Loudcloud: Early light on cloud computing
Editors' note: This is a guest post.
Every time that I see an article touting how great cloud computing is, I always chuckle and think to myself, "been there, done that."
Those who remember the emergence of the Internet era as a mainstream venue (circa 1995 to 2000) may also recall a company called Loudcloud, founded by Netscape pioneer Marc Andreessen. It is my opinion that you can trace the road that led to the current cloud-computing era back to Loudcloud's founding.
It started in 1999, when four visionaries who met while at Netscape--Marc Andreessen, Ben Horwitz, Tim Howes, and In Sik Rhee--saw a pressing problem facing Web-based start-ups. If these emerging companies wanted to establish a presence on the Internet, they were forced to buy a lot of expensive hardware, diverting precious resources that they otherwise might have been able to invest in their core businesses.
Loudcloud thus started with the vision of various "clouds"--mail, database, network, application server clouds, etc.--so that any enterprise could fractionally rent out what it needed. Customers would pay based on what they rented and for how long.
The formula was an instant success in a valley full of impatient Internet start-ups waiting to show what they could do. And all that changed when the dot-com bubble burst, and the names of many of the companies that had been hosted by Loudcloud started appearing in the pages of F***edCompany.com.
Being an employee at Loudcloud at that time gave me a ringside view of the company, as we tried to aggressively reinvent ourselves toward selling the concept of managed infrastructure to the bigger enterprises when the Internet economy collapsed. However, the idea met quite a bit of resistance from companies questioning the effectiveness of security controls in a managed infrastructure environment.
This was not a valid reason then and is not the valid reason now. These enterprises resisted moving to a cloud environment out of fear that they would lose control over the data that was sacrosanct to them. Interestingly, many of these same customers were more interested in learning about the automation platform that ran our multiple data centers across Europe and North America.
In 2002, Loudcloud's founders decided that it was in the best interest of the shareholders and the company's longevity to jettison the managed-services business (fixed equipment costs played a huge part in this) and to move toward producing a data center automation software platform that would help enterprise customers run their own data centers efficiently. So they spun out a separate software company, Opsware, which Hewlett-Packard bought in 2007 for about $1.6 billion.
Other cloud-computing companies can learn a couple of valuable lessons from Loudcloud's example.
Convincing an enterprise customer to let you host its crown jewels (business data) is going to be the last thing you should attempt. Instead, focus on applications that are important to an enterprise but not in their "critical" path--as the first wave of cloud-computing adoption has shown us (such as with Salesforce.com and NetSuite).
Be prepared to adapt and adopt. Loudcloud survived--and then thrived--as Opsware because it was able to refine its message to enterprise customers when the bubble burst.
Finally, cloud-computing companies have one thing that is going for them that Loudcloud didn't have nine years ago: enterprises are now more "comfortable" with the concept of the Internet and the maturity of Web-enabled technologies, and that has made them more open and receptive to taking advantage of the flexibility, speed, and agility that cloud computing offers. So much so that the newly anointed federal chief information officer, Vivek Kundra, has embarked on a cloud-computing strategy for the federal government.
Cloud computing is getting its due these days, but let's remember that Loudcloud first proved the viability of the concept.
Girish Venkat is a sales consulting manager for Oracle. He is a 16-year industry veteran of six start-ups, including Loudcloud. He is not an employee of CNET, nor is he a member of the CNET Blog Network. 
You point out a very good observation in commercial world cloud computing, but in government it would make sense as one could buy hardware and software centrally and allow business users compute their data over the intranet. Leads to savings and removes silos of data centers.
Regards,
Jay
I work with some of the largest financial institutions in the world and they *all* host some -- if not most -- of their "crown jewel" data with 3rd party providers. Not just for storage but for processing as well. E.g. IBM runs numerous hosted data centers for banks, brokerages, and insurance companies around the world. So does HP (+ EDS).
What do you think companies like ADP have been doing for all these years? Enterprises trust them to process ultra-sensitive information, including employee HR records, payroll data, electronic funds transfers, etc.
I worked with one of the largest brokerage firms in the world, and guess what, their "book of records" is outsourced to a financial service provider. It doesn't get more sensitive than that, at least in the commercial world. Ask your brokerage firm where they hold their data... you might be surprised with the answer (if they give you an honest answer.)
Even in the government sector, I know a major government PKI infrastructure (non-military) that's outsourced to an outside service provider. We're talking hosting PKI certificates that's the key to government computing facilities. Needless to say everyone in the project had to hold appropriate security clearance.
The cloud as a concept is NOT new to enterprises. They've been doing it for years. Of course they don't use "soap", "web services", etc. They do it using "old-school" technologies... mostly shipping jobs to be run on virtual partitions on massive mainframes hosted by the IBMs and the HPs of the world.
Enterprises host their data elsewhere, all the time. They just don't want to do it with puny companies like LoudCloud, for good reason. If IBM or HP or ADP f***k up, they have huge financial and other resources to make things right for the customer. LoudCloud had nothing to back them up. Marc Andreessen's entire net worth couldn't cover ONE DAY of failure liabilities.
Having worked at the company as a Project Manager and Project Engineer, my observation was that most of the customers were serviced through the implementation of custom silos of equipment which sometimes (not always) made use of shared infratsructure and tools, but not often (if ever) shared systems at the core of their infrastcructure. Some customers would make use of shared networking services (load balancing and firewalls, for example), but the major customers still worked with Loudcloud to deploy dedicated network equipment. All the web, app and database systems on the projects I saw and worked on (some of the bigger ones) were dedicated. Only the management and administration systems were shared (logging, bastion hosts, etc.)
That's not to say Loudcloud didn't do good work -- there were a lot of bright minds there -- but they mostly provided outsourced consulting and advanced hosting services to customes who either couldn't or didn't want to handle 24x7 application hosting and the associated operational tasks in house. FWIW, OpsWare, the "special sauce" was a valuable tool on Unix/Linux deployments, but it was almost complete vaporware for Windows deployments at that time -- maybe it got better before the OpsWare spin-off.
Lastly, what would a Loudcloud post be without a special shout out to In Sik Rhee, the useless executive who seemed to dislike Loudcloud as time went buy. At one point, I ran into a serious internal equipment ordering delay that was pi**ing off one of our largest customers and costing the company substantial revenues (we coudn't start charging the customer for our services until we had that equipment to take them live). The ordering staff responsible for the delay were under his direct control, but he refused to help and he told me, "F**k the customer -- from now on, we do what's best for Loudcloud."
That's when I knew it was time to go, so I left shortly thereafter.
-Mister Winky
I believe that these features along with our strong security protection (such as SSL encryption and folder privacy) may make AirSet a viable tool for businesses. If you can please take a moment to check out our service at www.airset.com and leave us your feedback.
- by katsnelson March 23, 2009 8:10 PM PDT
- As some of the comments point out, Cloud Computing did not start with the internet. When I started my IT career in 1985 Timesharing of mainframe resources has already reached its peek. Instead of Amazon there were companies like Control Data Corporation and others that charged you for the time you used on their "cloud" i.e. a mainframe.
- Reply to this comment
-
(7 Comments)Any time you have an expensive resource like IT infrastructure or a condo in Florida, you will have people who would rather rent than buy. And there will always be people who will just have an aversion to using a shared resource. They may use arguments like "security" to defend their position not to use a shared resource. And for some it is a legitimate reason but for many it is just an excuse. Reality is that most enterprises outsource and outsourcing is not much different from cloud deployment from security point of view. In most cases it is much riskier as the data center you are outsourcing to is likely to be less secure than those provided by the large cloud vendors like Amazon, IBM etc. Most of the private data centers are actually a lot less secure than what top tier cloud providers can offer. There are legitimate issues with physical access for audit and use of public network infrastructure for connectivity but these, with time, will be overcome with the changes in audit policies and broader use of VPN. Another important point is that most data breaches are not perpetrated by hackers but by insiders and there is no better way to keep insiders away from the data than to move the IT infrastructure away in to a third party facility like that provided by a cloud vendor. This is especially true when considering things like backup where today many enterprises back up to tape and then courier their tapes to an iron mountain for safekeeping. Much of the data loss occurs during the physical transfer of the tapes. Just Google on lost backup tape.
I am responsible for the strategy for IBM databases (DB2 and IDS) in the cloud, and security and resiliency are two topics that always come up in any cloud conversation. However, I do find that most people do change their stance once the facts are laid out for them and the methods for securing data in the cloud are clearly explained. So, in my mind, the enterprises will go cloud computing but it will take time and effort to educate. We've been trying to do our part by running webinars, offering free products and creating best practices type of documents. I try to put some of it on http://FreeDB2.com so check it out if you have a minute.