Blog Post

Lessons from the trenches: Joyent co-founder on current opportunities in the cloud

[protected-iframe id=”4145bb0ad7ad1f8dad80c18a04ed0d3a-14960843-25766478″ info=”″ width=”640″ height=”360″ frameborder=”0″ scrolling=”no”]
Session: GigaOm Structure Europe 2013: Conversation.

Chris Albrecht
Om Malik
Jason Hoffman

Chris Albrecht 00:02

Thank you David. Thank you Ulf. We have a bit of a change up in the schedule but it’s going to be awesome. We have Om Malik is going to be talking with Jason Hoffman, they are both founders. Om founded GigaOm obviously, and Jason Hoffman founded Joyent, and they’re going to be kind of having a founder- to-founder discussion. So let’s bring them both to the stage right now. Welcome Om and Jason.


Om Malik 00:39

So there’s a little mix-up Marco was supposed to come and have a conversation with me. He indisposed because of some family issues and I hope they get resolved very soon. So we send him our best wishes. Here is freshly unemployed Jason Hoffman. He is as of–

Jason Hoffman 01:05

I’ve been calling it retired, but you can–unemployed fines.

Om Malik 01:08

Unemployed dude. Come on let’s keep it real.


Jason Hoffman 01:13

Retired pretty real.

Om Malik 01:19

Retired means–

Jason Hoffman 01:20

It’s okay to be jealous though. [chuckles]

Om Malik 01:23

I don’t want to be to be in your shores too. Trust me. Those shoes, no. Jason and I have been friends for a very long time, obviously–

Jason Hoffman 01:31

A long time actually.

Om Malik 01:32

We were the early believers in cloud long before everybody else got onto the cloud band wagon and the world changed and so on and so forth. I’ve always enjoyed talking to Jason because he’s been one of those people who can see around the corners better than anybody and before everybody.

Jason Hoffman 01:55

That’s true. I agree with you.

Om Malik 01:56

Today I was hoping we can ask him to look back and share some of the lessons he’s learned from building a cloud infrastructure company and when folks in the audience go back and apply–why they can apply to their business. Jason tell us what you’ve learned in ten years of Joyent? From an infrastructure standpoint. I know you know a lot, but let’s just keep it to the cloud for now.

Jason Hoffman 02:28

I think if you look at the way I think of the cloud it’s been largely an industrialization of IT effort. So you have this sort of trend an industrialized factory manufacturing type of approach to how you’re building infrastructure. Everything is in fact now becoming a service so even the transformation and internal IT organizations have to do is becoming a service provider – even if it’s internal. What we’ve really seen the rise of in particular over the last ten years is really the rise of the developer. The global phenomena is developer based and you look at whether it’s a fortune one all the way up to sort of the rest of the global 2000 to start-up. The sort of IT buyer now and people who are fundamentally making the decisions are actually writing applications and they have data. This sort of developer cloud versus enterprise cloud, versus these sorts of things is a bit of a false dichotomy. That’s been that sort of industrialization effort and the rise of the developer and really sort of sits down and says you have to do a lot more developer focused type services and so on.

Jason Hoffman 03:53

From a business standpoint I think what’s been interesting around the infrastructure as a service side that a lot of people sort of don’t understand is–take a pretty simple example. If I look and say 2009’s, 2013 at Joyent if you look at 2009 prices – and our prices have always been about say what Amazon’s prices are. Those prices are discounted 83% relative to their 2009 number and then internally the cost is discounted 98% to sort of deliver an equivalent size VM. You see a lot of these even when people do these cloud efforts and they last one year, two years -they don’t really do three years, four years. There’s this whole thing around tail and life cycle management and what you’re doing in year three, four, five from an infrastructure standpoint, and how you’re almost dealing with that type of heterogeneity verses a vendor heterogeneity. What most people are messing up on is the tails end – most people are messing up on later years.

Om Malik 04:57

It’s more of a financial, arbitrage in the aspect of the cloud which nobody talks about. But that seems to be the fundamental difference between Amazon and everybody else, and Google and everybody – Amazon included.

Jason Hoffman 05:11

What Amazon is really exceptionally good at is Amazon is exceptionally at just being customer focused and really being able to drive buying behaviors. It’s not surprising considering what their core business has been, but as far as going and analyzing what people actually doing. What do people actually want? Rapidly iterating on delivering those types of things. It turns out when you sort of take what you need to do from a retail model point of view you can in fact apply it. You can imply that type of intelligence to going and delivering any time of commercial offering – in particularly they’re very good at that.

Om Malik 05:53

Would you think there’s any hope for others to compete with Amazon in the market? It seems like they have the scale advantage they have the financial arbitrage advantage. They also have the understanding of the market better than everybody else – you competed with them. What do you think can somebody actually take them on and actually win against that?

Jason Hoffman 06:15

Yeah. I think so. I think so because if you look at the–right now it really comes down to if you look at the big players and big meaning people that are spending more than $2 billion a year on CapX around this – that’s basically Amazon, Microsoft and Google now. Depending on how you slice the market, when you look and say if Gartner goes and counts the numbers of VMs they’ve always set the top five from VM deployed standpoint was…Well they said Amazon of course was in there and then Joyent was in the top five. Presumably Rackspace and Terremark is probably in there too or something like that from that magic quadrant standpoint. We almost talk about compute too much. We talk about running VM’s and everything like that. If you look at what’s been the exponentially growing services at Amazon and just for all of us they’ve been data services. So really it’s like an Jason Hoffman, it’s Redshift, it’s dynamo, it’s doing things like the Manta storage service. It’s basically sitting there and saying, “You need to do an object store. You need to do a relationally store for people. You need to do a key value store. You need to do a data-warehousing thing. You need to have some sort of map produce framework.” Fundamentally it’s around providing data services at the end of the day, not just running VM’s.

Om Malik 07:39

Do you think with all of the advantages you’ve listed out for all the public cloud vendors; you, Amazon, Google even Microsoft. What’s the role–do enterprise cloud efforts even have a shot? Like guys at VMware are trying to do different things and Dell has its plans and other people have made their plans. What do you think?

Jason Hoffman 08:02

That’s really sort of a physical server to virtual server business. That’s sort of selling just a lot of the similar things. It’s still not taking say a full service oriented type approach to it. Again if you look at we almost talk about things like a persistent compute service or an EC2, we talk about that a little too much. And again that’s actually the easy thing to do, it’s actually easy to run servers and it’s easy to run VM’s from a relative sort of point of view. The hard part becomes, how do you go and have these multi-data center distributed data services -it really ends up saying that. This is where Goggle is extremely interesting. No one questions the fact that Google can scale up a data service – no one questions that. It’s pretty clear that you look at things like Jason Hoffman and Dynamo there’s no questions that – Amazon can scale that. There’s no question that Microsoft can do the SQL server type service – there’s no question that they can scale that. I think as we start moving from talking about having servers and then having physical servers and having virtual servers and having things. Even this conversation where we look at the enterprise – typically enterprise stuff – and we say, “Well we write these applications.” The conversations becoming a lot more about data. What data do we have? Where should it be? What applications consume it? What applications generate it? When you take that approach you’re actually significantly more likely to look at cloud provider and say. “Well instead of bringing my Oracle database to the cloud I’ll actually use their relational database store. That is in fact dumping a bunch of your data in an object store or using those types of services that are actually driving that. All of the other players effectively have to get into…What they should be doing is they should be buying up noSQL companies. They should be picking these. They should be developing. You only have to do about eight to 12 different services to cover every data type and every query type to go after–

Om Malik 10:20

Even internally? You’re talking about the big companies; Amazon, Google. I was talking to the CEO of LSI and he argued that the companies which are going to do well in cloud aren’t the ones which are HyperScale, which is the VM companies is the ones he talked about. In China he listed out Buy-Do and Tennsco, Revol and folks like that. So from that standpoint do you have any thoughts on the Europe open cloud opportunities given that Europe doesn’t have an OS company or a large-scale storage company or large web company at that gigantic scale. They have a lot of great companies just not at that scale. You think that puts the continent at some kind of a disadvantage or you think of it–?

Jason Hoffman 11:15

I think it could be a timing issue, somewhat as well. Because if you look at the hard problems in infrastructure – the hard problems do basically boil down to operating system problems, storage problems and effectively how you’re doing distributed applications. How you’re actually doing leader election and consensus and all of those types of things. If you look at the companies that say just do operating systems, you’re right. We did a full operating system and hypervisor at Joyent. There’s a lot of kernel engineers that clearly – Microsoft, Google, Amazon as well. But a lot of the – they’re like me. You have to do operating system work to sort of go and do a cloud. I think it very often comes down to just pool of talent, because at the end of the day people innovate companies don’t innovate. People sell to people, companies don’t sell to people. A company is only as good as the people that’s in it. So for example, I can’t imagine how like old engineering at Joyent would have existed if some micro systems didn’t exist. Because almost everyone there is basically former Sun, former Google, former Amazon. You have to have that sort of workforce mobility and what people are doing like that.

Jason Hoffman 12:53

Then at the same time if you look at that I think like that sort of approach or commenting earlier around data services really being sort of the next thing that’s driving really the cloud adoption. Clearly a very large player like SAP here in Europe can be potentially well positioned to take those types of things. Say, “Okay. Well here’s a panel of actual data offerings and we’ve do the data and we’ll do this and we’ll do that. Well sort of move from there.” That’s probably the likeness.

Om Malik 13:28

Part of Nokia and [inaudible] given that boat actually focus on the next generation of the network, right?

Jason Hoffman 13:36


Om Malik 13:36

Which is the mobile network and mobile applications. Do you think they should be thinking about offering mobile specialty cloud services or cloud platform?

Jason Hoffman 13:45

Yeah. Absolutely. If you look at something–the opportunity for say network equipment vendors that do network services would be then–instead of talking about servers and storage we can say normally we’d do network equipment that we do as a network service and then we do compute equipment that we do as a compute service. And we do data equipment that we do data services on instead of calling it storage and servers. So potentially it’s expanding from network equipment to compute equipment to data equipment, then running those types things as services. It is definitely something those guys can move into, namely because they’re actually pretty familiar with running large scale services. It’s been on the mobile side. The application phenomena is a mobile app phenomena and as far as I can tell no one’s actually put say an infrastructure as a service or a Pass on a mobile network. I can’t exactly buy a VN that’s got an IP address from a mobile network on it.

Om Malik 14:44

I think Facebook is trying to do that when they bought Parse and try to build out some kind of infrastructure play there. What about you? What are you doing to do now? Are you going to start another cloud company?

Jason Hoffman 14:57

No, probably not.

Om Malik 14:59

Why not?

Jason Hoffman 15:00

You know how this goes. It’s like when you–we sort of I came to the end of what was a big product launch and a three year road map. You do something for ten years and feel like doing something else. I actually don’t sort of know exactly what that is yet. I am continuously fascinated by this mobile machine to machine the wearable type stuff. You toss all this big data conversation in there as well to use another catch phrase. All those things actually potentially have–

Om Malik 15:37

That can get you funded, right? That’s a good PowerPoint right there.

Jason Hoffman 15:41

Yeah. But all of those I think in particular if you look at–those have a very high potential of impacting say human health for example.

Om Malik 15:49

So would there be–are you talking about analytical, analytical tools or are you talking about actual products? When you think about this emerging world where there is many buildings of connected devices. What is the real cloud or infrastructure opportunity in your mind?

Jason Hoffman 16:06

Well just as something that’s simple, if you started looking at connected cars or wearables that now even become a medical device. So that the human being is now dependent on that device being connected. Or you don’t want to all of a sudden a cardiologist to think that all of his patients died. When you start thinking around that, there can be a simple things as–well if you go to the AT&T park at San Francisco and you go to a Giants games everyone’s tweeting and taking pictures and everything like that. You don’t want to drive a connected car by a busy baseball game, or have your pacemaker send stuff. And just because you went down a very congested mobile network your car stops working and your pacemaker’s doesn’t send telemetry. So when you start looking around the infrastructure there it becomes, how do you scale the combination of that connectivity and everything like that? And make it so that these actually now have devices connected to them that you’re entirely dependent on.

Om Malik 17:10

So you’re saying there needs to be a software layer, an intelligence layer in between the network, the hardware and the end customers?

Jason Hoffman 17:20

Yeah. If you have a self-driving car and it has to be on the network and all of a sudden everyone tweets and Instagrams it’s be bad if all of those cars stopped working.

Om Malik 17:30

I think one of the key things I think about is how does this impact our infrastructure needs? Because the first version of cloud – the first five years – since 2008 to today has been about certain kind of automation and industrialization. But next five years are about connecting a billion devices or more. Just the sheer stress you put on the infrastructure not to mention the network is at a scale we don’t even know how to project.

Jason Hoffman 18:07

They have to be. The current operating systems, run time, middle ware, hardware they actually have to be a billion times better than they are today.

Om Malik 18:18

So it seems like there’s a lot of opportunities still left–

Jason Hoffman 18:21

There’s a ton.

Om Malik 18:21

–in cloud.

Jason Hoffman 18:22

If anyone says there’s never opportunity their lying–

Om Malik 18:23

No I’m just saying.

Jason Hoffman 18:27

–there’s always.

Om Malik 18:26

You said maybe no cloud company again, maybe you need to take six months off come back and start another company. And come and tell us next year what you’re up to.

Jason Hoffman 18:34


Om Malik 18:34

Okay. Thank you Jason.

Jason Hoffman 18:36

Thanks Om.