In previous posts, I've outlined the important, exciting, but oft-overlooked concept of "Intra-Cloud Computing." As more and more big players made HUGE capital investments in building out their own Clouds, presumably with the goal of attracting software projects, the question arises: What strategy should they use to build the "killer cloud" ?
Clearly there's a certain degree of infrastructure investment: Computers, networking, power, redundancy, security, etc. Arista Networks clearly sees this opportunity and is targetting Intra-Cloud Networking to help cloud builders gain an advantage on networking. I for one have experienced things like memcached performance being 2-3x slower in EC2 than it would be on a small cluster in a data center. If anything, we who build apps in clouds should expect them to have better network performance than a data center, not worse.
Related to my earlier post regarding AWS Start-up finalists and the importance (apparently to Amazon) of perception, I should re-phrase the question that drives the thesis of this blog: What strategy should cloud builders use to build the perception that theirs is "the killer cloud?"
The hardware, virtualization technology and associated tools, and the network are incredibly important to cloud builders. They must focus their to build a reputation of reliability, security, and performance. When it comes to reliability, the market is VERY sensitive to even one event, as we've seen with Amazon S3's outages. Just imagine the first major release of "private" information from a cloud and the associated fall-out in the market for the unlucky cloud builder. In these still early days of building out clouds, these vendors should stay focused on reliability, security, and performance. Also key to their perceived value is the customers' ability to provision machines on demand - and just as importantly, to de-provision when demand falls. Unlike other data center models, virtualization within "A Cloud" is of key importance. This bodes well for companies like RightScale and Elastra.
The services that AWS provides, e.g. Simple Storage Service (S3), Simple Queue Service (SQS), etc., are great value adds to the EC2 environment. These make certain things possible and/or easier that are critical to leveraging "A Cloud." What's interesting is that neither S3 nor SQS are true intra-cloud services. In fact, S3 pre-dated EC2 historically, and let's face it, for the types of files we use with S3, network performance and latency are not always all that important. Still, these services are less expensive and better performance when used from within the AWS cloud.
What Cloud Builders need to differentiate the completeness of their eco-system are true intra-cloud services. These services should be as easy to use as S3 or SQS, but should leverage something that didn't exist before: the reliability, low latency, bandwidth, and security that exists within a cloud.
Since Cloud Builders are so busy building their secure, reliable, scalable, easily-managed infrastructures, they should partner with software vendors who specialize in intra-cloud services - easy to use web services which allow software developers to focus on their core and outsource (cloud-source?) the rest.
Tuesday, November 11, 2008
AWS Start-up Challenge - Who got picked and why?
As I mentioned in my previous blog, those who are promoting the use of "A Cloud" should work on making their eco-systems superior to all others. To be more accurate, they would naturally attempt to improve the perception that their Cloud is better than all the rest.
Amazon chose 7 finalists, including:
Clearly it is cheaper with Cloud Computing, with AWS in particular, to start these businesses than it would have been a few years ago.
The area of Cloud Computing which is most over-looked, even by Amazon itself, are those business models which were not even conceivable without a cloud. I would call this area "Intra-Cloud Computing." Arista Networks' business model is clearly focused on intra-cloud, on making clouds better, faster, more reliable.
I guess a couple of businesses out there have hybrid intra-/inter-cloud computing solutions, ones that are primarily focused on the general "Hey don't worry, we're in The Cloud" (and happen to be in AWS), but also point out that some things work better or are cheaper if you as their client happen to be in the same cloud. For example, if you live in the AWS cloud, then both you and Encoding.com save money on inter-cloud bandwidth (AWS doesn't charge for intra-cloud bandwidth), and presumably Encoding.com would pass some of that savings on to you their customer. Likewise, Soasta does load testing on your web app, regardless of your web app's location, but if you also happen to be in the AWS cloud, you should save some money.
What I envision over the next few years are software businesses who focus on Intra-Cloud computing, who live and work in A Cloud and are targetting customers in that same cloud. Certainly it's not the case that such a business must live in only one cloud; they can have "cross-cloud" support the same way a browser works on PC and Mac and Linux. Still, just like most client-side software starts on PC, establishes a viable business, then expands to other platforms, it will be a trend for such intra-cloud businesses to get up and running on one cloud before supporting others.
Amazon would be wise to promote these types of businesses - are you listening Jeff Bezos? If Amazon somehow loses their vaunted status as the "PC of Cloud Computing," where truly intra-cloud apps start their life, then they will become the "Mac of Cloud Computing," with support for various capabilities coming later to their cloud instead of earlier. Alternatively, any operator of "A Cloud" should be working on making their eco-system as "complete" as possible, focusing on infrastructure, commodity pricing, exceptional networking, and a wide variety of value-added services that help I.T. projects hit the ground running.
Amazon chose 7 finalists, including:
- Encoding.com - on-demand video encoding
- MedCommons - all of a patient's information in one location
- Knewton - online education platform
- Pixily - what I used to call "scanmycrap.com," an idea I had in 1998
- Sonian - compliant archival
- Yieldex - online advertising analytics
- Zephyr - enterprise software testing sold as service
Clearly it is cheaper with Cloud Computing, with AWS in particular, to start these businesses than it would have been a few years ago.
The area of Cloud Computing which is most over-looked, even by Amazon itself, are those business models which were not even conceivable without a cloud. I would call this area "Intra-Cloud Computing." Arista Networks' business model is clearly focused on intra-cloud, on making clouds better, faster, more reliable.
I guess a couple of businesses out there have hybrid intra-/inter-cloud computing solutions, ones that are primarily focused on the general "Hey don't worry, we're in The Cloud" (and happen to be in AWS), but also point out that some things work better or are cheaper if you as their client happen to be in the same cloud. For example, if you live in the AWS cloud, then both you and Encoding.com save money on inter-cloud bandwidth (AWS doesn't charge for intra-cloud bandwidth), and presumably Encoding.com would pass some of that savings on to you their customer. Likewise, Soasta does load testing on your web app, regardless of your web app's location, but if you also happen to be in the AWS cloud, you should save some money.
What I envision over the next few years are software businesses who focus on Intra-Cloud computing, who live and work in A Cloud and are targetting customers in that same cloud. Certainly it's not the case that such a business must live in only one cloud; they can have "cross-cloud" support the same way a browser works on PC and Mac and Linux. Still, just like most client-side software starts on PC, establishes a viable business, then expands to other platforms, it will be a trend for such intra-cloud businesses to get up and running on one cloud before supporting others.
Amazon would be wise to promote these types of businesses - are you listening Jeff Bezos? If Amazon somehow loses their vaunted status as the "PC of Cloud Computing," where truly intra-cloud apps start their life, then they will become the "Mac of Cloud Computing," with support for various capabilities coming later to their cloud instead of earlier. Alternatively, any operator of "A Cloud" should be working on making their eco-system as "complete" as possible, focusing on infrastructure, commodity pricing, exceptional networking, and a wide variety of value-added services that help I.T. projects hit the ground running.
Monday, November 10, 2008
"The Cloud" vs. "A Cloud" vs. "Intra-Cloud"
There is a great deal of confusion about Cloud Computing. Let's shed some light on what's out there, what's new, and what is possible now that was never even conceivable before Cloud Computing.
First, we have "The Cloud" aka The World Wide Web aka The Internet. The idea here is simple and dates back to 1996: Some day, everyone will have such a fast internet connection, that they no longer need to store files on their local system. They can store everything, do everything online, leveraging servers. Someone else takes care of backups, software updates, etc. One of the first attempts at this was the Network Computer from Oracle. Ironic that Oracle's Larry Ellison is one of the biggest cloud computing bashers. CherryPal, founded in 2007 a decade later, does basically the same thing - leverages "The Cloud."
Next, we have "A Cloud" - and Amazon is the clear innovator here. The key differentiation between "A Cloud" and "The Cloud" is that unlike the Internet Cloud, Amazon controls EVERYTHING about "The Amazon Cloud." This means that within this one specific cloud, developers can expect a certain level of reliability, performance, latency, redundancy, security, etc. One startup that validates the importance of this "intra-cloud computing" is Arista Networks, which focuses on selling network hardware to companies like Amazon to improve their "intra-cloud" performance, latency, scalability, and reliability.
In the "A Cloud" category, we've had several MEGA players making big announcements about their multi-billion dollar investments: Sun, IBM, HP/Intel, Yahoo, Google, AT&T . What is clear is that each of these clouds will seek ways to differentiate to compete with each other (all chasing Amazon right now), to persuade new software engineering projects to begin their efforts on their cloud. They will seek to make their respective intra-cloud "Eco-systems" superior in terms of features and functionality, but also in terms of network performance, latency, reliability, bandwidth, etc.
So as a builder of "A Cloud," where would you focus first? Clearly the first focus should be on the network, the computing, flexibility, security - in a word, infrastructure. However, clearly there are services that would also make their eco-system compelling in terms of completeness. Amazon once again is the example here: Their Simple Storage Service (S3) and Simple Queue Service (SQS) are key factors to their success and leadership, in addition to the infrastructure of the Elastic Computing Cloud (EC2).
Therefore, there is a window of opportunity for "Intra-Cloud Service Providers", companies which provide "piped" services like SQS, perhaps leveraging the "A Cloud"'s infrastructure to reliably and elastically deliver these metered services.
First, we have "The Cloud" aka The World Wide Web aka The Internet. The idea here is simple and dates back to 1996: Some day, everyone will have such a fast internet connection, that they no longer need to store files on their local system. They can store everything, do everything online, leveraging servers. Someone else takes care of backups, software updates, etc. One of the first attempts at this was the Network Computer from Oracle. Ironic that Oracle's Larry Ellison is one of the biggest cloud computing bashers. CherryPal, founded in 2007 a decade later, does basically the same thing - leverages "The Cloud."
Next, we have "A Cloud" - and Amazon is the clear innovator here. The key differentiation between "A Cloud" and "The Cloud" is that unlike the Internet Cloud, Amazon controls EVERYTHING about "The Amazon Cloud." This means that within this one specific cloud, developers can expect a certain level of reliability, performance, latency, redundancy, security, etc. One startup that validates the importance of this "intra-cloud computing" is Arista Networks, which focuses on selling network hardware to companies like Amazon to improve their "intra-cloud" performance, latency, scalability, and reliability.
In the "A Cloud" category, we've had several MEGA players making big announcements about their multi-billion dollar investments: Sun, IBM, HP/Intel, Yahoo, Google, AT&T . What is clear is that each of these clouds will seek ways to differentiate to compete with each other (all chasing Amazon right now), to persuade new software engineering projects to begin their efforts on their cloud. They will seek to make their respective intra-cloud "Eco-systems" superior in terms of features and functionality, but also in terms of network performance, latency, reliability, bandwidth, etc.
So as a builder of "A Cloud," where would you focus first? Clearly the first focus should be on the network, the computing, flexibility, security - in a word, infrastructure. However, clearly there are services that would also make their eco-system compelling in terms of completeness. Amazon once again is the example here: Their Simple Storage Service (S3) and Simple Queue Service (SQS) are key factors to their success and leadership, in addition to the infrastructure of the Elastic Computing Cloud (EC2).
Therefore, there is a window of opportunity for "Intra-Cloud Service Providers", companies which provide "piped" services like SQS, perhaps leveraging the "A Cloud"'s infrastructure to reliably and elastically deliver these metered services.
Thursday, April 12, 2007
Outlook Extensions
This blog is dedicated to those who wish to learn to develop, test, and deploy Outlook Extensions.
First off: vbscript sux. msft sux. The outlook programming model REALLY sux. But outlook dominates the world we live in, so we adjust.
Here are some of the hardest lessons I've had to learn so far:
First off: vbscript sux. msft sux. The outlook programming model REALLY sux. But outlook dominates the world we live in, so we adjust.
Here are some of the hardest lessons I've had to learn so far:
- Outlook, and therefore the user's experience, is very precariously balanced upon each and every extension installed. There are lots of ways to start outlook, and lots of ways to shut it down. If you don't write your code for each and every one of these, you'll hork outlook. In fact, there are ways to start outlook wherein it is absolutely impossible for your extension to be properly notified and setup. And somehow this is your fault.
- Outlook is more than ready to blame your extension when things go wrong. They'll ask the user, "Hey, wanna disable this bad boy?" with your company name and logo everywhere.
- Unlike java, python, and every other non-prehistoric programming language, vbscript has no garbage collection (at least not in the outlook context). As far as I know, there's no "lint" for vbscript (if you find one, email me at dr.marc.byrd@gmail.com). And yet if you try to free something that's already been freed... - see numbers 1 and 2.
Subscribe to:
Comments (Atom)