According to him, the cost of bandwidth is probably the issue, and would make Oracle a compelling choice even if Zoom didn't have negotiating leverage.
Nota bene: it gives me no joy to say good things about Oracle. I hope it will not happen again.
This is absolutely wrong unless you have rank n00bz leading your technology. This isn't even touching the implication that anything is harder to do outside of cloud hosting providers, it's just accounting for, well, the entire industry prior to, oh, let's say 10 years ago. It's not as prehistoric as it sounds, and in many (if not most) cases, the cloud providers are just slapping new trademarkable names on those classic architecture choices.
It makes sense when they have revenue targets to reach.
So, if that reputation is well earned, it's important to consider whether you'll actually come out ahead in the negotiations. One thing that's almost certain is that Oracle has a lot more experience in negotiating this particular form of contract (where one side tries to screw the other while still somehow retaining their business) than you or your company does.
It is still something to think about though when shopping providers, I assume any major cloud provider will have great peering, but a gigabit line from a random dedicated server provider will be noticeably slower for many customers.
Things like fixed cost throughput links, and fixed network pricing is available with contracting for large companies.
It’s likely oracle simply “bought the business” and is loosing money on the hosting. Hoping to make it up with a co-sell commission based contract, or by leveraging this as PR/marketing.
Cory Quinn is right. List price is expensive. But if you spend 1mil/year or more you shouldn’t be paying list. If you are zoom getting 90% off egress is totally negotiable.
Corey Quinn seems like a smart guy with a lot of connections. If he doesn't regularly have big fish in the AWS ecosystem telling him (off the record!) "here's the deals we got", he's not doing his job right.
If someone similarly knowledge comes back and say "actually, Oracle doesn't beat AWS on bandwidth when you're a big fish", I'll say not to trust this article.
AWS technically doesn’t do discounts. They just invent new usage tiers with phenomenal pricing. Which effectively means bulk discounts.
I’m sure Zoom is getting a great deal on outbound bandwidth. Especially since they are multi-cloud and can shift traffic between providers at a whim to squeeze their providers.
Ironically they would probably qualify for much bigger discounts if all of their workloads were in a single cloud.
But it's not 100% clear to me: are you saying it's likely their AWS bill is low enough to invalidate this analysis?
Also, if they're as cloud agnostic as it sounds, wouldn't they have evaluated the cost-savings of single cloud while negotiating?
I'd say the analysis is accurate in directionality but maybe not magnitude.
But more importantly I think it's just Corey's way of poking Amazon about their high prices. :)
But I will say that Amazon is pretty clear that you'll get a better discount if your workload is 100% with them. Like if you put 1/3 of your workload in AWS, it will be more than 1/3 the price of putting your whole workload there. Or to put another way, if you triple your workload you won't triple your price.
Can't be too hard. It's Oracle so they let you in nicely enough, then they screw you later when you're locked in, increasing cost substantially on every renewal, changing terms and billing arbitrary things.
I think Zoom could get away with any terms really (AWS is not hard to beat on bandwidth to begin with). Oracle cloud is desperate for customers, even more than Azure and GCP, but Oracle has the willingness to increase costs tenfold and change terms arbitrarily later unlike competitors, so all things considered don't expect them to have too much regards for the (initial) terms.
On the database they are notorious for trying to charge test databases (normally don't need a full license), any server they can find (up to you to evidence it's not running oracle) or even your entire virtualized datacenter because VMs can migrate so the entire datacenter should be the machine when it comes to billing and they bill by core.
From a PR standpoint it makes sense for Oracle to offer Zoom a deal that they initially make no money off of, bask in the PR value, and prioritize work that reduces the labor and equipment cost for managing a customer like Zoom until they do make money.
What work is AWS or Google going to do to reduce overhead on a Zoom-sized customer today? Wouldn't they have already done it?
The Oracle news is news because it’s a deviation from the business as usual, I suppose.
If my traffic goes up by 10x and now I'm overpaying 30 FTE's worth of cash, I am going to have a hard time stomaching that fact, and now the opportunity cost situation is reversed. What could I do with another team?
1. "AWS/GCP/Azure are price competitive with Oracle for this use case, but Oracle will give Zoom a better deal because they're trying to get business" is pretty boring. 2. "AWS/GCP/Azure overprice their bandwidth and Oracle doesn't" is interesting.
Wouldn't you want to send the data peer-to-peer between the users rather than funneling everything through the Zoom server?
You'd get better latency, and you wouldn't have to pay for the network streaming costs...
Source? Some apps like bittorrent have increased in the past decade, have they not?
Peer to peer simply doesn't work since NAT. Skype used to be peer to peer but they backpedaled a while ago because this and more reasons that simply made p2p obsolete and broken.
It's too long to cover in one message. Would you be interested if I write a blog post? How peer to peer architecture became obsolete and why no software is using it.
In a scenario of 12 people video chatting around the world, I don't believe you could guarantee consistent high-speed connections between the involved parties.
Teredo tunneling was one of the coolest solutions to this problem, providing both NAT-busting and IPv6 support but unfortunately it never gained popularity.
Ya, I'd be interested in reading that! Could you address why it wouldn't work with NAT the same way bittorrent deals with it? I'm probably missing something, but I don't see why bittorrent can work so easily but video conferencing cannot. thanks
Edit: Does bittorrent require you to port forward? I didn't think it did, but I haven't seen it configured in years so maybe I am forgetting that step
Some machines can't be reached. Many others have UPnP capable routers and can reconfigure NAT for port forwarding on the fly.
Additionally more and more users have IPv6 at home. It's now high enough to be worth trying along with other techniques like STUN.
P2P is great if you're small, but eventually it makes sense to do client-server with servers in the data centers to get that extra 10% of customers. Skype and Spotify are both formerly-P2P applications that went client-server.