back
6 comments
If all 6 million G Suite customers, with an average number of users at 25 per G Suite account, paying the $20/user fee, requested the three day credit for this breach in the SLA contract for the outage, it'd cost Google about 300 million dollars.

Which is .22% of there COH this quarter...

Or on the regular basis, they'd get $300M every other month (exclude any fee)
Your nines or their nines?

I bet if you personally can't use it, but their overall reliability meets the bar, then they're within SLA.

Don't ask why I know this.

You know there are companies who build crawlers or health check agents exactly for this purpose, so that they know precisely from when to when the service they pay or they need for their business doesn't work and went out of SLA. I think it's brilliant and the only way to make any company pay. I believe you can sometimes get away with a couple of pingdom checks/jobs.
Like they have any bargaining power against Google.
I don't know how it works in such cases, I mean, this is publicly known that there was an outage.

What I believe is that customers will probably get free GCP credits and that's it, everything is good as before.

This person service-provides
In [1] it says: "Customer Must Request Service Credit." Do you know how to request it?
Admins can go to admin.google.com and click the help button to start a support request.
It seems like they claim around 45m

https://techcrunch.com/2020/12/14/gmail-youtube-google-docs-...

Is this official announcement? I don't find any link to google in your posted url.
not an announcement, seems like just a statement. I assume an official statement will come with the post mortem blog post in the next few days.
If you click on the red status dots, it has report with timing.
SLAs are largely bullshit.
why?
This topic just came up recently on a podcast I was on where someone said a large service was down for X amount of time and the service being down tanked his entire business while it was down for days. But he was compensated in hosting credits for the exact amount of down time for the 1 service that caused the issue. It took so long to resolve because it took support a while to figure out it was their service, not his site.

So then I jokingly responded with that being like going to a restaurant, getting massive food poisoning, almost dying, ending up with a $150,000 hospital bill and then the restaurant emails you with "Dear valued customer, we're sorry for the inconvenience and have decided to award you a $50 gift card for any of our restaurants, thanks!".

If your SLA agreement is only for precisely calculated credits, that's not really going to help in the grand scheme of things.

I like your anecdote, I might steal that one.

IANAL, but I negotiate a lot of enterprise SaaS agreements. When considering the SLA, it is important to remember it is a legal document, not an engineering one. It has engineering impact and is up to engineering to satisfy, but the actual contents of it are better considered when wearing your lawyer hat, not your engineering one.

e.g., What you're referring to is related to the limitation of liability clauses and especially "special" or "consequential" damages -- a category of damages that are not 'direct' damages but secondary. [1]

Accepting _any_ liability for special or consequential damages is always a point of negotiation. As a service provider, you always try to avoid it because it is so hard to estimate the magnitude, and thus judge how much insurance coverage you need.

Related, those paragraphs also contain a limitation of liability clause, often at capped at X times annual cost. Doesn't make much sense to sign up a client for $10k per year but accept $10M+ liability exposure for them.

This is just scratching the surface -- tons of color and depth here that is nuanced for every company and situation. It's why you employe attorneys!

1 - https://www.lexisnexis.com/lexis-practical-guidance/the-jour...

> Doesn't make much sense to sign up a client for $10k per year but accept $10M+ liability exposure for them.

Businesses do this all the time, this is how they make money. And they use a combination of insurance and not %@$#@*! up.

I've never seen an SLA that compensates for anything more than credit off your bill. I can't imagine a service that pays for loss of productivity, one outage and the whole company could be bankrupt. If your business depends on a cloud service for productivity you should have a backup plan if that service goes down.
I haven't seen one (at least for a SaaS company) that will compensate for loss of productivity/revenue etc, but something like Slack's SLA[0] seems like it's moving in the right direction. They guarantee a 99.99% uptime (max downtime of 4 min/22 seconds per month) and give 10x credits for any downtime.

Granted, there's probably not many businesses that are losing major revenue because slack's down for half an hour, but it's nice to at least see them acknowledge that 1 minute down deserves more than 1 minute of refunds!

[0] https://slack.com/terms/service-level-agreement

> I haven't seen one (at least for a SaaS company) that will compensate for loss of productivity/revenue

They won't show up on automated systems aimed at SMEs, but anybody taking out an "enterprise plan" with tailored pricing from a SaaS, will likely ask for tailored SLA conditions too (or rather should ask for them).

It's hard to give a compensation for profit loss, as then you would have to know the profit of the customer beforehand and put an adequate pricing including that risk. It's almost like insurance!
In financial markets I have seen SLA's where you will make people whole on the losses they incur due to downtime you inflict on them.
Seems like you want insurance. As with the hospital bill you'd generally be paying a bunch of extra money for your health insurance plan to not get stuck with the bill.

Not sure that exists for businesses, but I'd expect you'd need to go shopping separately if you want that.

Seems like a good business idea if it doesn't exist.

I think the idea here is that if the payment for SLA breach is just "don't pay for the time we were down" or (as I've seen in other SLAs) "half off during the time we were down" that doesn't feel like much of an incentive on the service provider.

They have other incentives, obviously, like if everyone talks about how Google is down then that's bad for future business. But when thinking of SLAs I'm always surprised when they're not more drastic. Like "over 0.1% downtime: free service for a month".

Independent 'a service was down' insurance isn't the same though. It is important for the cost to come out of the provider's pocket, thus giving them a huge financial incentive to not be down. Having that incentive in place is the most important part of an SLA.
Even with insurance, some of the cost will come out of the provider's pockets - as increased premiums at renewal (or even immediately, in some cases). Insurers might also force other onerous conditions on the provider as a prerequisite for continued coverage.
I hear you, but there's going to be a cost for that. For the sake of argument, say Google changes the SLA as you wish and ups the cost of their offering accordingly.

Would they gain or lose market share?

I don't think it's obvious one way or the other.

> . . . like going to a restaurant, getting massive food poisoning, almost dying, ending up with a $150,000 hospital bill and then the restaurant emails you with "Dear valued customer, we're sorry for the inconvenience and have decided to award you a $50 gift card for any of our restaurants, thanks!".

It's even slightly worse than that. SLAs generally refund you for the prorated portion of your monthly fee the service was out, so it's more like "here's a gift card for the exact value of the single dish we've determined caused your food poisoning." Hehe.

You're right and the funny thing is that's exactly what he said after I chimed in.
Completely agree with your analogy but have you ever seen any SLA that provides any additional liability? I haven't seen them - you are stuck either relying on hosting services SLA or DIY.
I cannot share details for the obvious reason, but yes - there are SLAs signed into contract directly behind the scenes which result in automatic payouts of a condition isn't met and it's not a simple credit.

Enterprise level SLAs are crafted by lawyers in negotiations behind the scenes and are not the same as what you see on random public services. Our customers have them with us, and we have them with our vendors. Contract negotiations take months at the $$$$ level.

That is a fair point. Is this a situation where you asymmetrically powerful? I have to imagine you would have considerable clout to represent a fair bit of their revenue in order to dictate terms. When I wrote my comment it was in the vein of a smaller organization.
I am but a technical cog in the machine my friend, while I know about what goes on in business and contract negoatiations I cannot comment on power dynamics. I would assume it's like any other negotiation - whomever has the greatest leverage has the power, I doubt it's ever fairly balanced.
Or purchasing business continuity insurance.
Not OP, but how do you measure them? Let's say, for example, you can send and receive email, but attaching files does not work. Is the service up or down?

What if the majority of your users can access the service, but one of your BGP peers is not routing properly and some of your users are unable to access?

Google do a very good job of defining their SLAs.

In answer to your question, they'll accept evidence from your own monitoring system when you claim on the SLA. They pair that up to their own knowledge about how the system was performing, then make the grant.

Google are exceptionally good at this, from my experience. Far better than most other companies, who aim to provide as little detail as possible while getting away with 'providing an SLA'.

The SLA itself should specify the way availability is measured.
Down because email attachments are base64 encoded files written in plaintext into the body. So if those are not working, email itself is not working.
that was a bad example. i guess the comment was trying to say "how do you account for partial service degradations".

(i dont think SLAs are BS btw)

In this example: you get free days. Which depending on your business might be worthless if you have suffered more monetary loss due to the downtime than the free days are worth.
But still better than nothing. And for some (most?) people/businesses, probably worth more than any resulting monetary loss
Exactly; downtime doesn't cost a cloud service much. At worst it causes reputation damage, with possibly large companies deciding to go for a competitor, losing a contract worth tens or hundreds of millions.