How many times did Google Apps Engine went down since it's launch

[ahmed adel]

I need to know how many time did Google apps engine went down since
launching this service also I need to know for how many hours and what
is the time exactly.

Also I need to know your opinion about the below.

We are planning to develop a product over Google apps engine , a big
project actually for our company.

so should I start using Google apps engine or should I use another
PAAS and if so what is the recommended PAAS.

thx

[Joshua Smith]

Depends how you define "down."

If you only consider HR, it hasn't gone completely down ever. Some bugs have happened (like losing the ability to send mail to gmail addresses with .'s in them) which reduced functionality for HR, and there have been times when you could not update your app. But I don't think HR has ever stopped serving since it was introduced.

M/S has been down lots of times, sometimes for extended periods. But that's irrelevant if you are using HR, which you should be if you care about downtime.

-Joshua

> --
> You received this message because you are subscribed to the Google Groups "Google App Engine" group.
> To post to this group, send email to google-a...@googlegroups.com.
> To unsubscribe from this group, send email to google-appengi...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.
>

[Brandon Wirtz]

JeffProbst.com that I host on AppEngine has had 100% uptime for 6 months. At
its peak it served just shy of 1million visitors in 12 hours.

[Greg]

I haven't seen any downtime since I moved to the HR datastore. I run
my entire (small) business on Appengine now, and am very comfortable
with that decision.

Having said that, it depends on your application. If it is a general
business app, then you'll be fine on Appengine. But if people or your
business will die if it goes down for ten minutes, you need to make
other (very expensive) arrangements.

[Diego Ariel Fejgelis]

Ahmed, maybe this could help you a little bit http://code.google.com/status/appengine

[ahmed adel]

What do you mean by HR?

What I meant is , If I have application hosted in Google Apps Engine ,
will it be affected in the future for example due to data center crash
like what happened before.

We will need this application be available starting from 7 AM to 1 PM
Cairo time. because of this I am asking about the crash time in Google
Apps engine.

I am asking this because my manager wanna be sure our app won't be
available for use. (internal and external).

Also I heard that Google apps engine was down many times and according
to this many companies changed the host from Apps engine to another
PAAS provider.



Another reason for my question is I suggested to develop over Google
Apps Engine for our Customers , but also I told Our management that we
can not trust Google Apps Engine to do this for now , because as you
know and correct me if I am wrong (Google don't care much about Apps
Engine Support)


so my question will be now , can I trust Google Apps engine to use it
for our Clients? in case of yes or no please tell me reasons for your
answer.

[Joshua Smith]

Google had a lot of trouble with their original database, known as M/S. It was responsible for the downtime you heard about.

So they invented a new database, known as HR, and it has never gone down since they released it.

So if you use HR (which is the default for new projects), history suggests that you will not see outages.

But be realistic: Everything goes down. Everything. There is no 100% uptime guarantee. Google is promising 99.95% which means they are OK with 438 hours of downtime a year. And that's only a soft promise (like most SLAs, your recourse is negligible if you suffer losses because of their failures).

I'm quite sure that GAE on HR has better uptime than I could achieve with anything I could build and manage.

[Philip]

Joshua, your calculation is wrong. A year has 8760 hours, 0.05% of
that time the service can be down -> 8760 * 0.0005 = 4.38 hours of
downtime.

[Steve Sherrie]

Close enough!

[Brandon Wirtz]

Josh lives on a planetoid most of the way to Neptune so his year is longer
than the rest of us.

SLA is really only worth something if it is monetarily backed. Hopefully
your hosting cost is not greater than your operational income. If Hosting
is 5% of your total budget on a site making $100k a year each hour of
downtime costs $11. 4 hours costs $44.

If someone else has 99.99% uptime then down time costs $4 a year.

In both of these cases the refund based on the SLA being exceeded by 48 due
to catastrophic failure (like Amazon had a while back) is going to be
Pennies.

Take this to an enterprise level... Like at say Howcast.com that was a
former employer, 50-ish employees Hosting was about 1% of their budget, much
of their revenue came from Promoted content that was released on a schedule.
Missing a deployment or being down for 48 hours during the launch of a
campaign could have cost 100s of Thousands of dollars.

For this reason I look much less at the SLA, and far more at the reputation
and track record. (and I always have a hot spare ready to fail over to.

If you want to make sure your APP stays up during "the end times", make sure
that you have designed it to have at least read only functionality in the
event that the service drops in to that mode. HR isn't supposed to but I
predict once in the next 2 years it will happen that HR has 6 hours of read
only time.

Deploy a duplicate version of your app on MS, do enough syncs that this
could operate when HR is down. So far outages have been on either MS or HR
not both.

Route all GoogleBot traffic through a Caching proxy, so that all of your
indexed pages exist in a static form on a Non-GAE service. There are other
advantages to this technique as well, so I would recommend this even if you
aren't just optimizing for up time.

-----Original Message-----
From: google-a...@googlegroups.com
[mailto:google-a...@googlegroups.com] On Behalf Of Philip
Sent: Saturday, October 08, 2011 11:16 AM
To: Google App Engine
Subject: [google-appengine] Re: How many times did Google Apps Engine went
down since it's launch

[Maximillian Dornseif]

On 8 Okt., 17:42, ahmed adel <dev.ahmeda...@gmail.com> wrote:
> What do you mean by HR?

"High Replication" the recommended setup introduced about 8 Months
ago.


We switched a internal Application from a single Server hosted at
Rackspace to AppEngine.
Below you can see an analysis of a very expensive (->slow) internal
page.

We are testing from 12 Locations around the globe. See percentage of
requests which didn't
return success within 10 seconds:
http://static.23.nu/md/Pictures/reachability_6E56359C.png

And second: latency of that page:
http://static.23.nu/md/Pictures/latency_0985D51B.png

Can you tell what day we switched to AppEngine? It is an incredibly
useful product and I trust it much more than servers being run by
myself.

There are occasionally issues where you can't reprogram ("deploy") app
engine for a few hours. That is annoying. but user facing stability of
HR applications is great.

--md

[Joshua Smith]

D'oh! I knew that looked too big! :)

[Thiago Rossi]

Fat bill? :)

By the way, I never really understood how the CPU time is counted. Let's say my doGet or doPost takes 20 seconds to run. So 20 seconds for each request? Because well, the servlets are there, available, always...