Feeds

back to article Microsoft's cloudy chief: Azure reliability knocks your own kit for six

Microsoft ships an update to Azure every three weeks. Scott Guthrie, Microsoft's vice president for Windows Azure, reckons it's adding features and capacity in an effort to catch up with Amazon. Just announced is a new Azure datacentre in São Paulo, Brazil. "We have more regions than Amazon, we have coverage in places like …

COMMENTS

This topic is closed for new posts.

...not just a service for customers, but used extensively by Microsoft itself

Isn't that a bit ironic. People are scared of cloud specifically because it's NOT IN HOUSE - they lose control of the servers, data etc.

Of course MS would use azure because it IS IN HOUSE for them. Would they run their infrastructure on google or amazon? no. then why would they expect anyone who's not Microsoft to run their infrastructure on Azure.

16
1
Silver badge

VisualStudio online TFS

Login

Wait

Try login again

Go to msdn blog to see if any outage report - nope

Got to Reddit to see if anyone else has a problem - yes

Keep trying to login

Read story next day on reg about 12hour Azure outage

It probably is more reliable than an in-house server. But at least I know if the in-house server is borked and don't just sit there hopefully pressing retry.

13
2

Re: VisualStudio online TFS

Indeed your right and im no Azure fan, more AWS myself. I know if you can get involved and try and debug the problem yourself with an in house server you probably feel better because your doing something, but are you really going to solve it quicker than the types of sys admin experts the cloud operators employ? I mean, seeing as your not a full time sys admin im guessing you don't know as much as them, so it is perceivable you may never fix it?

3
5
Silver badge

Re: VisualStudio online TFS

With my own server the problem is TFS.

With their service; it might be my client, might be my router, might be my ISP, might be any one of the multiple Microsoft Online/Live/Hotmail/Corporate/Office365 accounts I seem to need to have to login to any Microsoft service, it migth be their Visual Studio service, it might be Azure, it might be one of their data centers - I have no idea.

It's like saying that since I'm not a professional olympic athlete cycling to work can't be as reliable as the mixture of Bus+Train+Tube because they are all run by professionals.

(It also doesn't help that half of their help pages for hosted TFS tell you to login into the server and start entering TFS commandline admin commands.)

6
0
Anonymous Coward

Re: VisualStudio online TFS

"Go to msdn blog to see if any outage report - nope"

Next time try looking on the Azure service status console.....

2
5
Silver badge

Re: VisualStudio online TFS

That doesn't help me if it's the visual studio server that is down.

Azure currently says "compute service performance degradation" doesn't say if that's what's stopping me logging into TFS, doesn't say when it will be fixed.

Having to know their internal details of which host their hosted service uses is hardly the point of cloudy-ness

4
0
Silver badge

Re: VisualStudio online TFS

'Read story next day on reg about 12hour Azure outage'.

Does Azure run on Linux or Unix? If not, there's your problem. Surely a no-brainer for anyone wanting to run lots of networked servers with high reliability and security.

4
0
Anonymous Coward

Re: VisualStudio online TFS

"Does Azure run on Linux or Unix? If not, there's your problem. Surely a no-brainer for anyone wanting to run lots of networked servers with high reliability and security."

Azure runs on a dedicated Hypervisor layer in a similar model to VMWare. which has a number of advantages of a Linux / Unix type monolithic hypervisor - such as much lower security vulnerability counts and smaller attack surface, better performance and better security.

0
5
Anonymous Coward

But you can do something about your own kit. You're completely at the mercy of Microsoft if their kit fails.

5
0
Silver badge

O reeely?

One of my machines was first powered up in 2001. It runs every day. What is actually worse is that it is switched on and off twice a day. The only failure I've had in that time was when the mains went nuts in a thunder storm. Replacing the PSU with a spare cured the problem.

Now then microsoft, how long has your cloudy thing been running, and how many times has it failed for more than about an hour?

6
2
Anonymous Coward

Re: O reeely?

For billions of compute hours, and not very many (fewer times than say Amazon S3...)

2
5
Silver badge

Re: O reeely?

Ah yes. The 5 person company that's 20 years old saying they have 100 years experience.

To make anything aproximating a reasonable comparison you would need to compare with enough individual computers to add up to a similar processing power.

Good luck with that!

P.S. Amazon being (apparently) worse doens't make them any good.

2
2
Bronze badge

Re: O reeely?

Except that's BS - your stats are always BS, all MSFT communication is full of &^%$ as they keep repeating the same nonsense about *average* hourly stats, conveniently forgetting about the nature of their failures: they are SYSTEM-WIDE and last HOURS unlike, say, Google's usually limited, short/intermittent, rare outages...

...which one would you rather live with, dear (paid) MSFT-troll: outages that bring all your services down for HOURS ~2x a year or short, few minutes long disruptions few times a year eg IMAP slows down significantly but web UI works fine etc?

HINT: any sane person with cloud-based critical services would go for the second one (thus stays away from MSFT.)

10
1
Anonymous Coward

Re: O reeely?

switched on and off twice a day

You mean you reboot it at lunchtime, coz it's started running slowly?

2
1
Anonymous Coward

Re: O reeely?

You mean like this system wide failure that lasted hours?

http://www.computerworlduk.com/news/security/10155/more-outages-hit-amazons-s3-storage-service/

Or this one?

http://www.computerworlduk.com/news/cloud-computing/7538/outage-hits-amazon-s3-storage-service/?intcmp=in_article;related

Or this?

http://www.networkworld.com/news/2012/122612-netflix-restores-streaming-video-after-265361.html

0
2
Anonymous Coward

Re: O reeely?

Look at the date on those articles you cite - surely it's conceivable that they have improved their reliability over the last 5 years?!

0
0
Anonymous Coward

Re: O reeely?

"surely it's conceivable that they have improved their reliability over the last 5 years?!"

Nope:

http://www.datacenterknowledge.com/archives/2012/12/25/major-christmas-outage-for-amazons-cloud/

http://www.datacenterknowledge.com/archives/2011/04/21/major-amazon-outage-ripples-across-web/

http://www.datacenterknowledge.com/archives/2013/09/13/network-issues-cause-amazon-cloud-outage/

http://www.zdnet.com/amazon-web-services-suffers-outage-takes-down-vine-instagram-flipboard-with-it-7000019842/

0
2
Bronze badge

Right - but somehow you forgot to mention that your cloud also...

...FAILS MANY MORE TIMES times than all the others (probably even combined) and you are usually unable to ISOLATE the outage, it always seem to become a CASCADING cluster*&^%$...?

5
0
Anonymous Coward

But...

The VMS cluster we have at work has NOT been down once since it was put in 1992. That's 21 years folks.

Sure the H/W has changed. It has move from 32bit Vaxen through Alpha's to Itanium. The storage has also changed but the rub is WFT should we change if it works. Sure HP has called time on VMS (a totally stupid move if you ask me) but with the current plans we will see it working well into the 2020's.

Would someone please tell us what moving to the cloud would give us apart from a direct feed of our data from the cloud to the NSA?

Answers in a pin head please.

12
0
Gold badge

Re: "what moving to the cloud would give us"

Well if I read the article correctly...

"instant, scalable web sites with integrated application deployment built into Visual Studio"

...the MS cloud has a handy button in Visual Studio that lets a programmer who is too stupid to deploy changes through controlled procedures nevertheless hack the production servers and push "go" just before knocking off for the night.

3
2
Anonymous Coward

Re: "what moving to the cloud would give us"

"...the MS cloud has a handy button in Visual Studio that lets a programmer who is too stupid to deploy changes through controlled procedures nevertheless hack the production servers and push "go" just before knocking off for the night.

The MS cloud model includes full approval workflow and automation capabilities, so that you can choose it you want programmers to be able to make that sort of change with enforced change control or not...or if you want it automated after approval....

See some examples are in http://blogs.technet.com/b/keithmayer/archive/2013/11/27/free-ebook-hybrid-cloud-management-with-system-center-2012-r2-app-controller.aspx#.UrDh2Y3uND8

0
1

It's a strange one...

Cloudy stuff.

Just as hardware is becoming cheaper, VM's easier and bandwidth faster - everyone is pitching for us to abandon our in house expertise. It is, in many respects, counter-intuitive.

It reminds me of 7-8 years ago when gigabit to the desktop was just taking off at the same time as decent wireless, all the big cheeses wanted both, with no understanding of the impact of either.

In the few years we have been dealing with cloudy stuff, perhaps a dozen toes and ourselves I have been shocked at how a fairly humble Amazon setup can spiral out of cost control due to all the little "ryanair" addons and a rogue process or two.

I have been aghast at the monkeys selling Google services and happily copying up gigabytes worth of data with no further business continuity, backup or other sensible plans. I have been horrified at subsequent lack of support when things go wrong (Mail routing, corrupt docs, mx transfers etc). Quite frankly I have been ashamed to be part of the industry when said monkeys have quoted for simple things to happen (merging of account, primary domain changes etc etc).

I have yet to see any proper change control docs for the web interfaces or apps required to use these services. I can go for a piss and come back top find apps across 50 devices are 'slightly different' due to on the fly updates.

I think most people in the tech world enjoy the fact that they are the 'go-to' people. The buck stops with them. If they can't fix it, they know a man who can. If it can't be fixed, then it can be worked around. We like solving problems, we like solving problems eloquently and we actually revel in knowing our shit. We accept that people only call us when things go wrong. We like that.

When you have a hung instance on a Azure / Amazon / faceless compute node, when email just isn't being accepted at an SMTP interface, when files are all different versions or when half the workforces internet is down... then in a (fully) cloudy world we have no answers. And with no answers we have no purpose.

10
1
Stop

Scott

You need to get out more and meet people who are not Microsoft customers and ask them about their downtime.

4
0
Bronze badge

When all your eggs are in one azure coloured basket . .

. . you better have a pretty strong one.

Or an escape plan when the interweb goes AWOL.

p.s. It will also pick it's moment to fail, just after you said "We'll test the backup tomorrow".

2
0
Silver badge

Re: When all your eggs are in one azure coloured basket . .

Every imperial mothership has escape pods!

1
0
Silver badge

(*)

(*) Offer not valid on leap day.

3
0

Microsoft can go

suck as can the cloud

3
1
JDX
Gold badge

Re: Microsoft can go

I bet 20 years ago you were advising companies "don't waste time with the internet, it's just a fad".

1
5
Silver badge

Re: Microsoft can go

In fact it was Microsoft that initially took that stance!

3
0
Windows

Re: Microsoft can go @JDX

I bet 20 years ago you were saying dump Unix and go NT: it's far more secure, crash proof and you will have near zero downtime.

0
0
Happy

Can give global scale to dev departments with ease

From my point of view there has been the occasional serious issue- mainly the leap date and SSL ones. However as a platform for .Net developers it has become increasingly feature rich and simple to use. It is very easy to build applications with global scale that cost very little to run. For us it wasn't a case of in-house as we would never have been able to build many of the apps as we don't have global data centres. I just love the freedom not to worry about infrastructural details and focus on getting the job done - great feeling scaling out successful apps.

If you are building .Net web apps that have to good reliability and global reach then it certainly has some great features. The development experience has improved greatly with much improved integration in VS. I think MS are really getting the agile thing and are impressively evolving their current development product range. I'd like to see them offer more add ons from 3rd parties e.g. RavenDB - there are good ones - but not enough.

Obviously I would like to see less outages and more reliability. Hopefully each outage is a lesson learned. The incident reports read like they are taking steps - with the scale of what they are doing there will always be issues. Hopefully there becomes less and less types that can occur.I just don't see the average in-house being so prepared for such a range of issues.

The big cloud providers are going to eventually be rock solid so I'd rather be ready for that than living in the past. As great as Azure is we have certainly isolated Azure specific code so it can replaced - I just don't see it being required. It's worth the odd issue to have the freedom not worry about hosting globally scaled apps.

3
6
Anonymous Coward

Re: Can give global scale to dev departments with ease

Thank you for joining the Register today to place this advert !

6
0
Silver badge

Who's reading your data?

They still have the problem that being an American company, you never quite know who's reading your data. Even with non-US cloud providers there's still some doubt.

At least if it's your own server you have a better idea on that score, unless your network security is lax.

4
0
Windows

"From a general availability and reliability perspective I feel pretty good that our systems are better than pretty much every customer system I have met,"

He's basically saying that Windows reliability sucks ... if he says so, fanbois can shut up.

More to the subject, having applications in the cloud is OK, I think, like a global process automation solution. Having your data in the cloud is dumb, though, having your security in the cloud is even dumber ... it would not only give the miscreant access to your data in the cloud, but also your on-site data. Then again, what do Window cleaners know about reliability or security ? Right ...

5
0
Coat

"Advert Warning" missing

El Reg, how much did you get to write this fluff piece?

Is it worth getting a reputation over?

6
0
Bronze badge

hapless Microsoft followers

Microsoft's Azure cloud chief is obviously joking by stating that " Azure reliability knocks your own kit for six" when the service has crashed twice in the past two months, or gone "tits up worldwide" as 'artfully' reported in TheRegister.co.uk recently, and four times in 2013 alone..

Fortunately for Microsoft many corporate executives who worship the Microsoft brand and personally admired Bill Gates as the world's richest technologist, will fall for this factually baseless hype and pour all their trust and data into the Microsoft botnet cloud hole.

Good luck to them, the sad, hapless ignorant fools.

5
0
Thumb Up

"Just announced is a new Azure datacentre in São Paulo, Brazil"...built to NSA specs in order to become intimate with a certain Guradian reporter there.

0
0
This topic is closed for new posts.