The question is, "Why buy, when you can build?"
Let's say I'm Rackspace. I have a cloudkick account that I used when I was evaluating them to see if I wanted to buy them. I know exactly what they offer.
So, let's say I take that and I create an engineering specification. Build a service that does this, this, this, and this, duplicating cloudkick, right down to the screen layout of the dashboards, if you're so inclined.
Now, go hire 10 engineers. No, make it 20. Now hire 5 project managers. Pay each engineer $100K + benefits, say they cost me $150K per year. Pay each project manager $150K + benefits, say they cost me $200K per year. So now I have 20 engineers and 5 project managers whose combined salaries + benefits cost me $4 million a year. Let them work for two years. For $8MM, I have my own cloudkick, right?
Why would I pay more than $8MM? Perhaps they didn't, I don't know. Again, my questions isn't about cloudkick. It's about why you see companies make the buy decision, when it often seems like it would be well within their ability to build it themselves, if they are so willing to spend money.
(2) Building (obviously) takes more time; the time you spend building is time-to-market that you're conceding to competitors. When you buy a competitor, that gap is erased; you get the benefit of their earlier time-to-market.
(3) Building comes with enormous risks. Deciding to build a cloud monitoring platform is not the same thing as executing on a cloud monitoring platform that the market will embrace. It may in fact be more likely that you will, on some important but perhaps subtle axis, fail utterly. Now you've made a huge investment in your target market and gotten negative returns.
(4) Acquirees are proven in the market. If an acquiree sucks, you buy one of their competitors (unless they suck on an axis you can readily remediate, in which case you probably get a big discount).
(5) Acquirees come with customer traction.
Fundamentally, and quite succinctly, you've explained why building a successful business/project/technology is hard. And because it's so hard, those that have made something successful and with traction have something of value.
Renting you let someone else maintain the property and can walk away at any time. This is what CloudKick's current customers all do when they consume the cloud service.
Buying means gaining control. You build equity and have total control over all the maintenance and improvements. And you have all the knowledge that hindsight offers. Is the house in a thriving neighborhood? Was it built well? Did the previous owners take good care of it?
Building means what? Total control over the entire job, from finding architects and contractors, having say in every detail of the design, the budget, bill of materials, schedule, etc.
But the amount of risk is massive here. If funding takes a hit, or budget blows up, what do you do?
Back to software, more projects die in incubation than mature to CloudKick status, no matter how mature the software companies are. Think really hard about how fragile software development practices are.
And there's further risk in the market. If you set out to build your own CloudKick style infrastructure, CloudKick is still out there furiously innovating and battle testing their software. It can be difficult or impossible to catch up.
You are proposing to spend $8MM to build the cloudkick of today in 2 years time. That might be ok if you're just worried about internal tools, but that's no good if you want to offer the service to other providers.
Assembling all the pieces that make a product does not make a product. There are lots of risks involved in creating new products and generally as companies get larger they become more risk-averse.
Even assembling the pieces involves jumping hurdles in any significantly large company (ie. various levels of approval, staffing, inertia). Acquisitions avoid product development risks and the acquirer also get a group of people that demonstrated they can create a successful product.
There you answered your question. Cloudkick's already worked on the problem for two years.
Time matters, it's all we've got to sell.
The other reason you sometimes pay a premium is to get them off the table for a competitor. So even if you could build this solution in two years for 8MM. What if your big competitor decided to come buy them and then starts expanding their business. In two years you come out and say, "Hey look, we have that product from a couple of years ago!"
It doesn't mean of course that buying a company is always the best decision.. it boils down to whether the acquired product and team ends up being a good fit.
I can see Rackspace extracting the monitoring features/code from cloudkick and add it to their cloud servers control panel, and leave CK going as it's own entity. That's pretty much what they did with Slicehost iirc.
Sometimes the outside party just built a better solution, and it makes sense to get a whole company, lots of good staff, good PR and new products in one simple purchase.
As long as the return on investment is as-good or better than the ROI of building it yourself, then it's a win in terms of time factor alone.
Consider how many different people/organizations were trying to accomplish what cloudkick has? You're cutting a check to get the best version of all their efforts.
Additionally, Rackspace has become fairly large, I'm sure there are sufficient politics in that organizaiton to force any in-house app right off the rails. The marketing manager would want the app to only measure all the places where rackspace wins. Any top shelf talent would be poached by the orgs core business units etc... Even starting with the winner in this area, they may still fall behind.
Isn't that...illegal?
SL gets margins by being efficient whereas Rackspace gets margins by having an incredibly large and efficient sales team. SL is a hacker's hosting company which has made very solid progress selling to the enterprise.
To top that off, Salesforce just bought Heroku, telling the world that they really intend to compete in cloud hosting against Amazon, and any other clouds that matter (do any other clouds matter?). The competition is fierce, the stakes are high, and we're seeing an arms race. I wouldn't be surprised if HP got into the mix as well (there must be some reason they bought loudcloud and EDS beyond just rounding out their EYP division).
Any start-up who can help some of the big dedicated server firms (Softlayer, Rackspace, Peer1) either appear to be a sexy alternative to Amazon (to try and woo new startups, which is where all of the dedicated server companies are sort of lackluster, primarily due to a lack of advertising and a lack of a presence in the valley) has a high potential for acquisition in the next year. Look at the past three years of operations and cloud related start-ups. For example I'll be really surprised if either Puppet or Opscode are still independent entities in 2012.
I don't think so, not directly anyway. Heroku is platform-as-a-service (applications), Amazon is infrastructure.
This is always a problem when a company is purchased by one of the vendors it sits above in the value chain, esp when impartiality and independence is important in a space such as vendor monitoring.
I can't see RS buying CloudKick for internal monitoring of its own customer's servers and instances because they would already have a mature solution by now.
I'm not sure I want to be providing RackSpace with performance data of my servers running on their competitor's instances, the CloudKick agent is placed to send (albeit transparently) all sorts of data back to RS. And don't be surprised if other vendors raise concerns to this effect too.
I just want to respond to one point.
> I can't see RS buying CloudKick for internal
> monitoring of its own customer's servers and
> instances because they would already have a
> mature solution by now.
(that's not a disrespect to CloudKick, but an examination of scales of economy)
But whatever. If this is about bolstering RS's own monitoring does that mean CloudKick will no longer be monitoring 3rd party vendors?
I'm guessing CloudKick will still continue to be more than happy to monitor Amazon EC2, Linode, GoGrid, etc out of the box.
And so even if this acquisition is going to help RackSpace's own monitoring, the point still stands that RS now has a massive competitive intelligence operation going on here too. If the data is available inside the company then it even has a duty to shareholders, as a publicly listed company, to make the most out of that data to enhance profits.
This is why Cisco acquires companies started by former Cisco employees on a regular basis that some folks have been acquired into Cisco 3 or 4 times in their careers!
Also, the competitive intelligence is not as useful as it may seem. Data for the sake of data isn't valuable. And when you figure out what they could turn it into, it doesn't seem like it would show much more than economic and market forces already show, even if CK got to scale.
For RackSpace, monitoring is a cost center and not a core competency.
"If the data is available inside the company then it even has a duty to shareholders, as a publicly listed company, to make the most out of that data to enhance profits."
This cliche, like most cliches, is not only annoying, it is wrong:
http://crookedtimber.org/2008/07/25/what-obligation-maximise...
I think they would have grown much faster with a smoother price curve. Make a plan with 2 or 3 servers of full monitoring for $30 a month to grab the low end of the market when people are just starting and putting in the foundational pieces of their technology stack.
Overall though, they're really good. They seem to have done plenty well anyway. Congratulations.
I wanted very basic things (namely application event monitoring for fraud), so I built my own analytics package around cron jobs.
Maybe I'm mistaken but I don't see how it is even remotely similar to what these monitoring services provide. Can you explain that better?
If it runs on your server then how does it monitor my application? Is there an agent or something? Why is it better than running cron on (say) Google App Engine?
If it runs on my server, then exactly why would I want it? I think I can live with running every 5 minutes or 20 seconds instead of 7 minutes or 17 seconds.
So.. I just don't get it? Where can I read more (without registering)?
In that price range you have Pingdom and AlertFox...
Plus there is always Pingdom, AlertFox, Webmetrics, Keynote...
For some background Jungle Disk still supports multiple clouds as well.
I feel bad for Cloudkick as a company, though. San Antonio is a million miles away from San Francisco, culturally. Good job on having an exit, good luck on not hating yourselves in a year.
Now back to work as i wait patiently from my turn. :)