Monday, December 28, 2009

Will latency kill Cloud Computing (like it did kill ASPs)?

Interesting comments from the TM Forum's Management World Americas event in Orlando (December 2009) around the problematic latency.

The real question here is about the networks and their development in the near future, as this is where the latency comes from (most of the times…). Being located in a remote office and working from my home office from time to time, I experience the issue of latency every day – in my business and private live.
I am an extensive user of Google’s Cloud services and many of our internal applications at Progress Software are also hosted on internal or external servers and made available via a web interface. While Google has many data centers all over the world (and one in Zurich, very close to me…), the applications located on internal or external servers are sometimes only accessible through the corporate intranet and often lack performance.

I am also using Salesforce.com which host their services at Equinix (they manage to run their whole IT on just 1000 servers!) but the performance (despite the local data center in Zurich) is rather average.

So will latency kill Cloud Computing??

I think the answer is NO! Because ‘real’ cloud computing and cloud services (where it doesn’t matter where they run) are actually the answer to the issue of latency and poor performance because the computing can happen very close to the user and network latency is reduced to a minimum.
Google is a very good example of how that can work out nicely. I never had major performance issues despite using Google services in various places, on different devices and sometimes with high data volume.

We just have to distinguish true Cloud Computing from hosted Web Applications that run on dedicated servers and do not utilize the real power of the Cloud (or Cloud Services).

Saturday, November 21, 2009

Chrome OS – the future of computing or just a niche?

image Google announced their Chrome OS on Thursday – which is basically an Operating System that relies on cloud services like Google Apps, Google Mail, YouTube, etc.

The news for me was that the OS will require a modified BIOS and thus support form the device manufactures (probably mainly netbooks). The time to the login screen was demoed with 7 seconds (3 more to get to the apps / browser). That is lightning fast compared to a Windows or Linux boot sequence, but since I never shutdown my laptops or desktops I have similar wakeup times from the Windows 7 sleep mode…

The idea of getting rid of all the unnecessary overhead of a full blown operating system and solely focus on the browser is a brilliant one, though. Google is the only one at the moment that actually has the power to build, spread and maintain both sides – the client side and the necessary cloud services.

With the ever increasing bandwidth available and ubiquitous internet access, more and more data will move into the cloud. Chrome OS is the consequent advancement of this movement and with the reservations regarding security and hosting private data somewhere in the cloud going away, I see this model becoming really popular in the mid-term.

Chrome OS actually might become the standard interface for many cloud services that require more functionality provided by HTML, Ajax or even Flash / Silverlight. With the capability of discovering and adding apps through the built in Apps-Menu, it will become fairly easy for users to add new functionality and applications with a few clicks. The Apple App-Store has shown how powerful this can become…

Good thing is that Google has announced the release of Chrome OS only for the end of 2010, so they will have some time to make sure all the standard use cases for notebook users and at least the basic hardware support (printers, scanners, cameras, webcams, etc.) is covered. Because the acceptance of a pure cloud based OS will stay and fall with the ability to do the stuff that you do on a notebook currently. Very few would accept major cuts there…

My feeling is that we will see many netbooks and small tablet pc before Christmas next year that offer a dual installation of Windows (maybe even Linux?) and Chrome OS. For a quick checking of emails, chatting with friends or updating Facebook with some photos, Chrome OS will be the faster option. Windows will step in for everything that requires a regular operating system. Maybe it will even be possible to launch Chrome OS virtualized inside Windows or Linux. That would safe us some reboots (which we know can take long if its not Chrome OS :).

Very excited to get my hands on this next year!

Thursday, August 27, 2009

What about “Virtual Private Clouds” ?

Amazon recently introduced what they call “Amazon Virtual Private Cloud (Amazon VPC)”. which basically is a secure and seamless bridge between a company’s existing IT infrastructure and the AWS cloud.

The service is currently in public beta and is supposed to give companies a chance to re-use their existing IT investments while still leveraging the Amazon cloud computing offering. The EC2 instances will be isolated and connected through a secured VPN tunnel.

VPC_Diagram[1]

Graphic from  http://aws.typepad.com/aws/2009/08/introducing-amazon-virtual-private-cloud-vpc.html

This makes perfect sense for many companies that have spent big money on their computing and data center and cannot move into the cloud entirely in a big bang. This solution gives them the possibility to take a phased approach and move service by service into the cloud. Additionally the business critical systems (and the data) might be kept in-house if required.

Monday, June 1, 2009

Do “Private Clouds” make sense?

I actually should better ask “is a Private Cloud really Cloud Computing?”

Quite frankly i do not think that what Gartner and many others call a <Private Cloud> is anything near Cloud Computing. Agreed there is a trend towards virtualization and this makes perfect sense for large corporations – but Cloud Computing for me implies that the user (and even the IT department) does not know where the application or service is running and where data is stored.

The reason to use a Private instead of the Public Cloud is most of the times security related, i.e. I want to assure that sensitive data is not leaving the company in order to reduce risk or ensure compliancy with corporate policy or laws. Distributing the available internal computing power and the available storage across many users, apps, departments and locations is more or like a task of clever load-balancing. Some of the technology is probably similar to the one used in the Cloud, but all this is still very much in the control of corporate IT.
The fact that I am sure that the data does not leave the companies firewall shows that I always know where it actually (and physically) is. Isn’t this is an antagonism to the concept of Cloud Computing?

Saturday, March 28, 2009

Cloud Evolution

Dion Hinchcliffe posted an excellent article about the current state of Cloud Computing.

Interesting to see the evolution of this new service evolve and generate the agility that companies want from their IT infrastructure. I am curious to see who the big players in this arena will be in one or two years from now… Amazon, EMC, IBM? Somebody completely unknown today?

As soon as the market agrees on standards, the fight for market share will start and we will see a huge amount of investments being made… increasing competition will then put pressure on prices, playing into the hands of consumers.

Friday, March 13, 2009

Governance in the Cloud – who and how?

image One one side it is very attractive to put data and processing into the cloud and avoid the costs and problems of up- (and down-) scaling of the own IT. But there are some questions that need to be asked (and solved….) before mission critical data and functionality can be moved outside a controllable environment:

 

  • Who guarantees Data Security (and how)?
  • How are SLAs controlled and enforced?
  • Which law (i.e. which country) will apply?
  • Who will make sure the law is enforced (and how)?
  • What happens if the provider goes insolvent?
  • What if the provider is acquired (and HQ moves to a different country)?

As technology moves much faster than regulations and laws, there is a lot of uncertainty involved at the moment. As long as my datacenter resides in Germany, German law will apply and the CTO is (kind of) responsible for the compliance to the respective regulations. Even if the IT is outsourced, the company providing the hosting services is responsible and can be sued. In a Cloud Computing scenario, my data is theoretically distributed all over the world, which brings up the question which law applies and who is responsible for it.

Some of the data might reside in China or India while it is processed in Europe or North America. What if a country changes laws unexpected or the company providing the cloud services is acquired or insolvent? The missing control could easily bring a business down when relying on cloud services that do not deliver anymore or when mission critical data is inaccessible for a longer period.

Outages of Amazons Cloud Services or Google showed that this scenario is not so unrealistic and the risk is rather high. But what to do when cost reductions do not leave any other option other than Cloud Computing? What rights and possibilities do people and companies have if the provider of cloud services abuses my data? How would I even know where my data is and how I can get access to it?

Tuesday, March 10, 2009

Cloud Computing is the new Grid Computing?

Comparing the idea of Grid computing to the new Buzzword “Cloud Computing” shows that - while the basic idea is the same - there are major differences in the two concepts.

One difference is obviously the intended usage. Grid Computing was planned for (single) applications with high demand in processing power while the idea of Could Computing is using such a grid on an internet scale to balance the load of many applications running in parallel.

This requires virtualization in order to provide the sandbox for each application:

image

Virtual Appliances are ready-to-run software packages that are pre configured on a virtual OS (or multiple ones), delivering a quick and easy to use solution for a certain part of the business. In my eyes, this is very appealing to all kinds (and all sizes) of corporations that do not want to go through the hassle of installing a local IT and facing problems when their business is growing (or shrinking) fast…

The main differentiator to SaaS is the Multi-Instance vs. Multitenancy. The control over a SaaS delivered platform is limited, while a virtual appliance allows full control in the VM boundaries.

Let’s hope that the Open Virtual Machine Format (OVM) will establish a standard that allows to create portable VMs that can move form local IT to the Cloud (and back) if required.

VMWare already introduced a concept called “vApp” that allows to package a set of configured VMs that compose one application and are managed as a unit.

image

Interesting times ahead…

Saturday, January 24, 2009

Cloud Computing – The Power Grid of the modern IT

Roger Smith outlined a very nice analogy of Cloud Computing to the existing AC Power Grid.

1047129_transmission_towers Nice way of thinking about this… as we agreed on 240 Volts and 50 Hz (well, of course the US did not really agree on this) we need to agree on standards in Cloud Computing to make services and providers interchangeable and open the market for substantial growth.

This actually puts some pressure on all the proprietary APIs that are currently used and puts up the question on how the new standards will look alike? Are we talking about SOA and Web-Services here? Or will we agree on JSON or pure XML as standards? Will the future be RESTful?
Or maybe none of the mentioned will be able to pass the critical mass and the Cloud Computing providers will have to deal with a mix of standards that need to be supported in parallel?

Interesting times ahead!

Wednesday, January 14, 2009

Cloud-Computing to invalidate Moore’s Law?

An interesting article on InfoQ says that Cloud-Computing will dramatically reduce the computing power per Dollar and therefore grow the available processing power even faster than Moore predicted in 1965 – expecting an exponential growth.

Not only will it increase the available processing power, it also requires ultra-fast networks which will help network infrastructure providers like Cisco and Nortel Networks (are they still around?) to create and sell new products and services.

Providers like EMC will make storage available in the cloud – another crucial piece of the puzzle.

Wednesday, January 7, 2009

Welcome to the Cloud

I created this blog to share my thoughts on Cloud Computing which I predict to be the next big buzz.

As I am busy with my Masters' Thesis, do not expect to see many posts here until Summer 2009...