You are currently browsing the category archive for the ‘Hosting’ category.
I hear about vendors and service providers talking about virtualization, cloud computing, grid computing, and utility computing on a daily basis. Now throw in terms like unified computing from vendors like Cisco and you have a recipe for confusion in the marketplace. Staff members at my own company sometimes mix up terms like grid computing and cloud computing. And many marketing departments from traditional service providers take great liberties with these terms.
Virtualization is not cloud computing.
I hate to break it to ya but virtualization isn’t cloud computing. Virtualization is nothing more than a brick in the cloud computing house. It’s purely an infrastructure component. I don’t think it is necessarily even a requirement for a cloud computing infrastructure. Cloud computing abstracts computing resources from hardware from the perspective of the end user. You can’t just stand up a couple vmware servers and claim that you now have a cloud. I know vmware and some service providers would like you to believe that.
Grid computing is not cloud computing.
I’m probably going to upset some of the “grid hosting” providers out there when I say that their notion of a grid is marketechture at best. You can find true grid computing in academic circles and high-performance computing environments. Grid computing requires specialized architectures and purposefully written software applications that can run on distributed systems. You can’t run your Windows application on a grid. Think Beowulf clusters. Grid hosting marketechture is nothing more than a group of individual servers running hypervisors that are on the same vlan. These servers, and the computing resources they represent, are effectively silos.
ASP (application service provider) is not cloud computing.
Some vendors have related cloud computing to the ASP-model over the past year — usually in a dismissive tone. Their belief is that cloud computing is just another name for ASP 2.0. The big difference between cloud computing and the ASP model is that the cloud is designed to host a generic class of applications whereas the ASP model is purposely designed to handle specific applications. For example, the Amazon or Rackspace clouds can host an infinite set of applications. Whereas Microsoft Hosted Services are designed to provide rental of specific applications like Microsoft Exchange. You could say that Microsoft’s hosted services live in the “cloud”. But you would be hard pressed to say that this type of service makes Microsoft a cloud computing provider.
The Internet is not cloud computing.
I’ve seen this analogy creep into vendor-speak over the past year. Technical people traditionally referred to the Internet as the “cloud”. In fact many of us would literally draw a cloud when sketching out a network infrastructure connected to the Internet. So the thinking goes that if you connect a computer server to the Internet you now offer computing in the cloud — or cloud computing. This analogy ignores the business model and processes offered by cloud computing and focuses solely on infrastructure. See the “virtualization is not cloud computing” rant above. It is absolutely possible to build a cloud computing environment that is not connected to the Internet.
Utility computing could be cloud computing.
Utility computing is probably the closest traditional definition to what we call cloud computing today. Utility computing is defined as the packaging of computing resources as a metered service similar to a public utility. Utility computing allows the customer to rent resources versus paying for the acquisition of computing servers. This definition sounds awfully close to cloud computing. Cloud computing could be called utility computing 2.0. I’ll cover this more in the future.
I read the latest Daily T1R newsletter sent out by Tier1Research with interest this morning. Basically they were saying that traditional hosting companies need to carefully manage customer relationships and expectations as businesses become more aware of cloud computing.
I’ve talked about how utility pricing is a key advantage of cloud computing architectures. The challenge traditional hosters face is that as customers begin to understand the significant advantages of utility pricing they may seek new hosting vendors.
As an owner of a traditional hosting company I find myself in a tough position. I’ve had to build an infrastructure to service a diverse group of customers with ever changing needs. I know that many customers pay for resources they are not using. I’ve been able to lock in customer relationships for long periods of time using contracts. But customers have come to accept these constraints as part of the cost of doing business. Internet and telecommunications companies have relied on the oversubscription model and service contracts to propel growth for decades.
Now cloud computing is forcing service providers to reevaluate not only technology architectures but business models. At the same time we are dealing with increasingly savy business users that understand the cloud. How do we successfully transition from the old model of hosting to the new? How do we maintain margin much less propel growth in a world where the customer only pays for resources they are consuming?
I think part of the answer lies in the efficiency gains provided by a cloud computing architecture. Service providers will be able to leverage the computing resources of a cloud architecture more efficiently than racks of individual servers. Providers will also be able to provision and manage services more efficiently leading to labor cost savings.
According to this article change management is a pretty interesting ride at Google.
Making changes to Google’s search infrastructure is akin to “changing the tires on a car while you’re going at 60 down the freeway.” –Urs Holzle, Google
Mosso, a division of Rackspace, launched their new Cloud Server hosting service today. I’ve been following the chatter on this service development over the past 6 months.
When I read Tier1Research’s report on the launch last Friday I had to do a double-take. They reported that Cloud Server pricing started at $0.015/hour — or $10.95/mo. I thought that was clearly a mistake until I checked out the pricing list on the Mosso website. I don’t know how the heck they can price a virtual server that low — even though it is basically unmanaged. My guess is that they are expecting most customers to quickly ramp up into higher pricing tiers.
I tried to compare Mosso’s Cloud Server pricing to Amazon’s EC2 service. It’s not an apples-to-apples comparison because of the subtle differences in pricing models.
Many businesses are paying more for shared hosting than they would pay for a Mosso Cloud Server. If I was a website developer hosting sites for clients I would be ecstatic right now. The plethora of cheap hosting infrastructure offerings on the market is amazing. I’ll take it one step further. If I was starting a traditional mass-market hosting company today I wouldn’t buy a datacenter or servers. I would totally leverage existing infrastructure in the cloud. What a great time to invent new business models!
Disclosure: I am a Rackspace (RAX) shareholder.
Google’s Gmail service crashed earlier this morning impacting users around the world. It sounds like people were able to access their Gmail accounts through traditional mailbox protocols like POP and IMAP, but not via the web.
The sheer fact that this is national news just underscores the scale of Gmail and other global email services. I actually think that Gmail is pretty reliable for a free (or very cheap) email service. In my experience it has better uptime than most enterprise-class Microsoft Exchange email service providers.
Cloud hosting infrastructures have a number of unique advantages over traditional dedicated infrastructures. One of those advantages, a key part of any cloud infrastructure model, is the utility pricing model. Here’s a key point to understand: the customer always overpays for infrastructure and never underpays when using traditional hosting. Traditional hosters oversubscribe infrastructure such as bandwidth and storage. Customers pay for bandwidth and storage reservations as part of their selected hosting package. They also pay for any overages beyond their infrastructure reservations. In other words, a customer pays for a reserved amount of bandwidth or storage whether or not they are actually using that infrastructure. In the cloud model the customer only pays for the infrastructure they consume. They can scale the infrastructure up or down as the needs dictate. In many cases the infrastructure unit pricing goes down as the customer’s needs scale up.
Negotiating contracts for datacenter services is a tricky business. I’ve been fortunate to be on both the buying and the selling side of the datacenter business over the past ten years. During this time I have gained some insights into some of the “gotchas” in the ways contracts are written. Let me share some of these insights with you today.
Autorenewing contract terms
Most datacenter service contracts incorporate autorenewing contract terms. The customer must notify the datacenter provider that they intend to cancel the service within a specified amount of time before the contract ends. If the customer doesn’t notify the datacenter provider the contract is automatically extended a certain length of time. For example, a contract could state that the customer must notify the datacenter provider within 60 days of the contract expiration otherwise the contract is automatically extended one more year. Datacenter providers love this type of contract language because they know that most customers forget about this notification requirement. As a buyer I would flip this requirement and make the datacenter provider notify me before the contract expiration. If they fail to notify me then the contract goes month-to-month.
1 year or 3 year deal?
Most managed service providers want their customers to sign multi-year contracts. The fact is that most providers don’t make any profit the first 6-12 months of the deal. The datacenter provider is basically supplying all the upfront capital and betting that the customer will stick around a few years. That’s one of the reasons that most providers offer multi-year discount pricing — i.e., 10% discount for signing a 3-year deal. So should a customer sign a 1 year or a 3 year contact? I think it depends on the type of service. If we are talking pure colo and your organization doesn’t plan to move for a couple years and you don’t like renegotiating contracts then I think a multi-year contract makes sense. If you are looking at managed services then a cost-benefit analysis is required. You need to understand that if you sign a multi-year contract your equipment will not be refreshed for the term of the contract. So while you are receiving a 10% discount in year 3 you will be using equipment that is over 2 years old. As a buyer I would rather sign a one-year contract and refresh my equipment on an annual basis. This makes even more sense if my infrastructure is completely virtualized. Because every year I would upgrade the performance and capacity of my services at little or no additional OPEX.
Co-terminus contract lengths
Here’s a very common gotcha in the service provider business. A customer signs a 2-year colo contract. After one year they decide they need more power. The customer then signs a 2-year power agreement. What they don’t realize is that they just extended their original colo contract for one more year. Service providers conveniently forget to tell their customers about that contract extension provision. As a buyer I would ensure that the contract length for any additional services I buy from the service provider expire at the same time as existing services — co-terminus contracts. So the customer that signed the power contract should have signed a 1-year agreement that terminated at the same time as their original colo agreement.
America was founded by people willing to take a chance. They were men and women that risked everything for a better opportunity. To fail meant nothing less than their lives. The ones that survived were rewarded with new opportunities and new found freedom.
Today we live in a country where some companies are too big to fail. We live in a country where banks are bailed out before businesses. We live in a country where politicians manipulate financial controls to win votes.
Our children’s children will have more debt than any beings on this planet. They will be forever indebted to the failures of our generation and the generation before us.
I came across an interesting article in Data Center Knowledge about a device known as a heat wheel, or Kyoto Cooling, that takes advantage of outside air to cool a datacenter. I’m pretty familiar with the concept of air-side economization but I had never heard of this technology before. Sounds pretty cool. My datacenter location probably has the right climate profile to take advantage of this type of technology.
Well VISI pulled off a solid VISION 2008 technology conference yesterday in downtown St. Paul. Kudos’s go out to the VISI marketing team for the event organization and the respectable attendance. I don’t know the final attendance count but my guess is it landed in the 150-200 range. The food was good and the networking opportunities were plentiful.
I attended two sessions: Jim Akers from NetApp talking about next generation storage and Misha Govshteyn from Alert Logic talking about security. I’ve known Jim almost my whole career starting back in the old ISP days when he worked at Cisco. I found his discussion of storage virtualization and clones particularly interesting. Misha is the CTO and Co-Founder of Alert Logic and I thought he gave a very high-level, strategic overview of current security threats and defense strategies. His presentation was particularly refreshing since I’m so used to hearing thinly veiled vendor pitches during these types of presentations. I was very interested to learn how worms are becoming more stealthy. A few years ago worms were used as a way to damage computers and now they are used to extract financially valuable information.
I also had the opportunity to meet up with several prominent local bloggers and entrepreneurs including Graeme Thickens, Steve Borsch and Steve Kickert. Steve’s company, Riverock Technologies, recently launched a new team collaboration and productivity solution called OnePlace. The SaaS-based solution is written in Ruby-on-Rails and hosted at Engine Yard.
UPDATE: Graeme Thickens wrote a wonderful blog article detailing Robert Steven’s keynote speech at VISION 2008.
UPDATE2: Steve Brosch wrote an article in Minnov8 describing his experience at VISION 2008.
