In the 90s it was difficult to imagine Novell ever fading away in the enterprise server market. There was no alternative comparable to Netware and NDS. Microsoft completely killed them off with NT Server and then Active Directory with Win2k. We had around 50 clients running Netware, and within 12 months of Active Directory being released, we had switched almsot all of them over to Windows (usually with Linux servers alongside for web stuff).
Some customers were so desperate to move over that we actually installed and rolled out preview releases of Win2k - since it allowed them to manage their network using the GUI and it was a total cost reduction.
Novell simply never responded. They were still stuck on DOS when Win2k provided a familiar GUI and much easier management. In our bespoke rollouts, even we recognized that Win2k wasn't the best at everything, which is why we would almost install some Linux or BSD servers as well (with Samba) (we did so much of these hybrid installs that I ended up writing NDS drivers for Squid, Apache and a bunch of other open source apps - I think that same code is still being used today). Running Windows also meant that you had one unifying network protocol - TCP/IP, instead of IPX/SPX on the network and TCP/IP on the outside.
RFP's from government and large corporations flooded out with requests for Windows rollouts to replace Netware. It was amazing how quickly this happen. A lot of consultants made a lot of bank at that time ripping Novell out of the enterprise and replacing it with Windows, and often Linux as well.
By the time Novell picked up SuSE, it was too late - they had blown the massive lead they had in the enterprise server market and never regained it. Sorta sad what happen to them, since Netware was a great product for its time.
Enterprise Windows, and to an extent, enterprise Linux, owe a lot to the poor strategic decisions Novell made (or didn't make)
I hope Ximian find a good home, there is no mention if they (and Mono etc.) are part of the Linux deal. I wouldn't be surprised if the second buyer is IBM, but Microsoft have also poured a lot of money into Novell the past 5 years, so it could be them.
Hmmm ... I had no difficultly imagining that Netware would die a hard death in the early 90s. My memory of the exact details are a bit fuzzy at this remove (I might be e.g. confounding versions 2 and 3), but its fatal flaw was no memory protection. And as I recall, at least early on, the entire directory structure was in memory (this was an issue for my company in provisioning them for document imaging storage).
So if you wanted to run any 3rd party software on your Netware servers you were making a big bet on both its stability and safety (in not scribbling over the rest of Netware, including most especially whatever memory it was using to keep track of files and directories, no matter whether it had all of the latter in memory or not).
As soon as NT became stable enough (which didn't take very long at all as these things go, that group really had its act together from launch to 3.51 SP1) the value proposition was pretty clear and simple: run Netware for file and print services or run NT for both plus whatever other server software you might want (NDS was never a big issue for the sorts of systems I was involved with so I don't know the story there).
So as I recall the bleeding started before Win2K, although I can see becoming critical with the release of Active Directory.
Of course, all this assumed that Novell wouldn't make the proper strategic decisions in time, but it didn't take too long to realize they just didn't have what it took in that area.
> My memory of the exact details are a bit fuzzy at this remove (I might be e.g. confounding versions 2 and 3), but its fatal flaw was no memory protection.
You are correct, Netware 3 ran everything in Ring 0, which also made it so fast. Memory protection was added in Netware 4, but you can still opt for running applications in non-protected address space all the way up to Netware 6.5 (latest, and almost certainly the last version).
Wikipedia claims "In 4.x and earlier versions, NetWare did not support preemption, virtual memory, graphical user interfaces, etc. Processes and services running under the NetWare OS were expected to be cooperative, that is to process a request and return control to the OS in a timely fashion. On the down side, this trust of application processes to manage themselves could lead to a misbehaving application bringing down the server."
If true, memory protection without preemptive multitasking is still basically unacceptable (just not something you can base a healthy server ecosystem on; we had enough "fun" of that sort with non-preemptive 16 bit Windows client).
The article also notes that TCP/IP was a second class citizen at best until Netware 5.x in October 1998, "during a time when NetWare market share dropped precipitously...."
In general it notes Netware's kernel had a big speed advantage which became less important as systems steadily became faster (and I suspect its tunable caching became less useful).
So what could be VMware's plans for SuSE Linux? I am not familiar enough with their strategy to be able to speculate on this question, but I am curious - it certainly does not look like an "obvious" acquisition to me.
Keep in mind that SuSE Linux was not profitable under Novell - IIRC the SuSE venture finally broke even in the last FY, so it's not like VMware are buying SuSE to rake in the Enterprise Linux revenue.
VMware's newer products for enterprise virtualization (VSphere, 4i) are built on top of the SuSE Linux kernel... so my guess would be a purchase to essentially own the underlying OS kernel they build their product on top of. IIRC, their older products were built using Red Hat instead, they made the switch a couple years ago.
I didn't know that vSphere used a SuSE kernel - that's interesting. I assumed they roll their own for sure.
There must be more to it though - a couple of hundred million dollars for a Linux kernel which you can have for free anyway would be a steep price, even if we consider it as a talent acquisition to get the engineering team which built that kernel.
When I was in one of the vSphere training classes the instructor informed us that the hypervisor is completely homegrown and that only the Management Console (which is a VM that provides CLI access) runs on top of a Linux kernel.
If you watch ESX boot, the Linux kernel starts first - the 'VMkernel' is started via the vmkmod.ko module.
Not saying that VMware Inc are wrong, but Linux runs VMkernel. VMkernel may them 'subsume' its parent kernel somehow but I'm not sure how they'd accomplish this, nor do VMware like talking about how this is achieved.
This is my first comment here. Oh Boy! how easy and fast it was to sign up!!
I sign just to comment because the comment "(VSphere, 4i) are built on top of the SuSE Linux kernel" is so wrong. The Classic ESX has Service Console is is customized RHEL. ESXi does not have the Service Console. It still does have a command console where one can run ESX commands and a limited set of Unix commands which AFAIK are Busybox's. It also has dropbear sshd.
So coming back to the topic, I too have no idea what VMware is going to do with SuSe Linux.
"ESX 3.5 was built on top of RHEL" again far from accurate. The COS has been based on RHEL since the genesis of ESX. Whereas VMkernel is VMware's own creation. At least thats what they claim.
Even if the COS in ESX 4.X is based on SLES instead of RHEL, your point does not still make sense because VMware's future roadmap is to phase out Classic ESX. Sometime in the future, there will be only ESXi.
Maybe they are planning on entering the cloud hosting market with an OS, with some nice SaS included, that can either be hosted on their cloud or an internally hosted cloud? Their purchases of Zimbra, SpringSource, Gemstone, Dunes, TriCipher, Itegrien and hiring of the main Redis developer seem to indicate they are concocting something interesting.
The acquisition of SpringSource was the the first step towards this. They catered towards the Java/Spring/hibernate enterprise developers. The SuSE branding and kernel will be used to bolster cloud development (internal & external clouds) for the LAMP crowd.
I don't think it does - it is just the SuSE venture that is being sold to VMware. I would imagine that the copyrights are held by the "rest of Novell".
Vmware is one of the few software I have ever bought, I am pretty hardcore open source otherwise.
Vmware works very well, has top notch clone/backup/rollback feature, can automatically install tons of different OS. That's a real time saver.
I use kvm at work where we don't have vmware licenses, and while it does work nicely, there is still no working rollback that I know of, and the network configuration is a PITA in general. Vmware frees me from all that hassle that I frankly don't want to deal with.
Have you ever seen a crashed ESX Server with no community support, tons of home-brew utilities with wired command line arguments which throw a bizarre error messages to you that even google know nothing about? I did. ^_^
The idea is about reducing complexity and reusing community-tested solutions, not creating a new layer of complexity with a coded in a rush, badly tested poorly documented closed-source layer.
btw, if I use native lvm I can get a lot of flexibility.
edit: some BS removed.
btw, there is still no x86_64 xen-host for a modern kernels.
I feel sorry for your experience, it just does not match mine. Note that we are not talking about the same products - I mainly use vmware workstation/player/fusion (I deal a lot with cross platform issues in some medium-size open source projects).
Concerning things coded in a rush, I cannot say about vmware, but libvirt and the associated tools certainly are far from mature in my experience, at least for an end user.
I had once a system which runs 64-bit FreeBSD, RHEL and Solaris on one Core-Quad box with Fedora 13. It used for testing mostly while FreeBSD instance was used as a Windows VPN server (mpd5) and OpenVPN.
Btw, there is an issues with last updates of qemu-kvm in Fedora 13. beware.
ESX Server does not have community support, but it has commercial support.
If you are tenacious and patient enough to make it past the first tier of support, you will get your crashed server fixed, possibly faster than it would take to wait for an answer on a mailing list.
Oh come on. ESX isn't the shining example you make it out to be, true KVM doesn't have as nice snapshots as ESX, but ESX has snapshots that are implemented very poorly - a never-ending delta that grows and grows and grows and takes performance with it eventually too. How many times have I lose a VM while cleaning up snapshots? Not enough to count on more than one hand, but that's still too many times.
I'm using openSuse 10.1. I got 11.3 but it was awful. I don't know how they messed up. I went back to 10.1.
Either way, Ubuntu didn't find my video card. but Suse did (it's a 1920x1200 Dell laptop). I don't know why Ubuntu didn't.
Suse is easy to configure and it's amazing to think how much a Sun box would have cost that did what this can do (my Dell M70 cost just £200 second-hand).
There is a massive truckload of .net developers who create iPad and iPhone apps using Mono (particularly via Unity3D) and even more who would if Mono had better promotion.
They'd also be able to get .net into the browser as a standard.
36 comments
[ 4.5 ms ] story [ 84.4 ms ] threadSome customers were so desperate to move over that we actually installed and rolled out preview releases of Win2k - since it allowed them to manage their network using the GUI and it was a total cost reduction.
Novell simply never responded. They were still stuck on DOS when Win2k provided a familiar GUI and much easier management. In our bespoke rollouts, even we recognized that Win2k wasn't the best at everything, which is why we would almost install some Linux or BSD servers as well (with Samba) (we did so much of these hybrid installs that I ended up writing NDS drivers for Squid, Apache and a bunch of other open source apps - I think that same code is still being used today). Running Windows also meant that you had one unifying network protocol - TCP/IP, instead of IPX/SPX on the network and TCP/IP on the outside.
RFP's from government and large corporations flooded out with requests for Windows rollouts to replace Netware. It was amazing how quickly this happen. A lot of consultants made a lot of bank at that time ripping Novell out of the enterprise and replacing it with Windows, and often Linux as well.
By the time Novell picked up SuSE, it was too late - they had blown the massive lead they had in the enterprise server market and never regained it. Sorta sad what happen to them, since Netware was a great product for its time.
Enterprise Windows, and to an extent, enterprise Linux, owe a lot to the poor strategic decisions Novell made (or didn't make)
I hope Ximian find a good home, there is no mention if they (and Mono etc.) are part of the Linux deal. I wouldn't be surprised if the second buyer is IBM, but Microsoft have also poured a lot of money into Novell the past 5 years, so it could be them.
So if you wanted to run any 3rd party software on your Netware servers you were making a big bet on both its stability and safety (in not scribbling over the rest of Netware, including most especially whatever memory it was using to keep track of files and directories, no matter whether it had all of the latter in memory or not).
As soon as NT became stable enough (which didn't take very long at all as these things go, that group really had its act together from launch to 3.51 SP1) the value proposition was pretty clear and simple: run Netware for file and print services or run NT for both plus whatever other server software you might want (NDS was never a big issue for the sorts of systems I was involved with so I don't know the story there).
So as I recall the bleeding started before Win2K, although I can see becoming critical with the release of Active Directory.
Of course, all this assumed that Novell wouldn't make the proper strategic decisions in time, but it didn't take too long to realize they just didn't have what it took in that area.
You are correct, Netware 3 ran everything in Ring 0, which also made it so fast. Memory protection was added in Netware 4, but you can still opt for running applications in non-protected address space all the way up to Netware 6.5 (latest, and almost certainly the last version).
Wikipedia claims "In 4.x and earlier versions, NetWare did not support preemption, virtual memory, graphical user interfaces, etc. Processes and services running under the NetWare OS were expected to be cooperative, that is to process a request and return control to the OS in a timely fashion. On the down side, this trust of application processes to manage themselves could lead to a misbehaving application bringing down the server."
If true, memory protection without preemptive multitasking is still basically unacceptable (just not something you can base a healthy server ecosystem on; we had enough "fun" of that sort with non-preemptive 16 bit Windows client).
The article also notes that TCP/IP was a second class citizen at best until Netware 5.x in October 1998, "during a time when NetWare market share dropped precipitously...."
In general it notes Netware's kernel had a big speed advantage which became less important as systems steadily became faster (and I suspect its tunable caching became less useful).
Keep in mind that SuSE Linux was not profitable under Novell - IIRC the SuSE venture finally broke even in the last FY, so it's not like VMware are buying SuSE to rake in the Enterprise Linux revenue.
There must be more to it though - a couple of hundred million dollars for a Linux kernel which you can have for free anyway would be a steep price, even if we consider it as a talent acquisition to get the engineering team which built that kernel.
Not saying that VMware Inc are wrong, but Linux runs VMkernel. VMkernel may them 'subsume' its parent kernel somehow but I'm not sure how they'd accomplish this, nor do VMware like talking about how this is achieved.
I sign just to comment because the comment "(VSphere, 4i) are built on top of the SuSE Linux kernel" is so wrong. The Classic ESX has Service Console is is customized RHEL. ESXi does not have the Service Console. It still does have a command console where one can run ESX commands and a limited set of Unix commands which AFAIK are Busybox's. It also has dropbear sshd.
So coming back to the topic, I too have no idea what VMware is going to do with SuSe Linux.
The Classic ESX has Service Console is is customized RHEL
ESX 3.5 was built on top of RHEL, but vSphere 4 was completely redone. Ask anyone at VMware my friend.
"ESX 3.5 was built on top of RHEL" again far from accurate. The COS has been based on RHEL since the genesis of ESX. Whereas VMkernel is VMware's own creation. At least thats what they claim.
Even if the COS in ESX 4.X is based on SLES instead of RHEL, your point does not still make sense because VMware's future roadmap is to phase out Classic ESX. Sometime in the future, there will be only ESXi.
The acquisition of SpringSource was the the first step towards this. They catered towards the Java/Spring/hibernate enterprise developers. The SuSE branding and kernel will be used to bolster cloud development (internal & external clouds) for the LAMP crowd.
Vmware works very well, has top notch clone/backup/rollback feature, can automatically install tons of different OS. That's a real time saver.
I use kvm at work where we don't have vmware licenses, and while it does work nicely, there is still no working rollback that I know of, and the network configuration is a PITA in general. Vmware frees me from all that hassle that I frankly don't want to deal with.
The idea is about reducing complexity and reusing community-tested solutions, not creating a new layer of complexity with a coded in a rush, badly tested poorly documented closed-source layer.
btw, if I use native lvm I can get a lot of flexibility.
edit: some BS removed.
btw, there is still no x86_64 xen-host for a modern kernels.
Concerning things coded in a rush, I cannot say about vmware, but libvirt and the associated tools certainly are far from mature in my experience, at least for an end user.
Btw, there is an issues with last updates of qemu-kvm in Fedora 13. beware.
If you are tenacious and patient enough to make it past the first tier of support, you will get your crashed server fixed, possibly faster than it would take to wait for an answer on a mailing list.
Either way, Ubuntu didn't find my video card. but Suse did (it's a 1920x1200 Dell laptop). I don't know why Ubuntu didn't.
Suse is easy to configure and it's amazing to think how much a Sun box would have cost that did what this can do (my Dell M70 cost just £200 second-hand).
There is a massive truckload of .net developers who create iPad and iPhone apps using Mono (particularly via Unity3D) and even more who would if Mono had better promotion.
They'd also be able to get .net into the browser as a standard.