Ubuntu's Mark Shuttleworth Pulls No Punches on Red Hat and VMware in OpenStack Cloud (zdnet.com)

At OpenStack Summit in Vancouver, Canada this week, Canonical CEO and Ubuntu Linux founder Mark Shuttleworth came out firing at two of his major enterprise OpenStack competitors: Red Hat and VMware. He claimed that Canonical OpenStack is a better deal than either Red Hat or VMware's OpenStack offerings. From a report: Shuttleworth opened quietly enough, saying, "Mission is to remove all the friction from deploying OpenStack. We can deliver OpenStack deployments with two people in less two weeks anywhere in the world." So far, so typical for a keynote speech. But, then Shuttleworth started to heat things up: "Amazon increased efficiency, so now everyone is driving down cost of infrastructure. Everyone engages with Ubuntu, not Red Hat or VMware. Google, IBM, Microsoft are investing and innovating to drive down the cost of infrastructure. Every single one of those companies works with Canonical to deliver public services."

Then, Shuttleworth got down to brass tacks: "Not one of them engages with VMware to offer those public services. They can't afford to. Clearly, they have the cash, but they have to compete on efficiencies, and so does your private cloud." So, Canonical is rolling rolling out a migration service to help users shift from VMware to a "fully managed" version of Canonical's Ubuntu OpenStack distribution. Customers want this, Shuttleworth said, because, "When we take out VMware we are regularly told that our fully managed OpenStack solution costs half of the equivalent VMware service."

Comments

Re:Ouch

[KiloByte]

Looks like it only supports Linux containers

Well, you can use KVM if you want, but it's usually not a good idea. Containers are drastically more efficient than paravirtualization like Xen, and that in turn is drastically more efficient than dumb old virtualization. Yes, full-blown virtualization offers better separation of virtual machines, but for example the recent crop of Intel bugs allow breaking out of a VM just "fine".

Ubuntu RedHat

[bigmacx]

RedHat became the Microsoft of Linux very soon after that crappy IPO and re-running the share auction that f*cked a lot of us. My Etrade screenshots haunt me.

Ubuntu #1 sever. Linux Mint #1 desktop. There are no close seconds IMNSHO

Re:VMware and RHEV, 1 person, 1 day

[LostMyBeaver]

I think it depends on the long term goal of what you're hoping to build.

I don't know RHEV but I am more than a little experienced with VMware. Sure, VMware can be setup pretty quickly. But VMware is a disaster after it's setup. You end up with bunches of VMs and bunches of LUNs and heaven forbid you use NSX or worse, Cisco ACI, you'll end up with a rats nest from hell. VMware, should never be used anywhere you need more than a few virtual machines.

See I guess the point is that if you're simply trying to build a system which facilitates 1990 style IT in a virtualized environment, VMware is ok. Probably the best tool for the job ever really. But this is 2018. We are supposed to be smarter than that.

I develop software for the enterprise and I've also taken over operations in our environment. They were wasting $1.6 million a year on VMware and servers. It was a nightmare.

I asked "Why are we spending all this money on 10Gb/s networking?", the answered "We need it for storage and for vMotion".

So I pointed out that the software itself would operate well on 10Mb/s networking, but that faster would decrease transaction latency as 10 times the line speed means 1/10th the time to transmit a single frame.

So I asked "My software works almost entirely using an in-RAM database. I don't read and write from disk except for cold storage. Why do you need that much performance for storage?". They attempted to make the case that it was required because they need to be able to migrate the virtual machines if needed across the hyper-converged storage.

I asked "Why would you migrate the VMs at all? They're designed to simply die and be replaced by newly spun up ones". And they explained that we can't depend on that.

I asked why they had 2 10Gb/s switches for each data center and two more in the middle. And I wanted to know why we were spending a lot of money on MPLS between two sites. And worse, if we lost a data center, would we be down to a single site/single point of failure. They explained that it was too expensive to add a third data center and that it wouldn't make sense and that the risk was acceptable as there was so much redundancy in each data center that if a catastrophe on that scale occurred, we had bigger problems.

So I talked to the CFO and asked him to give me a breakdown of what IT was costing us a year for the data center. I wanted the consultant and power costs as well. We were spending $3.0 million+

So, I started a grass roots project. Bought a 3D printer, 10 Cisco 8 port network switches, 40 Raspberry Pis and the additional accessories I would need. Then I built 10 "Pods" which is a switch and 4 Raspberry Pis. Then I setup Kubernetes and Couchbase and setup a CI/CD service. And then I simply deployed our applications (written in C#) and done. We use a distributed load balancer, but more for logging than anything else. We use IPv6 mobility and AnyCast for all our failover stuff.

We now have an IT cost for data center operations of about $25,000 a year. We have 10 sites for failover... what have a share-nothing database which can suffer MASSIVE outages and never miss a beat. We can scale to all our sites (1000+) if we want. And we can handle substantially more transactions a second than we ever could on VMware.

Oh and if any site goes down, we throw it in a box and put a new one there. It's not worth debugging.

Oh... the rest of our systems are pretty much SaaS these days. We did some research and realized that 95%+ of what we were running in our data centers was stuff that we needed just to keep the data center running. The remaining 5% of the system resources were used for communication, collaboration, accounting, identity and our internally developed applications. So we just stuck everything but our own stuff in the cloud.

When people go on about RHEV vs. VMware vs. Ubuntu, I can't help to think about a room full of men with hats discussing optimal horse breeding for pulling delivery carts while Henry Ford is building a factory to produce cars.