See what happens
When automatic updates are left enabled?
Four days have passed since a "procedural operations error" downed Azure SQL Reporting in Microsoft's East US data center, and Redmond is still trying to restore customer data. After saying on Thursday that full restoration from Monday's fail would have occurred by Friday, the recovery date has slipped again, according to …
When automatic updates are left enabled?
Is that the (IMO: self-proclaimed) "experts" on El Reg's chat this afternoon (regarding Win Server 2012) were quite specific when it came to Azure and its development model. I quote:
"Alun Rogers: Maybe this is more of a joining up with Azure release schedules? They iterate that like crazy"
"MJF: I think this move helps put MSFT's "Cloud OS" campaign into more context. The idea is Win Server should be the best OS for building/supporting cloud services. If that's true, it needs to be evolving in lockstep with Azure, not lagging it feature-wise for years at a time"
The topic at hand during these quotes was "Blue"; the assumed new setup where Microsoft pushes out new software more or less continuously (and on a subscription bases I might add) instead of releasing major versions.
It would appear as if both guests had no clue - what so ever - regarding the current dire state of Azure, something which IMO speaks for itself when reading the quotes. I shudder at the thought of a Windows server being released / developed at a quick pase as Azure, not merely for the current state Azure is in, but the idea in general scares me considering how Microsoft still firmly holds the reputation that a v1.0 release is usually filled with bugs and other nasty stuff.
Yet these two seemed to think that it would be the most ideal situation for customers to be in. Are you kidding me?
Now, this maybe a little bit of a cheap shot on my part, sure, but now taking the current state Azure is in into mind combined with the fact that they haven't been able to fix this within a whole week I'd say its safe to conclude that hasty releases on Enterprise level aren't the brightest of ideas.
Of course I'm no expert :-)
"Windows is not ready for the data centre"
Ha ha ha ha ha! What the hell do you think runs services? Organic lolcats?
MS have rather a better record than say Amazon - that use Linux - so I guess we should assume from your logic that Linux is even less ready for the data centre?
Not sure about the trillions pounds though, however:
They moved to Windows and .NET and CRASH!
Mind you, even when it didn't crash it couldn't deliver with 10% of speeds maintained by some Linux rivals (2-4 msec vs .4 msec). It's now allegedly performs at sub 150 mksecs speeds.
That's more about the choice of developers (Accenture) and technology (.NET) than anything.
When they moved to Linux they didn't run it on Java or Mono did they?
If they had written a C++ application on Windows in the first place it wouldn't have been so bad. You can't really compare an app running on a VM on Windows with a native one on Linux.
It's funny you should raise the whole .NET thing when so many mobile apps are running on VM type technology. Android being the big one. iOS applications are native using objective C.
Azure runs some of the biggest cloud services in the world like Xbox Live and Office 365, so to suggest that it doesnt scale is rediculous. They might not have as many customers as Amazon but they certainly have loads more cloud customers than Google.
I note that Azure has been far far more reliable than Amazon has - who have had a number of extensive, sustained and region wide and even wider outages. And they of course use Linux...
At the time that went in, Microsoft were already beating those speeds in other Stock Exchanges:
That experiment cost them a thousand thousand million pounds Eadon? I am of course not saying that it is impossible but I would really appreciate a link to that - if you can manage it, hmm?
I'm no Eadon fan, but when I read Trillion, I imagined Dr Evil saying it, with his little finger in his mouth, and Minime by his side, and got the point without getting all aspergers about it.
I'm sorry but Java is crap too.
When a system is under strain the garbage collector will eventually kick in and pause the app. While Java has had a lot of development there are so many systems out there using 1.4 still and that's horrendous. With 1.6 and 1.7 things are a lot faster.
How do you know Microsoft creates crap code, have you ever seen any of it? when the NT source code was leaked it was analysed and found to be of good quality.
If it was so bad then how are so many businesses managing to run so much on it? administering Linux is a very expensive affair and the emloyment pool is much smaller than Windows.
If Microsoft 'writes shoddy code' then how come they have so many fewer security vulnerabilities than Linux distributions do?
The story of my life with Microsoft.
"No data was harmed in the incident, but getting it back is taking a long time."
Well actually the data was totally fucked up, and getting it back is taking even longer than a long time... but yes, the statement is mostly correct.
Linux = Reliability.
Linux does not = Reliability any more than Windows = Reliability. Any sys admin worth their salt will tell you that things are a lot more complex than that. The issue is that they cant recover quickly form an outage and that is usually more to do with their operational procedures and safeguards than the particular OS. However its hurting their customers and thats bad for business.
@eadon. As Mr Stallman aka RMS is so fond of reminding us Linux is not the Kernel. I'm pretty sure that at some point Amazon is going to have another large scale outage and it runs Linux. Its the law of averages. The outage may have zero to do with the kernal and then all of the MS fan bois will be saying linux is crap. Ask Netflix how many of its customers cared that they were running on Linux when it went down on Christmas eve.
Linux is not proper Unix. Proper BSD has better uptimes.
It's not all about the kernel.
There is the filesystems too. Having run Linux as a desktop and had filesystem corruption you tend to want the much friendlier more reliable repair process of Windows and NTFS>
Backing up your ext3 filesystem to another drive and running some recovery process on the image isn't acceptable.
Since the NT kernel is basically a blatant ripoff of VMS, it's actually very good (except for them shoving the GUI layer into it for speed). The rest of the system however...
Having run Linux as a desktop and had filesystem corruption you tend to want the much friendlier more reliable repair process of Windows and NTFS
Please share your sci-fi story with us. A far as my experience is concerned, MS and Lenovo restore tools were not able to see a corrupted ntfs partition on the hardrive nor could they restore the user's backups. ( The commented story issue might be pretty similar to that.) A live Linux it was able to both mount it and copy user's data onto an external hdd.
In case of a similar accident happens to some Linux filesystem, you just boot to a single-user repair ncurses session and follow pretty straighfoward instructions there, or just boot off a rescue cd to run gparted (or is back-end e2fsck) on the affected partition.
Please, find some other case study from the "Get the facts" record, or just come up with something more plausible.
@seansaysthis: "Linux is not the Kernel"
Actually, Linux *is* the Kernel. That;'s the basis of RMS's gripe. What he wants to remind us of is that Linux is *not* the OS... which is why RMS keeps on referring to the OS as GNU/Linux.
Probably helps to UNMOUNT the file system first before running an fsck. A fool with a tool is a dangerous fool. Don't blame the hammer when you whack your thumb.
Well if individual experience is valid tot his discussion (although your choice of ext3 speaks volumes (pardon the expression) for your little piece of fiction) I have had the unpleasant experience of a chkdsk on windows running for 36 hours while the business fumed waiting for it to complete. Sort the disk off line? Oh right, then the system wants to run chkdsk to ensure it's all been run to its satisfaction. Meanwhile, I also have experience of a linux complete disk system failure at a site across an ocean and 3000 miles away from a sysadmin. Total business time lost - 4 hours, including the hardware being replaced.
Neither your comment nor my anecdotes have anything to do with the story under discussion though...
Actually as Windows has a hybrid micro kernel architecture, versus a large monolithic kernel in Linux, it is the Windows kernel that will be 'more secure, more robust, requires fewer reboots and is more resilient' - what you may or may not be right about is the rest of the OS around it - but claiming that on a kernel level that Windows is inferior is clearly BS. Windows has a fundamentally more modern and more streamlined kernel architecture with lower complexity and a smaller attack surface than Linux.
The User GUI layer is now removed from the kernel again for security / modularity. And it is still the fastest OS for Open GL benchmarks....
Some interesting information here:
No system is perfect. There is this idea that the cloud is magic and infallible. Its still hardware and software and many meshed systems its just sitting in someone else's facility. There are very few cloud providers that haven't had some issue that impacted customers. This happens in the enterprise and unless it impacts external customers you dont get to hear about it. Enterprises dont air their dirty linen in public unless they have to.
There is a cultural thing where the new cloud web 2.0 kids seem to think that just because you stamp cloud on a service that you forget all of the lesson learned by operational teams over the last 30 years. The basis tenants of keeping things running remain the same. As much fun as it is to poke fun at MS, Amazon, Rack space , IBM or HP it underlines the issue that no large environment is free from outages.
The best operational teams understand this and have good plans in place for recovery. They assume something will break and have factored this into their business and operational process so when something does go wrong they can recover quickly. The issue here is not that something went bang its that its taken them over a week to fix it.
And they're still not done fixing it.
Now you can generalize all you want, saying that cloud structures are all at risk etc. etc. You're not inherently wrong, you're just forgetting that Amazon and Google are managing much larger volumes of data and are doing so much more efficiently than Azure is apparently capable of.
Microsoft is building itself a history of failing major products in an embarrassingly public way. Some failures can be explained by market conditions, but this Azure failure is a technical one, and that is a stain that will simply not go away.
We all know how this is going to end. The service will be restored, Microsoft will triumphantly tout the excellence of its platform that lost no data, and the weeks it took to get to that point will be smothered under a pile of pillows. For Microsoft, this will be a success story.
For everyone else, this will be the baseline for Azure reliability : when it fails (and it will), it takes X weeks to get back online. As said in previous comments, Azure is already at a reliability rating of one 9, and that is a failure in any administrator's book.
I'm not defending MS , if it was my business being affected Id be pretty pissed. I dont think even MS PR department could spin this one out a success story. My point was that the recovery time is unacceptable an that adequate recovery procedures weren't in place or it would be sorted. MS have again shot themselves in the foot.
You realise this was fixed days ago?
Don't know if it will get past the censors but it was so satisfying to write...
If they do then Apple will take their place. People like commercial software of a good quality.
Free software of a hugely unpredictable and varying quality isn't what the masses want. People buy things these days, the days of building your own stuff is long gone.
People generally don't want free software any more than people want to knit their own jumper or make their own clothes.
Android is not free softwware. You already need to pay license fees to Microsoft. And possibly to Oracle and Apple in the future...
The reason Azure is flaky, is due to rushed implementation, backup process failures and a general slap dash approach to cloud by Microsoft!.
Build your Linux and Windows on VMware Cloud Director accompanied by robust DR concepts and and a more professional attitude to delivering Cloud properly and the situation will be much more robust if any of the componenets fail.
The whole Azure, Hyper-V, Windows Server 2012, SCCM is a rushed coded cobbled together ecosystem that is going to burn all the MS shop "follow like sheep" fanbois!. This will not be the last time customers will suffer long downtimes in all type (Azure, Hyprid, Public, Private) MS cloud mash ups!.
Pay the money and you get reliability. Its a no-brainer.
Hyper-V, Windows Server and SCCM are all very well proven and at least a match for the best of the competition. And don't forget SCOM, SCSM, Orchestration Manager, App Manager, etc...
Azure is relatively recent and is still introducing new features, but has a better reliability record than some of it's largest rivals.