We've been using Oracle VM in production for years with very few issues. Their management software is total garbage, but it's still xen underneath which has been very solid for us. We've used xen for our virtualization of non-production servers well before Oracle came out with their product, so I use the command line to manage vm's more than that horrible web app. When Oracle offered support their database running on what amounts to xen with Oracle branding, we moved over and haven't looked back.
You put your virtualization on the new machines not on the hand-me-down stuff, silly. Your old machines weren't speced for them.
No, technically they weren't. But the specs fit the need.
PS: This is one reason businesses lease servers instead of buy them. It makes it easy to cycle out the old junk every few years.
Very true. But the cost of buying out the lease of existing servers to use as dom0's was more cost effective than leasing new servers that would be overkill for the entire length of the lease. It's cost justification. Next leasing cycle, specs need to be re-evaluated.
If the CPU doesn't have hardware VM built on it, it isn't an ideal candidate for serving virtual machines period.
Wrong. The CPU's still work great with paravirtualization for creating linux vm's. Just because everyone tends to ignore xen's ability to paravirtualize, doesn't mean that it doesn't exist. It works very well and is very ideal for some needs.
Why would I want to spend money on buying new servers when the ones I own already work fine? Just because the next servers I buy will probably have vt-x, doesn't change the fact that the ones I currently own don't.
Using the argument that people just need to upgrade their hardware is garbage. Virtualization is supposed to CUT costs, not incur new hardware costs. The fact that I could take old machines that have run their course as database servers and convert them into several virtual machines for more lightweight stuff (web,dns,ldap,etc.) is what drew me into virtualization in the first place.
Re:Isn't everyone like just using KVM?
on
The Book of Xen
·
· Score: 2, Informative
No, those of use with older servers (without vmx/svm processors) are not. The machines work just fine and XEN allows us to paravirtualize several linux machines on each one. Unfortunately RedHat has lost sight of those customers. As far as I'm aware KVM still requires the intel vt/amd-v technologies to create any guests, which XEN only requires for HVM guests (i.e. windows).
Re:Xen slowly being discarded ?
on
Fedora 10 Released
·
· Score: 3, Informative
Xen virtualization is still strong, fedora just isn't porting the kernel patches upstream anymore. I believe Novell has a 2.6.27 kernel with xen patches. I would think it to be possible to pull down the kernel source from Fedora, build the config, then pull in suse kernel source, run make oldconfig, and compile your own kernel for fedora using the suse sources. I've never tried it and fully understand that this is an unacceptable option for most fedora users.
I've been following the fedora-xen mailing list and they would still like to put xen back in, but not until it's in the upstream kernel. As stated at the F9 release, they feel it is counterproductive to maintain 2 different kernels, which I can't disagree with them on. The latest news I heard is that they were hoping the 3.4 release of xen to have pv_ops dom0 in the kernel. Wether or not that it's Linus' kernel, they haven't really stated. If fedora is waiting for pv_ops dom0 to make it into Linus' kernel, then I would have to agree that xen will slowly dissapear and KVM will be the way to go. I see that as an uphill battle for the xen team since kvm is already in the kernel.
http://fedoraproject.org/wiki/Docs/Beats/Virtualization for more information on f10 virtualization.
puercopop@localhost ~ $ which which
which: no which in (/usr/bin:/usr/local/bin:/usr/bin:/bin:/opt/bin:/usr/x86_64-pc-linux-gnu/gcc-bin/4.1.2:/usr/qt/3/bin:/usr/games/bin)
It may not be bad for consumers in general, but for Oracle customers it limits choice. While virtualizing a database may not be the best idea from a performance standpoint, virtualizing a middle tier or 2 may be very attractive.
Oracle is only supporting virtualization with their VM product. While this makes sense from their perspective, consumers are now choiceless if they use Oracle. For instance, we've implemented open source Xen to consilidate several other servers, if we planned to virtualize any of our Oracle servers at any point we'd have to use Oracle's VM to do so, making management of VM solutions more complicated.
This also means Oracle doesn't *have* to compete in performance if they don't want to. "Use us or don't use virtualization at all with our products".
This is the exact fear I have with their "Unbreakable Linux". Thankfully they came to the game too late to pull that card with linux... but then again this is Ellison we're talking about and he could change his mind at any second.
Actually Oracle's VM is much more akin to the xenserver product from citrix or ESX from vmware. Its not a hypervisor that runs on an existing OS, it is a stand alone OS. Oracle's VM is not a rebrand of RHEL this time.
And btw, I'm pretty sure oracle's "unbreakable linux" (the red hat clone) already has Xen as a virtualization option, as it is based on RHEL 5 which includes Xen (even in the 5.0 release).
We've upgraded all our 5 servers to 5.1 in the past 2 days, all of which are Xen dom0. The live migration of fully virtualized (hvm) guests is now supported and works swimmingly well. There is 0 downtime, only a small hiccup in the network connection, which is not noticeable unless you are watching for it. We've transferred mid-download on the domU and have not dropped a packet. The only issue we've really had is having to re-setup the NIC cards of HVM guests after the upgrade. They apparently see a different (better?) piece of hardware for the virutalized network card. Live migration of paravirtualized guests has always worked well and continues to do so.
ACPI is now supported in windows guests, which is a big bonus for us. 32-bit paravirtualized guest also work on 64-bit dom0's. This is only a "technology preview" but so far has worked pretty good (for the day and half we've had a system running on it). However live migration from a 64-bit host to a 32-bit host (and vice versa) does not appear to work. I've not delved into it enough to find the problem though.
We just started setting up xen for our webservers this week.
We have plans to move our legacy cobol applications to a vmserver in the future. Those applications have a lot of disk i/o. We've found that NFS works nicely in xen, so our plan was to put the files that need editing on a fast network share.
Would doing this still cause an i/o bottleneck in the vm? My assumption was that only 'local' i/o would be an issue.
Does anyone have experience in that area?
Granted, I'm not a microsoft admin, but some of our 3rd party apps still run on server 2003. I understand that some companies may not have had patches out until recently, but we started back in January with our Oracle and Java patches. Our approach was to bite of small junks as early and often as possible. We also pushed back on any 3rd party applications that we pay support to, to get DST patches as quickly as possible. We've had most patches (all mission critical ones) in place on test systems since early February, giving our users ample time to test.
The past 2 weeks have been relaxing AFAIC compared to the previous 4-6.
You're using Internet Explorer, aren't you? Turn in your geek card, you're not welcome here.:P
Canvascape uses Google's IE Adapter to provide Canvas support. It's functional, but not particularly quick. If you try it in Firefox/Seamonkey or Opera, I think you'll find performance to be at least an order of magnitude better. There's no frame counter, but I'd estimate that I get about 30 frames per second. Which isn't bad for a software implementation of a 3D game, programmed in Javascript!
While it works 'ok' on opera for me, the Firefox implementation was extremely slow. I'm running linux so I can't compare it to the internet explorer version.
Still, i agree, it is impressive that is programmed in javascript.
I disagree.
I knew several people with PS2's when they came out, and not a single person even used the DVD player, let alone decided to purchase a PS2 because of it. They all loved the original Playstation and/or liked the games that were out for PS2. None of them said "Oh there's a cheap dvd player! and bonus, it also plays a few games!"
I doubt very much that many people bought a PS2 just because it had a dvd player built in, just like i highly doubt many people will buy a PS3 just because it has a blu-ray player built in.
The PS2 was successful because it had exclusive games that people wanted, a control format that people were used to, and a price that was reasonable for most people.... plus they could still play their PS1 games that they invested money into.
When I read this, I had a very similar thought.
Ellison and the MS camp certainly aren't friendly... with Oracle sudo-owning a linux distro, perhaps Microsoft felt they needed to compete?
Oracle 'steals' the biggest market share of enterprise linux, and two weeks later MS goes out and 'partners' with the 2nd most popular version enterprise linux...
Seems too odd to be coincedence in my mind.
I just got my first ipod last week, its a 30G ipod video. I won it as a prize. I have a PDA which i've always felt sufficient for listening to music and watching movies, although limited in storage space. I figured i'd give the ipod a shot because it has a longer battery life and a lot more storage. I don't know if its because i'm coming from a touch screen interface to the clickwheel, but i didn't like the clickwheel. Although i think the design to be 1 handed is nice, i constantly missed my mark and 'bumped' it inadvertently. I would've much preferred depressing the buttons to navigate around. I'm looking forward to the new interface.
Working on an OpenLaszlo http://www.openlaszlo.org/ project, which occasionally caused the flash player to timeout while testing. Its nice that i can now "continue running the script" instead of being forced to close out, fix one problem, come back in only to find another problem. Flash 7 on fedora didn't allow me to continue. Unfortunately flash 9 doesn't feel any faster in the application than flash 7 did.
Slashdot Mirror
We've been using Oracle VM in production for years with very few issues. Their management software is total garbage, but it's still xen underneath which has been very solid for us. We've used xen for our virtualization of non-production servers well before Oracle came out with their product, so I use the command line to manage vm's more than that horrible web app. When Oracle offered support their database running on what amounts to xen with Oracle branding, we moved over and haven't looked back.
You put your virtualization on the new machines not on the hand-me-down stuff, silly. Your old machines weren't speced for them.
No, technically they weren't. But the specs fit the need.
PS: This is one reason businesses lease servers instead of buy them. It makes it easy to cycle out the old junk every few years.
Very true. But the cost of buying out the lease of existing servers to use as dom0's was more cost effective than leasing new servers that would be overkill for the entire length of the lease. It's cost justification. Next leasing cycle, specs need to be re-evaluated.
If the CPU doesn't have hardware VM built on it, it isn't an ideal candidate for serving virtual machines period.
Wrong. The CPU's still work great with paravirtualization for creating linux vm's. Just because everyone tends to ignore xen's ability to paravirtualize, doesn't mean that it doesn't exist. It works very well and is very ideal for some needs.
Why would I want to spend money on buying new servers when the ones I own already work fine? Just because the next servers I buy will probably have vt-x, doesn't change the fact that the ones I currently own don't.
Using the argument that people just need to upgrade their hardware is garbage. Virtualization is supposed to CUT costs, not incur new hardware costs. The fact that I could take old machines that have run their course as database servers and convert them into several virtual machines for more lightweight stuff (web,dns,ldap,etc.) is what drew me into virtualization in the first place.
No, those of use with older servers (without vmx/svm processors) are not. The machines work just fine and XEN allows us to paravirtualize several linux machines on each one. Unfortunately RedHat has lost sight of those customers. As far as I'm aware KVM still requires the intel vt/amd-v technologies to create any guests, which XEN only requires for HVM guests (i.e. windows).
He used to be a hilarious DJ on WXDX in Pittsburgh that I listened to in college.
http://en.wikipedia.org/wiki/Alan_Cox_(radio_presenter)
oh wait...
Xen virtualization is still strong, fedora just isn't porting the kernel patches upstream anymore. I believe Novell has a 2.6.27 kernel with xen patches. I would think it to be possible to pull down the kernel source from Fedora, build the config, then pull in suse kernel source, run make oldconfig, and compile your own kernel for fedora using the suse sources. I've never tried it and fully understand that this is an unacceptable option for most fedora users.
I've been following the fedora-xen mailing list and they would still like to put xen back in, but not until it's in the upstream kernel. As stated at the F9 release, they feel it is counterproductive to maintain 2 different kernels, which I can't disagree with them on. The latest news I heard is that they were hoping the 3.4 release of xen to have pv_ops dom0 in the kernel. Wether or not that it's Linus' kernel, they haven't really stated. If fedora is waiting for pv_ops dom0 to make it into Linus' kernel, then I would have to agree that xen will slowly dissapear and KVM will be the way to go. I see that as an uphill battle for the xen team since kvm is already in the kernel.
http://fedoraproject.org/wiki/Docs/Beats/Virtualization for more information on f10 virtualization.
puercopop@localhost ~ $ which which
which: no which in (/usr/bin:/usr/local/bin:/usr/bin:/bin:/opt/bin:/usr/x86_64-pc-linux-gnu/gcc-bin/4.1.2:/usr/qt/3/bin:/usr/games/bin)
Burn her anyway!
It may not be bad for consumers in general, but for Oracle customers it limits choice.
While virtualizing a database may not be the best idea from a performance standpoint, virtualizing a middle tier or 2 may be very attractive.
Oracle is only supporting virtualization with their VM product. While this makes sense from their perspective, consumers are now choiceless if they use Oracle. For instance, we've implemented open source Xen to consilidate several other servers, if we planned to virtualize any of our Oracle servers at any point we'd have to use Oracle's VM to do so, making management of VM solutions more complicated.
This also means Oracle doesn't *have* to compete in performance if they don't want to. "Use us or don't use virtualization at all with our products".
This is the exact fear I have with their "Unbreakable Linux". Thankfully they came to the game too late to pull that card with linux... but then again this is Ellison we're talking about and he could change his mind at any second.
Actually Oracle's VM is much more akin to the xenserver product from citrix or ESX from vmware. Its not a hypervisor that runs on an existing OS, it is a stand alone OS. Oracle's VM is not a rebrand of RHEL this time. And btw, I'm pretty sure oracle's "unbreakable linux" (the red hat clone) already has Xen as a virtualization option, as it is based on RHEL 5 which includes Xen (even in the 5.0 release).
We've upgraded all our 5 servers to 5.1 in the past 2 days, all of which are Xen dom0.
The live migration of fully virtualized (hvm) guests is now supported and works swimmingly well. There is 0 downtime, only a small hiccup in the network connection, which is not noticeable unless you are watching for it. We've transferred mid-download on the domU and have not dropped a packet.
The only issue we've really had is having to re-setup the NIC cards of HVM guests after the upgrade. They apparently see a different (better?) piece of hardware for the virutalized network card.
Live migration of paravirtualized guests has always worked well and continues to do so.
ACPI is now supported in windows guests, which is a big bonus for us.
32-bit paravirtualized guest also work on 64-bit dom0's. This is only a "technology preview" but so far has worked pretty good (for the day and half we've had a system running on it). However live migration from a 64-bit host to a 32-bit host (and vice versa) does not appear to work. I've not delved into it enough to find the problem though.
what about a foo?
fo shizzle
of course you could just do select count(1)
bah... my bad, forgot to preview
i meant "tab search terms"
nope, not faster... same number of keystrokes as " search terms" which is what i use
We just started setting up xen for our webservers this week. We have plans to move our legacy cobol applications to a vmserver in the future. Those applications have a lot of disk i/o. We've found that NFS works nicely in xen, so our plan was to put the files that need editing on a fast network share. Would doing this still cause an i/o bottleneck in the vm? My assumption was that only 'local' i/o would be an issue. Does anyone have experience in that area?
2 weeks?
Granted, I'm not a microsoft admin, but some of our 3rd party apps still run on server 2003. I understand that some companies may not have had patches out until recently, but we started back in January with our Oracle and Java patches. Our approach was to bite of small junks as early and often as possible.
We also pushed back on any 3rd party applications that we pay support to, to get DST patches as quickly as possible.
We've had most patches (all mission critical ones) in place on test systems since early February, giving our users ample time to test.
The past 2 weeks have been relaxing AFAIC compared to the previous 4-6.
Canvascape uses Google's IE Adapter to provide Canvas support. It's functional, but not particularly quick. If you try it in Firefox/Seamonkey or Opera, I think you'll find performance to be at least an order of magnitude better. There's no frame counter, but I'd estimate that I get about 30 frames per second. Which isn't bad for a software implementation of a 3D game, programmed in Javascript!
While it works 'ok' on opera for me, the Firefox implementation was extremely slow. I'm running linux so I can't compare it to the internet explorer version.Still, i agree, it is impressive that is programmed in javascript.
got any job openings?
I disagree. I knew several people with PS2's when they came out, and not a single person even used the DVD player, let alone decided to purchase a PS2 because of it. They all loved the original Playstation and/or liked the games that were out for PS2. None of them said "Oh there's a cheap dvd player! and bonus, it also plays a few games!" I doubt very much that many people bought a PS2 just because it had a dvd player built in, just like i highly doubt many people will buy a PS3 just because it has a blu-ray player built in. The PS2 was successful because it had exclusive games that people wanted, a control format that people were used to, and a price that was reasonable for most people.... plus they could still play their PS1 games that they invested money into.
When I read this, I had a very similar thought. Ellison and the MS camp certainly aren't friendly... with Oracle sudo-owning a linux distro, perhaps Microsoft felt they needed to compete? Oracle 'steals' the biggest market share of enterprise linux, and two weeks later MS goes out and 'partners' with the 2nd most popular version enterprise linux... Seems too odd to be coincedence in my mind.
I just got my first ipod last week, its a 30G ipod video. I won it as a prize. I have a PDA which i've always felt sufficient for listening to music and watching movies, although limited in storage space. I figured i'd give the ipod a shot because it has a longer battery life and a lot more storage. I don't know if its because i'm coming from a touch screen interface to the clickwheel, but i didn't like the clickwheel. Although i think the design to be 1 handed is nice, i constantly missed my mark and 'bumped' it inadvertently. I would've much preferred depressing the buttons to navigate around.
I'm looking forward to the new interface.
Working on an OpenLaszlo http://www.openlaszlo.org/ project, which occasionally caused the flash player to timeout while testing. Its nice that i can now "continue running the script" instead of being forced to close out, fix one problem, come back in only to find another problem. Flash 7 on fedora didn't allow me to continue. Unfortunately flash 9 doesn't feel any faster in the application than flash 7 did.