Google suffers international outages, slowdowns Earlier today, web users across the globe were reporting outages on myriad Google services, including Gmail, Google News, Google Docs, Google Calendar, Google Analytics, Google Maps, Google AdSense, and Google Search. Google has acknowledged the problem and says it has now been solved, blaming the traffic slowdown on a routing …
They take the data around by plane? Someone introduce these guys to the Internerd or something. No wonder so much carbon is generated by a Google search -- although the cheeseburgers and newspapers might have something to say about it too...
Don't forget to add the icon.... done. Don't we have an icon for bad joke? I guess that's what the coat one is for, then.
what's with reporting every single minor issue with accessing some google service for an hour or so? good job the register never has outages
http://uptime.alertra.com/uptime.php?accountID=067807&abrv=LREG
...oops!
it's not the number of outages that matter really though but rather how you deal with outages - which google seems to be quite good at! (their architecture is apparently based on the expectation that some of their servers will die all the time) - unfortunately routing issues can't really be completely prevented and it's impossible to stop routing issues causing outages unless you have 2 separate networks with separate routes - which is probably overkill for anything that isn't in charge of keeping people alive (and if you're running systems that critical over public networks, they are probably already dead by now)
anon because i criticised el reg, and that tick box will make it impossible for them to figure out who i am ;)
"it's impossible to stop routing issues causing outages unless you have 2 separate networks with separate routes - which is probably overkill for anything that isn't in charge of keeping people alive"
OK, so you're not in charge of critical websites then. And you don't know why a Google outage is a big deal.
Why are you arrogant, then?
FYI, these days the world+dog run their crap on multiple networks. There are even specialized large scale ISPs that trunk multiple carriers together, when you plug in your fiber cable.
And google outages are important for their relevance to the reliability of cloud computing. you really do not want to be the CTO of a company that keeps their docs at google today.
If you add the above two things together, you're talking multibillion dollars businesses. That's why this is important.
If you're thinking Google, think Microsoft - BIG TARGET. So why bother when there are superior services out there? Dogpile for searches (no Google profiling or search rank manipulation), my ISP's interweb mail thingy that plugs into Outlook. In my book, anyone who uses Google for anything is a) behind the times and b) dumping their data on the US government - seen who Google execs share a little airport with?
Ancient Wisdom of the Human Race describes the Thing Then Unheard-Of yet again. Sorted!
As for that meet-up with the Bilderbergers mentioned by Spiteful God @ 17:50 GMT: That does it. Given only the accuracy of Spiteful God's assertion, the ol' "Don't Be Evil" GoogleMotto is here and now declared Officially and Infandously Hypocritical, per Backchannel-Informed Fiat of Walking Turtle, All the appropriate caveats and boyctts do thus hereby apply to every emptor.
Gah. Blarg. Plugh!
Mine's the one with the somewhat battered copy of the Complete Works of Aesop in the pocket. Rght next to the paperback of Sun Tsu's Art of War.
multiple upstreams and multiple networks are different things
you can have 10 upstreams, but traffic from a specific source will pick one of those upstreams to come in through - for example from my home ISP traffic to google appears to go via LINX, if there's an issue at LINX i lose access to google - the only way to prevent this is if there is another IP Address not routed via LINX that my client can use instead
remember if a link has a fault then it might not always cause automatic re-routing if it isn't detected, and if it is detected then it will still take time before the routing propagates and service is restored via other routes
so yes the only way to ensure 100% availability is by having 2 different active routes which do not share any common equipment, which means effectively 2 networks (although if they are interconnected they could still form 1 physical network, whilst being 2 separate logical networks) - google do not do this, as all of their routes are advertised to their peers causing the peers to send all traffic over that same peering route
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
