You are currently browsing the tag archive for the ‘Amazon’ tag.
I was reading the description for a cloud educational session sponsored by a local development company when I came across this copy:
IT Leaders: You’d prefer to focus on critical data and security issues without the hassle of servers, software licenses, installation, and maintenance; and ultimately you want to be able to scale your infrastructure dynamically as the need arises. The cloud can give you that freedom.
I have to believe this organization was referring to Platform-as-a-Service (PaaS) clouds like Google Apps, Microsoft Azure and Salesforce Sites vs. Infrastructure-as-a-Service (IaaS) clouds like Amazon EC2 or Rackspace Cloud servers. Because IaaS clouds require a certain level of software installation and system management knowledge. This begs the question: Will IT leaders eventually favor PaaS cloud solutions over IaaS because of perceived management efficiencies? I don’t think so and here’s why:
1. PaaS is the ultimate form of vendor lock-in. If I create a Google application I won’t be able to port it over to Salesforce or Microsoft. I will never be able to walk away from my cloud provider once I’m fully invested.
2. PaaS can’t run 99.999% of the existing software on the planet. My investments in legacy applications and custom code are negated by PaaS clouds. Why would my company rebuild perfectly reliable IT solutions just so that they can work the way a particular PaaS cloud provider wants them to work?
3. My IT resources and expertise will cease to be a competitive advantage if everything runs in a PaaS cloud. And I will only be able to utilize the capabilities that my PaaS platform vendor offers me.
4. PaaS cloud security will never be as transparent as IaaS cloud security. One minute Google is showing your secret documents to the whole world and the next minute they aren’t. All my data is one username and password away from being compromised.
You might get the impression that I don’t like PaaS cloud architectures. Nothing of the sort. I think the PaaS model makes an awful lot of sense for SaaS-based startups and social applications. I don’t think it makes sense for most enterprise IT applications. Tell me why I’m wrong. 
Several big announcements rained down from the Internet clouds over the past few days. Amazon announced support for Microsoft Windows in their EC2 cloud. They will support both the Windows OS (AD, IIS, .NET, etc) and SQL Server. Additionally they are providing a 99.95% uptime SLA for their services. This is a respectable goal for a cloud hosting service.
Microsoft, not to be outdone, announced their own cloud computing offering called Azure. We all knew that Microsoft was working on a cloud hosting service over the past year. It’s nice to finally hear some of the details. Microsoft’s service is initially geared towards developers but it’s not hard to see that small businesses are the next likely target. Microsoft provides additional support in the cloud for technologies like Microsoft Dynamics, Sharepoint, and Live.
Someone recently asked me the question: “Are clouds going to compete with dedicated hosting services?” My response was both yes and no. Cloud hosting from Amazon and Microsoft will definitely compete with unmanaged dedicated hosters like The Planet and SoftLayer. Cloud hosting is nothing more than virtualized Platform-as-a-Service (PaaS). I also think cloud hosting will directly compete with smaller datacenter colocation deals. Think about it. If you are a company looking to colo half-a-dozen servers at a datacenter you could save significant capex and infrastructure headaches by using virtual servers in a cloud. Plus you could take advantage of the global business continuity features of a large cloud.
I don’t believe clouds will compete with managed hosting providers — yet. The simple reason is that cloud hosters don’t provide server management. They are purely infrastructure plays with some cool APIs. I’m sure some smart group of people is thinking about building a managed services business on top of a cloud platform. You just need the technical talent, marketing, and luck.
Every day a new vendor or service provider touts their new cloud computing technology and strategy. But what is cloud computing? It is quickly becoming one of those catch phrases which hosting marketing groups morph into every product pitch.
The origins of cloud computing stem from grid computing. You generally find grid computing in academic institutions and research laboratories. You’ve probably read articles about schools that build supercomputers using hundreds of off-the-shelf desktop machines. They use grid computing architectures and technologies to tie all those computers together. Think Borg. You can’t install Windows XP on this sort of massive cluster. These types of systems only support custom applications that are designed to take advantage of the distributed computing architecture.
Cloud computing incorporates the horizontal scalability of grid computing and packages it into a service that is delivered from a datacenter. The idea is that you can install your applications in a datacenter computing environment with little regard to the underlying infrastructure. Your applications live in the cloud. Cloud computing environments generally utilize virtual servers, networked storage, and significant software automation. Amazon’s EC2 is a good example of a modern cloud computing strategy. Their hosting service allows users to load applications on virtual servers and then scale those virtual servers — even across datacenters — to meet increased performance requirements. Google’s App Engine is another example of a cloud computing strategy. Do users really know where their applications live within Google’s infrastructure? No. And it doesn’t really matter. Google’s massively distributed infrastructure ensures that applications survive and thrive within the cloud.
The biggest challenge with cloud computing is that the specialized architecture requires specialized application development. It requires re-engineering time that most businesses don’t want to invest in. Many of Amazon’s bigger EC2 customers are companies that have strategically focused their application development on Amazon’s platform.
I believe that cloud computing will one day replace common hosting services such as shared websites, virtual private servers, and dedicated servers. But it will take a few more years before we see a standardized platform that both hosting providers and businesses can support.
During the conference yesterday I asked Phil Soran, CEO of Compellent, if the up-and-coming crop of Internet storage services like Amazon S3 are a competitive threat to his business.
His answer was two parts: 1) Compellent sees storage services like S3 as potential customers, and 2) Compellent focuses mainly on customers that need higher-end storage.
I understand his response but I respectfully disagree with some of his thinking.
First, storage services probably are not going to buy Compellent storage. It’s simply too expensive on a $/GB basis. It scales vertically pretty efficiently (by adding disk shelves) but not horizontally. If I were building a storage service I would look for the cheapest disk I could find. Then I would find a way to make it scale (virtualization, load balancing, mirrored storage, etc).
Second, it’s true that many companies need the storage performance that a company like Compellent can provide. Sadly services like S3 will still eat into Compellent’s growth — and their own story proves it. Compellent’s big product claim is their automated tiered storage. Basically a customer can install fast fiber-channel drives and slower sata drives in the same SAN. The SAN will then migrate less frequently accessed data blocks to the slower drives — keeping the faster drives free to store more high-performance data. It’s a smart concept. The problem is what happens when companies start replacing those sata drives with S3. Or, what if S3 becomes the third, and lowest, tier of storage. People just won’t need to buy as many new drives and shelves.
Sure, but companies will always need faster storage right? Well, not if the SaaS companies have something to say about it. My customers aren’t buying storage for Microsoft Exchange and Sharepoint anymore because I provide it at my datacenter. My customers aren’t going to buy storage for Microsoft CRM. Eventually they will be outsourcing their ERP and business intelligence applications to SaaS providers. And guess what? Some of those young SaaS providers are starting to build their platforms on top of the online storage services.
Phil made a good point that many companies don’t want to store confidential information on “shared” storage services. I agree that this is an impediment to the growth of the Internet storage platforms. But this is definitely starting to change — especially with SMB. The larger companies will follow suit as they outsource more of their applications to SaaS providers.
The data storage business is amazing.
It sort of reminds me of Churchill’s quote… how did that go… “never was so much owed by so many to so few”. We keep giving our blood sweat and tears to a handful of enterprise storage vendors. We are served the same old commodity hardware and complicated software. Let’s face it. Almost everyone is selling the same disks on the same chassis and touting the same features.
The storage sales guys by-and-large are marketing droids that don’t understand a damn thing about what I need and probably wouldn’t care if they did. Oh, I’m not saying they are bad people. They have to earn a living like everyone else. My beef is that they always have an agenda that includes finding my wallet and removing my money. When they tell me that they want to meet to better understand my infrastructure it feels like they are just snooping around for more business. If I have a real storage project believe me — I’ll find you.
My company primarily buys netapp storage. Most of the netapp sales guys are okay — certainly a little better than some vendors I’ve dealt with. The gear works. That’s certainly a positive. On the other hand it is really expensive. It is a limiting factor for growth in our business. It harms our competitiveness. Netapp wants to license every cock-a-mamy protocol and service separately. Need iscsi? cha-ching. Dedup? cha-ching. Oh, now you need replication? cha-ching cha-ching. You won’t know the true cost of your netapp infrastructure until a year down the road when you need to start adding expensive features. And here’s the real rub. You end up paying about 20% annually for service and maintenance on the hardware. So basically over five years you have just paid 2X for your storage. It’s such a great scam I’m kicking myself for not inventing a storage company.
I want google storage. No, not some kind of remote data storage service like Amazon S3. I need more performance than those storage services can offer. I want storage that scales like building blocks. I want to be able to plug in storage blocks and gain more space and more performance. I want the storage scaling to be automatic and dummy-proof. I don’t want a solution that simply mirrors raid10 disk on vaultA to raid10 disk on vaultB. That’s trivial and a waste of resources. I want true distributed storage with outstanding performance. I want redundancy if building block nodes fail. I want to replicate data to multiple datacenters. I don’t want to license separate features. Give me the whole enchilada. Netapp, if someone else steps up to the plate and is able to provide me with this type of storage solution, it’s… buh-bye.
