When will MS learn to use a decent server OS and not a scaled up single user, single microprocessor OS.
Last time Microsoft's Azure cloud went down, it was a sub-component that flaked out globally, and the time before that it was a certificate problem – now the service is inaccessible again, along with its status page. And the wobble has taken down the Xbox.com website on the Xbox One worldwide launch day. Other services having …
"NT has always been a multi-user multi-threaded OS since day 1, unlike, say, Linux which had such things bolted on later in life)."
Err, what? Linux was multi user and multi process from day 1. Threads came along later but then when you have multi process multi threading is a nice to have rather than an essential. Remind me - can Windows fork() processes yet? No, didn't think so.
As for NT multi user - sure , so long as you don't want more than 1 user logged on at the same time it was "multi" user. ie - fucking useless.
"Not entirely sure why you would want to "fork" a process under NT. Yes you can execute out to one, (with all the overhead that uses), but generally a thread or two inside the same process will suffice."
If you don't understand the pros and cons of mult process vs multi threaded programming then I suggest you go and read up about it. And executing out to one - ie starting up a brand new process from scratch - is NOT the same as doing a fork(). But then if you had a clue you'd have known that.
Yep, a thread or two inside the same process so they can gleefully trample each other to death, neatly sidestepping the advantages delivered by multiple address spaces.
Multiple threads have a place; IBM had them in their mainframe OS at least as far back as MVS (1974 or earlier) and used them extensively in products like CICS and DB2. Separate address spaces provide protection from other processes and, for many applicatons, a far simpler, if less flexible, programming environment. Both techniques are useful, each in its place.
But but but but...NT is a MICROkernel.
Have they trotted out that old nugget yet?
I still fondly remember the old Unix World article on NT where Davey Cutler proudly announced he was going to "kill UNIX." Didn't work then, not working now. But I hear he's doing GREAT things running the Xbone division!
(Because that's what the X-box developers needed, a bloody drill sergeant.)
@W.O.Frobozz - I think you'll find that while NT didn't kill UNIX, it certainly broke the stangle-hold of Big Iron UNIX vendors on the datacentre. There will always be a place for the older technologies, just in the same way that Big Iron UNIX didn't kill the mainframe, NT won't kill UNIX and Linux won't kill Windows NT, they'll all exist in a heterogeneous environment with each performing to its own strengths.
>When will MS learn to use a decent server OS and not a scaled up single user, single microprocessor OS.
Yeah, I always get blank stares from Windows users when I explain that Unix cut its teeth on university servers where everyone was trying to hack into everyone else's accounts.
(disclaimer: speaking, perhaps naively, as a dev, not a sysadmin)
However, I don't get the sense that Windows Server itself is the issue with all these Azure failures. The underlying OS may or may not be able to support "the cloud".
Rather, Microsoft as a cloud vendor, not an OS vendor, just does not seem to have the right mindset to _manage_ a cloud. They always seem to be stumbling into single-point-of-failure management component snafus.
Amazon has had a few of those in the last few years, but MS seems to just have more of these failures and be less good at learning from them.
i.e. MS would probably still screw up Azure if it was riding on top of Linux rather than Windows ;-)
First off, that's what Azure is... a cloud services platform. Where else would they run their cloud services?
Azure itself isn't down. Still more detail coming out but it looks like there was a global DNS issue which looks to be cleared up now. Unclear where the bad DNS info originated.
Fortunately it looks like the problem didn't last very long.
Yes but why not have an Azure enterprise and an Azure gaming, they don't have to mix. Much like running a business you don't mix personal and business finances, sure you could but your accountant will give you crossed eyes at all the trouble you have given him.
There was no detail when I started noticing the drop in services, DNS wasn't on the radar at that moment, but I'm curious now; playstation network has been having DNS issues last night, our internet service at the office lost DNS early morning. Man in the middle again, like Afghanistan blocking YouTube?
Can we now infer that the GTA:online cloud services are provided by azure too? They have been shit since launch, are guaranteed to fail every weekend and coincidentally many people are reporting time outs, emptying sessions & cloud unavailable issues today and it isn't even the weekend.
I have suspected for a while that MS were providing Rockstar's (abysmal) cloud services. Would someone at El Reg care to investigate (if web journalists do any investigating anymore?)
 Seems they are investigating, but R* don't want to talk, I'll try to reign in my cynicism in future :)
More cloud!!! Quick, everyone - jump on before you're left behind!
I've been rather sweary lately so I'll refrain for the moment but all the same, it's a getting to be a joke. Not because things go down - that happens - but because companies like MS are fairly FORCING people onto their cloud platforms. Please. Stop.
Oh look, it's my stalker!
If you look carefully at the article, you'll note that my "demonstrably untrue" comment has in fact just been clearly demonstrated, once again, by Microsoft. It even notes some of those other occasions when Microsoft demonstrated it. There are in fact articles going back decades that demonstrate how awful Windows networking is, and has always been, to the extent that it's become a cliché.
But don't let something as trivial as the facts get in your way.
No, not your stalker - I did see your comments the other day and suggested that if you don't want people to look at your posts and form an opinion about you because of them, there is always the "post anonymously" option.
BTW: What makes you think that you're interesting enough to be stalked?
As for your comment that "MS know nothing about networking" and suggesting that an outage to DNS proves this, you seem to conveniently disregard the fact that there a millions upon millions of devices from living room to desktop to server running MS networking software and doing it perfectly happily. One outage does not "know nothing" make. There are plenty of things to accuse MS of, but being clueless about networking really isn't one of them. Not since about DOS 3.1.
"All the IP stack stuff is in the registry"
Whaaaat ??? Since when did executable code live in the registry ??
Oh, I see, you meant to say "All the IP stack configuration stuff is in the registry"
Which broken configuration method that Windows offers you choose to employ when porting (and it was impressive for MS to come up with an even more broken system than .ini files) is pretty irrelevant to the heredity of the code you're porting.
You may not like the Registry, but storing configuration in a versioned, ACLd, database seems like a pretty good idea to me. You can't scatter config across the system as happens in some other OSes, you don't have to put up with a config file being an all or nothing access thing, so you can allow users the ability to control aspects of their settings in an extremely granular manner. You also don't have the problem of non-standard names for files, which occurs in other OSes.
There are plenty of things to accuse MS of, but being clueless about networking really isn't one of them. Not since about DOS 3.1.
Oh puhleez. We had to stick a Wollongong IP stack on MS Denial of Service 3.1 to make it use anything resembling TCP/IP, and it was still damn hard work as DOS 3x didn't really have too much in the way of ISO layering (even then, MS "integration" got in the way of Making Things Work (™). Heck, even ARCnet was hard work, but that was admittedly also a byproduct of how PCs worked in those days - the eternal fight with IRQs and other settings, and then the joy of trying to get as much of the TSRs above 640k so you could actually still get software to load so it did something for the user too. I can recall using PowerLAN in those days in the small company I was working at, and file exchange with our other offices was some stuff I cooked up with modems and batch files (just in case you thought not having sufficient IT budget was a modern thing :) ).
When MS then decided that networking was the thing (because it was losing its fight with Novell, even after altering Windows so it appeared "incompatible" with anything but MS DOS), it cooked up something called Windows for Workgroups which was soon dubbed Worries for Workgroups because it created a lot of interdependencies that could take a whole office down. In those days, Those Having A Clue were not MS, but Novell and IBM (well, ish - depends on your opinion and experience of Token Ring), with a helping of companies that started down the TCP/IP route over Ethernet.
It took Citrix to show MS that servers could do more than just sharing some files and print jobs, but the history of MS and networking is patchy, and that's being positive. Do I need to remind you that Gates dismissed the Internet as a fad? You know, the guy that told us that 640k ought to be enough for everyone and then forced us to upgrade kit and memory every time they came out with a new release?
You're confusing volume with quality. They still can't run a decent service - see all the outages. Anyone who creates a business dependency on any MS operated cloud needs IMHO his or her head examined.
"Don't let the fact that your comments are demonstrably untrue let you stop, though."
I present two examples:
- MS suggesting that VM's moving between HyperV hosts should get new MAC addresses. Maybe it was just a dumbass L3 support person who suggested this, but still...
If your switches can't handle dynamic MAC address changes, you should probably get more modern switches. Besides, it's perfectly reasonable to change MAC addresses for virtual machines which are moving between physical hosts, I fail to see why this might be a problem?
"If your switches can't handle dynamic MAC address changes, you should probably get more modern switches. Besides, it's perfectly reasonable to change MAC addresses for virtual machines which are moving between physical hosts, I fail to see why this might be a problem?"
It's not a switch problem - they will happily move the packets around. However, layer-3 devices don't know the MAC address is changed and continue to address packets to the old host MAC.
HyperV's competitors have an elegant solution to this same issue that shows some understanding of the networked environment.
Just read up on the MAC address problem - it appears that the MAC will only change at reboot time (all MACs are by default generated dynamically at boot time), if you do a normal migration between systems it shouldn't change, until reboot. You can use SCVMM to assign a permanent MAC (I assume there's a cmdlet to do this as well, should you not have SCVMM). If you're seeing other behaviour, you may want to check it out as it doesn't appear to be what's supposed to be happening.
By default the MAC address doesn't change - this was a MS technical support provided fix for an issue with HyperV hosts not communicating with clients after moving servers between HyperV hosts. When the HyperV host moves, it doesn't notify the switch that the MAC address has moved ports (as VMware does) and instead waits for the server to send outbound traffic. Not so great for a web host.
As a user there are clearly problems with the networking within Windows.
Shares contantly get lost across the network, which may be more of a browsing issue, but it's still under the umbrella or networking. Permissions go walkabout and often never come back.
We've had to bolt on an 'update' to XP machines (for update read frig) so that Windows 7 macbines can see XP machines. If that isn't a bodge then I don't know what one is.
W7 sometimes forgets wehter it's at Home, at Work or in an Internet Cafe. That's perhaps more of something going wrong in the higher client layers but to a user using the network that's still a network issue.
How many times in various incarnations of Windows have we had to use tools to flush the TCP/IP stacks?
It can't be that hard to follow the protocols and to to get them work, can it? If other operating systems can do it then surely Microsoft can.
Since I started this message a Win7 lost a share with an XP development box (i.e. the share is still there but the machine has vanished according to W7) but the reverse from the XP development box to the clueless W7 machine is still rock solid.
Networking is certainly not one of Microsoft's strengths. They almost had it nailed down in the days of XP but even then there was that possibility of it just grinding to a halt. With Windows 7 there seems to be extra layers of cotton wool and confusion.
I vote for the motion that Microsft is clueless about networking.
Sorry to say it, but there is something wrong with your environment - I've been working on extremely large heterogeneous environments (FSTE100 and US equivalent) for nigh on 20 years with Windows forming the vast majority of workstation/file/print and I've not seen any of the problems you are seeing.
It's always worth considering the problem is with your own IT before blaming the OS/hardware.
Where all your capabilities come to a screeching halt the very moment your service provider decides as much (planned or unplanned doesn't really matter here). And some people wonder why I'm quite opposed to software subscription models ;-)
I'm sure all those Office 365 users are quite thrilled right now, I also wonder if people can actually play any games at all on their XBox One.
In the mean time I simply start up Word 2010 by clicking on its icon in the start menu and can get to work. Even if I'm no longer connected to the Innernets, such an amazing achievement don't you think?
That shows a basic misunderstanding of what Office 365 is about - while you CAN use the web service to host and edit your documents, you can (and for an enterprise, you SHOULD and almost certainly WILL) also store them locally on your company server and edit them with an instance of Office 2013 installed locally on your workstation. Just like the old days. The idea is to give you flexibility, and it does that - whether the price is worth it to you is another question.
It is about pushing everyone to a subscription model and making their "local" Office suite phone home every now and then to enforce that.
And that means that you are open to having your "local" copy locked and not be able to continue working if you don't pay your subscription. Or are you expecting me to believe that the day I would stop paying I could still access my documents ? I don't think so.
As far as I'm concerned, Office 365 can go stuff it. I don't need to have my work stored outside of my purview and I will NOT accept to submit access to MY data to a third party.
....... the biggest boss was asking me about Office 363, and this is why I love Microsoft. When a techy needs help saying migrating to it at this time would be bad for us they create a service failure to back me up.
Keep it up MS, and ignore these guys as they still haven't worked out you are secretly helping us.
"Service Unavailable - DNS failure. The server is temporarily unable to service your request. Please try again later. Reference #11.27ddf180.1385076682.341919,"
speaking as one who's had problems when a DNS server failed the routing and I couldn't get access to fix it. that's definitely Akamai reporting a problem contacting servers. I'm guessing something propogated badly into Akamai and was harder to eradicate than anyone thought. It's a design I've seen used (and fail) before.
Quite handy really that our Helpdesk system is hosted in the cloud. We didn't have to log calls from all our users when they couldn't reach their email, since the helpdesk system was down too.
Anonymous, because embarrassingly this is true - the great and the good decided is was "better" to move many services to one cloud - what a great experience for our users. (NOT)
(It may or may not be Microsoft my company has used, it's the concept that's important)
This post has been deleted by its author
...I can only repeat myself: anyone is relying any Microsoft online service for his critical services must be fired on the spot, period. MSFT has the industry's WORST RELIABILITY TRACK RECORD and WORST ARCHITECTURAL SETUP when it comes to DEALING WITH CASCADING ISSUES, ISOLATING OUTAGES, it's a fact.
Biting the hand that feeds IT © 1998–2021