanting to switch my server from CentOS to ProxMox VE [http://proxmox.org/products/proxmox-ve] which is, quite simply, the best way to deploy open source virtualization, hands down. I've touted the merits of this distribution before. It's based on a Debian build with the kernel optimized for KVM and OpenVZ deployments, and it's got a really nice web interface. Supports clustering, live migration, shared storage ... the works.
The question for me was how to begin taking full backups of the VM images without shutting them down. I settled on a practice that is fairly common among system administrators once you begin to think of the virtual machines as "database-like" applications. In my new deployment, the virtual disks (.qcow2 files) reside in a filesystem that lives on an LVM2 logical volume. I have written a script that will perform a snapshot of the logical volume, mount the snapshot read-only, rsync it to the backup destination, then unmount and destroy the snapshot.
After a couple of test runs I can confirm that I still *do* get the benefit of the rsync speedup, even though we are copying virtual machine images instead of files within the guest OS.
Once things settle in a bit I'll probably have it sync to both on-site and off-site backups.
Sounds fishy. I think you may not get a time-consistent snapshot of the guest filesystems that you're backing up unless you perform xfs_freeze (or its equivalent under ext4 4, write_super_{un,}lockfs) in the *guest*
Wed Aug 31 2011 16:22:32 EDT from IGnatius T Foobar @ UncensoredHaving been toasted by software RAID failures twice now, I'm simply not going to run RAID at all this time. And I don't have the money to spend on a hardware RAID controller.
A hardware RAID can also fail, and take data with it.
Yeah, I experienced the hardware RAID failure problem at work, shortly after I started on this job.
We had replaced a drive, then Microsoft decided to restart the operating system (against our requests) while the hardware was busy updating the new drive. It cost us over $15,000 to recover the data lost by that event.
Well, if it's a true snapshot, the guest images will certainly not be cleanly
unmounted, but they ought to be time-consistent, because LVM snapshots are
COW-based...
Theoretically, a restored image ought to have the same level of consistency as a non-virtual server that was powered down during normal operation.
Obviously there will have to be quite a lot of verification of these backups, at least initially. I will probably also continue to do my rsync backups, perhaps to another destination. I may consider doing the VM image backups onsite and rsync backups offsite, or something like that...
Theoretically, a restored image ought to have the same level of consistency as a non-virtual server that was powered down during normal operation.
Obviously there will have to be quite a lot of verification of these backups, at least initially. I will probably also continue to do my rsync backups, perhaps to another destination. I may consider doing the VM image backups onsite and rsync backups offsite, or something like that...
ouch... 15 grand just because of microsoft... wait... that comment should be in the Microsoft Bashing room.
I've never worked with RAID setups before, but know how either side (soft or hard) works... I just use .img/.iso backups. I have it set to freeze the system and write to an .img every couple of days. I used to use the automated backup to .tar.gz approach but felt it necessary (and slightly easier) to do disk images...
--
Stephen D King
skpacman8629@gmail.com
We had replaced a drive, then Microsoft decided to restart the
operating system (against our requests) while the hardware was busy
updating the new drive. It cost us over $15,000 to recover the data
lost by that event.
We frequently have to remind people that RAID is *not a backup strategy*.
Even with the best RAID you still have to do off-host backups.
RAID is a technology intended for high availability, not for backups.
As for hardware vs. software, it doesn't really matter how much money you've sunk into it; if you have two disk failures in a RAID5 set, your volume is going bye-bye. If the application is mission critical, then hopefully you've invested in good storage with things like predictive failure detection, automatic and immediate rebuilding of failed disks onto hot spares, etc. High end storage systems will even "call home" so that the vendor can send a tech with replacement parts to you right away.
And of course you could always do RAID6 or even RAID60, and really be able to lose two disks and still keep going. Good luck getting that past your DBA, though. "The Book" says that anything other than RAID1 or RAID10 is Teh Evil And Must Never Be Implemented, so your paper DBA will usually refuse to actually analyze the application's performance requirements vs. the hardware's capabilities. DBA will instead see that you want to implement RAID5/50/6/60 and say "This is unacceptable we require RAID10 please do the needful"