Sorry, but from my professional background this is exactly how it is. Each business stream and product line has to work within strict business plans and P&L. Variations against plan have to go though strict change and approval processes. If you need an extra M€ then you need to get this approved on the basis of a business plan.chungy wrote:...Sun Microsystems has a revenue of ~$14 billion...I can't imagine that money is a real factor in this.
Sorry, but I've got to disagree with this one on two levels. First don't believe the MS achievements set the bounds of possibility. When I last looked at this, the best that MS had released was an active/passive system where one system was read-write and the other read-monitor, but with a heartbeat so that the control could flick if the active node stalls. However, Digital's VMS supported active-active clustering where nodes could symmetrically share file systems and other shared resources. This was robust and worked extremely well. The key thing you need is the OS to support a lean and fast Distributed Lock Manager (DLM) in the heart of its kernel, plus some other Comp Sci techniques that were pretty much hammered out in the 70s and 80s.chungy wrote:...The real factor here is the technical feat...and this just doesn't happen in the real world of computing.
It just doesn't work this way, and in reality FLOSS doesn't rely on individuals like me. The real reason for the success of the FLOSS model is the dominance of MS. The only way that the other vendors can compete is by federating their resources. GPL is primarily a way of doing this bit enabling companies to protect their commercial interests whilst pooling when they have to. They choose to donate N engineer-years of effort to develop FLOSS product because it is ultimately in their commercial interest to do so, and every incorporate company in the US has a legal obligation to maximise the profits of its shareholders. That's just the way of our capitalist world.chungy wrote:... So really, all this talk of Sun providing developers the monetary resources misses the whole picture.
Heh, I had no idea that it was even possible to cluster Windows computers. Guess you learn something new every dayTerryE wrote:Windows clusters might pool resources together to appear as a single filesystem but they all have independent disks at the low-level.
As you said, VBox doesn't need to worry about the DLM, it's the responsibility of clusterware(such as Veritas CVM) running on guest machines. So, for basic disk sharing, it shouldn't be that difficult to implement, and I don't believe this task can take a man-year. Once the basic disk sharing implemented, further enhancement should be put on SCSI3 PGR support emulation.TerryE wrote:Sorry, but from my professional background this is exactly how it is. Each business stream and product line has to work within strict business plans and P&L. Variations against plan have to go though strict change and approval processes. If you need an extra M€ then you need to get this approved on the basis of a business plan.chungy wrote:...Sun Microsystems has a revenue of ~$14 billion...I can't imagine that money is a real factor in this.
Sorry, but I've got to disagree with this one on two levels. First don't believe the MS achievements set the bounds of possibility. When I last looked at this, the best that MS had released was an active/passive system where one system was read-write and the other read-monitor, but with a heartbeat so that the control could flick if the active node stalls. However, Digital's VMS supported active-active clustering where nodes could symmetrically share file systems and other shared resources. This was robust and worked extremely well. The key thing you need is the OS to support a lean and fast Distributed Lock Manager (DLM) in the heart of its kernel, plus some other Comp Sci techniques that were pretty much hammered out in the 70s and 80s.chungy wrote:...The real factor here is the technical feat...and this just doesn't happen in the real world of computing.
My second point is that this whole issue of integrity is actually the business of the guest OSs and not the VMMs. All the VMM needs to do is to properly virtualise whatever the shared I/O channel mechanism to the multiple VMMs and again this is a standard problem.
Notwithstanding this, the devil is in the detail and if this takes a man-year, say $150K built up cost, to realise then someone needs to make a business case for an extra $1M revenue or whatever to cover this.