Ephelyon

Another point there: I trust myself to have "backups of backups" - or, in this case, both on-site D2D backups going back a fair while and nightly off-site replication of the entire system (at a cost of £0) - because I know I do. I don't need trust, because it's managed and delivered in-house, by me. The only thing better than the strongest trust is the lack of need for it.

Just gonna leave that here...

Higher.

I get the impression that those of us who attended the Microsoft event on Tuesday aren't really supposed to be discussing it publically yet. Nevertheless, what I can say is that the answer to the question that was the original purpose of this thread - "Is Microsoft looking to move out of the education market?" - is a resounding "no". If anything, they are very interested in a Microsoft-centric future for IT software and services in UK education.

There is a further level after that: Strategic. It does have a place in some schools (though admittedly not all).

Bless 'em. It's our job to think beyond that. We do - that's our job. It's also our livelihood. And without that, none of the rest happens. ... so why on earth would you move from a situation where you can access that without the Internet to a situation where you can't? Where is the business logic there? Good intent notwithstanding, that bodes ill. It's something we need to sort out. That is a pretty wild prediction. It's often the case that what we thought would be have come about by X Time is in fact surpassed by orders of magnitude... but it's also often the case that it's missed by miles. Where are the flying cars?! They will still exist... you've got to get that cloudy connection from classroom to classroom and from block to block, after all! Valid point - we can only guess what's going to happen there... :/

I hear you - there have been step-changes all the way along the line. But in terms of this one, what we're looking at is not just a step-change in how we model our own infrastructure, but whether it even remains our own infrastructure. That takes things to a new level. We've seen mainframe-based solutions where you had a central setup accessed by dumb terminals... then you had the introduction of "fat clients" that had to do complicated things like run their own OS that talks back to something else... then towards the back-end of the NT era Microsoft announced that they'd reinvented the mainframe concept with NT4 Terminal Server (and how we lolled at that)... then we decided to go for full-fat clients but with increased automation driven by the back-end (like GPO and more latterly GPP) to orchestrate many aspects of how they worked... and around the same time we started looking seriously at server virtualisation, which involved essentially the same setup but using fewer more beefy boxes to achieve mostly the same thing... ... and I see how it's easy to view the cloud as just the next logical progression on that journey. But, really, I don't think it is - there's something different about it. For all the changes that happened over the decades in respect of how you managed your environment, you still actually managed it - yourself. What the cloud means is that you don't do that; someone else does. There is a big difference between saying "we will adapt how we manage our environment" and saying "we will let someone else determine how our environment is managed"... oh, and it's not actually here either, it's somewhere else...

Perhaps viewing this as a software/hardware split is muddying the water a little... It's true that in the pre-virtualisation era it was a case of saying "X server does Y job", so if I now have Z requirement, I need another server. I mean, yes, you could consolidate things that e.g. Microsoft would now call "roles" on a single instance - and Linux boxes have been veritable workhorses when it comes to this kind of thing - but you would still have your DC server, file server, print server, database server, web server, terminal server (if you were lucky) etc... and at some point you would probably also have your "unlucky workhorse" server that did all the other little things. And, yes, that could get expensive... Repurposing those roles into "virtual instances" within your own infrastructure was revolutionary about a decade ago, yes. The idea that you didn't need "a box" to do "a thing" was new, and we had to deal with the idea that a few beefy boxes might replace e.g. a whole blade server cluster... but despite all of that, it was still your own infrastructure. You still designed it. You adjusted (kicking and screaming) to the new dynamic of how things would work... how you would plan for redundancy... how backups would be carried out... how you would distribute load... etc. But I don't think that comparing a move to the cloud to virtualisation really works out. In the latter model, you had to decide how you would repurpose your infrastructure. In the former, you're handing that decision off to someone else. It's not the same thing at all.

So... you don't find that any solution-oriented creativity ever stems from this "messing around" and "titting around"?

Well, I'd say there's room for all of us by all means, but it isn't right for one group to be effectively shown the door if there's no real need for that. If the cloud offers you benefits, then fine, but there are many setups where that's not necessarily the case and the nub of my argument is that, if you've managed your infrastructure strategically over the past few years, you really shouldn't need the cloud (by which I mean a fully cloud-based VM setup, not just lashings of O365 or GApps). Just to be clear, that's not meant to be a sly dig at anyone's competence - it also depends on whether your SLT have actually allowed you to do that properly! I think there's a good analogy to be drawn with all this "enforced academisation" stuff going around at the moment. Put bluntly, it's not hard to tell the difference between ideology and dogma, and I think in both "cloud first" and "academies everywhere", we're seeing a policy being dogmatically enforced without any real consideration of whether it's genuinely better for all those affected. A multitude of alternative setups is good; enforcement of any one in particular is bad. But that's perhaps for another thread...

... because that's the fun stuff? Seriously, if the nitty-gritty side of being a sysadmin is something you would want to "avoid", why did you go into it? Or did you not go into it directly? (This is more of a mini "opinion piece" and not intended to be a judgment per se. I'm asking these questions as much out of genuine curiosity as anything else.) For me, designing an infrastructure (more on the systems side than networks), setting up a server backup/failover scheme, planning/installing/configuring/maintaining the back-end server fleet, deploying and managing the system as a whole... that, that is why I went into IT system administration. I enjoy it and I have my main skillset in it. Why on earth would I want to just outsource all of that to someone else?! Of course it's not true to say that there's nothing left for a sysadmin to be doing if everything is cloud-based, but I do think it rather takes the "heart and soul" out of the job. I would struggle to believe I'm the only person who thinks this way too... :/

True - I guess I'm talking about what George Osborne might call 'normal times'...

I do agree with that in principle, but the reason I say both on this occasion is that I don't think it's a good idea to move to the cloud if you find that it is indeed cheaper (which it will be for some) but doesn't actually add to what you can already do. It's a big change and I would personally say you should be moving to the cloud because it offers you functionality you don't already have, and if it can save you money too then obviously that's great. But I wouldn't move to the cloud just because it was cheaper (unless it was massively so) because of the other risks and disadvantages it can bring, which are sidelined too often. At the end of the day, your Internet bandwidth is supposed to be used for Internet access - I don't see how it's efficient to start using bits of it to handle traffic that could be zipping around your internal network perfectly fine. That capacity should be going on cat videos, not print spooling! (Just on that point, I agree with @psydii about file storage. It's not necessarily the case that the document you're printing will itself be in the cloud, which raises the whole print server issue again.)

That's funny as that's what they themselves said... I'll check that out later on. @CyberNerd, agreed that a lot of it depends on protocols, but then of course that's out of our hands. Ping latency is helpful but can be misleading when trying to measure the performance of other things - multicasting for example. Either way, while we could debate the performance of various people's Internet connections and other measures by providers, I think the main point is that we're involving the wider interwebs in things they don't need to be involved in. Why would you send data all around the houses when it only needs to go from one place to another internally? It's a very inefficient process if there is no tangible benefit and that should be avoided out of principle if nothing else. If everyone started doing this with all their workloads the bandwidth use, however small, would begin to add up (thinking at the pipeline level here, not just for individual connections). That means ISPs would eventually have to expand their capacity, which requires a cost (which will be passed on to customers) that ultimately might not have been required if people didn't insist on chucking things into the cloud by default. The cloud should be used where it offers a clear benefit in terms of both functionality and cost - nowhere else. The reason we have nothing cloud-based here is because I've yet to see a situation where that would be the case for us.

Remember it's not just the bandwidth, it's also the latency - and reducing that much further would require a step-change in the way connectivity is delivered. Could we actually get latency down to similar levels to an internal network? I'll believe that when I see it!

^ This. I think that whoever we end up sending off to see M$ in May, and whatever else is discussed, we should agree to put forward a collective position along precisely these lines, and make a fairly big thing of it. Who would agree?

Also, even if the document is stored online, owing to the way the printing subsystem works, the data is transmitted by the workstation to the print server (which is necessary for many reasons, one of which being authentication), so that would involve the data coming down the line to the workstation and then back out again to the cloud-hosted print server VM. It won't be the case that the workstation simply instructs the server hosting the document on e.g. OneDrive to access an Azure VM in another area of that cloud directly for the data transfer. It's the same reason why, with file shares, it's best to copy from the machine you're logged on to from a remote one - if you're logged on to your workstation and you copy from one remote share to another, the traffic will come via you and it will take longer. Another example of workload I wouldn't want in the cloud for the same reason is Impero (or similar). When you monitor screens (which of course you can do in bulk), the data goes from the workstation to the server and then back to you (they have confirmed this). Do I really want the bandwidth usage of multiple internal workstations' screen monitoring to go from them, to the cloud and then back to my workstation, as it would be if the Impero Server VM were cloud-hosted? Or would I prefer all of that to stay on the same internal network with far less latency - key word there - so things aren't chugging around through Internet pipelines that don't need to be? I think you're right in that it will be possible to run just about anything from the cloud in a few years' time, but I don't think that means it will ever be the best choice for every conceivable workload.

Haven't received an e-mail - assuming I'm not selected? We've had overblocking issues upstream with County lately so thought I'd better check.

I don't know if that's true - I'm not sure when the business logic would ever apply that it makes sense to transmit a document from a workstation to a print server over the Internet for processing, then transmit it back down the line again to a printer on the local network. It's a clear example of a workload that should definitely remain on-site where you have the choice, especially as the connection bandwidth you do have will be under more strain from all your other cloud workloads by that point than it is today.

Reported, but not confirmed, although many would welcome it. The fact that many academies (particularly SATs) will choose to keep using the Local Authority for some central services even though they are no long required to is now well-established.

He means the services buyback such as payroll etc. - he doesn't mean any oversight or control.

The Conservative problem with public authorities in very broad terms is that, to their mind, "they're just not good at innovating". I will say for the record that I know exactly what they mean. Where I diverge from their approach - significantly - is that the answer to that problem is to fix it. Stimulate the public sector; streamline it to an extent, yes, but also put pay and progression structures in place that don't encourage everyone to stay where they are and don't encourage recruiting managers to match people to roles instead of matching roles to people. Do give people incentives to be excellent at what they do and find ways to ensure all staff are in some way "in contact" with the direct results of their work (which doesn't have to mean performance pay). Aside from pay compression, these issues have been the biggest problems with Single Status, and I do feel it's representative of a good deal of other thinking within the public sector over the past couple of decades. And the fact of the matter is, all of this is fixable. What it also is, however, is expensive - and that means the current lot won't touch it with a ten-foot barge pole. (Then again, I take the exact opposite position on Europe, so I guess it's my view that different solutions are best for different problems - but let's not derail this thread with that... )

When are we likely to know?

@ZeroHour, just read the e-mail you've sent. Am I right in thinking there will be two days held, or will MS just pick the day favoured by most? I ask as it would be tricky to put in an absence request for "this day or maybe that day"...

I found this bit in particular from another BBC article to be rather telling: