Friday, December 4, 2009

IGT Cloud Computing Summit – part1

Image Source: Wikimedia Commons.
Created by Sam Johnston using OminGroup's OmniGraffle and Inkscape (includes Computer.svg by Sasa Stefanovic)

I participated on December, 2nd in the first day of IGT2009 The World Summit of Cloud Computing. The interesting event is organized by The Israeli Grid Association which organized a similar event in 2008. The event's agenda includes presentations by global vendors (Amazon, IBM, Microsoft, eBay, Netapp, Sun Microsystems), local companies (e.g., Matrix) Analyst (Bitcurrent) an enterprise (Intel), local and international Guru (Yossi Vardi). In addition an Executives Panel and workshops took place.
This post does not summarize the presentations, which will be available in IGT's Web site. It is my take.

Enterprise Services vs. Social Computing
Most presenters talked about Enterprise Cloud Services.
Yossi Vardi's message was: people are the Killing Application of Cloud Computing. He presented examples of new creative collaborative patterns.


Cloud Computing Past, Present and Future

Comparing 2008 and 2009 conference agendas reveals the changes in Cloud Computing. 2008 agenda included many theoretical and visionary presentations, as well as some down to earth presentations (e.g. Amazon's presentation). This year agenda included concrete marketing messages from IBM and Microsoft including locations of future Cloud Computing services Data Centers and products descriptions and announcements (including availability dates).
Most of the panelists in the Executives Panel supported the opinion that in 10 years Cloud Computing will be the architecture of choice of most enterprises.
My opinion is different. I do think that Cloud Computing will be Mainstream; however, it will take more time until it will be implemented for Core Business Services for most organizations. My opinion is that in 10 years we will see enterprises with mixed environments: Internal Data Centers and External Cloud Services For more details see previous posts titled: SaaS is going Mainstream and Future Applications: SaaS or Traditional?
The following section may support my opinion.

The Case of Intel

In my opinion, Intel's Gregory Touretskey's presentation was the most interesting presentation. He described a failed Proof of Concept (POC) of External Cloud Computing executed by Intel. Intel's POC executed together with Cloud Computing services provider, checked the scenario of using External Cloud for Intel's Design processes (60% of Intel's workload).  Intel is a lot larger than typical high-end enterprise; therefore its conclusions may not apply to most enterprises. However, some insights are applicable to other enterprises.

The reason for executing a POC is lack of ability to add significant computing capacity in short time. This could happen in case of Capacity Planning error or in case of unplanned significant workload. Usage of External Cloud Computing Infrastructure as a Service (IaaS) services could be the answer to this problem.
My take: This issue is not Intel specific. A need for immediate Capacity, which could not be provided by increasing Data Center capacity, due to long time required for supply processes, could be handled by Cloud Computing services provider. Planning and testing should be done prior to the emergency situation. It could be part of BCP and DRP testing.

The main reasons that the POC failed are:
  • Software Licensing  
Software Licenses limited software usage to specific locations or distance from locations. The IaaS provider's Data Center could be outside the specified location or distance.
My take:  This problem is not Intel specific. Software licensing models should be adapted to Cloud Computing. However, Software vendors profits may decline, therefore expect for a long period until new Licensing models will be available.
  •     Security
The Cloud is less secure due to Multi-Tenancy. Even if appropriate security measures will be taken in order to protect the software and hardware, People (Cloud Computing Provider's employees), could breach them.
My Take: I do agree that people are the weakest Security link. I elaborate on this issue in a post titled: The Chain is as strong as the weakest link in the chain. However, this concern is applicable only for sensitive data like Intel's design data.
  •     Performance  
The reasons for not good enough performance were old generations processors (Intel upgrades its servers processors more frequently then the IaaS providers) and Virtual Machine overhead due to a larger number of Virtual Operating Systems Instances per Physical server in the IaaS Provider's environment in comparison to Intel.
My Take:  Intel is a very large enterprise which manufactures Processors. Therefore it probably upgrades processors more frequently than many other enterprises. The number of Virtual Operating System instances per Server is irrelevant for the IaaS Consumer as long as the Cloud Computing provider provides the service level defined in an Service level Agreement (SLA).
  •    Capacity
The IaaS Provider could not provide the required capacity in a single Data Center. Usage of a remote Data Center implies Network Latency Time which negatively effects performance.
My Take:  This issue could be is specific to Intel or other Very Large Enterprises. Smaller enterprises capacity could be allocated in a single IaaS provider Data Center but be aware of this effect, when signing your contract with your IaaS services provider. 
  •     No Return of Investment (ROI)
My Take: reducing costs is a major reason for consuming Cloud Computing services. I do think that this reason for unsuccessful POC is specific. In the 2008 conference Intel presented its Internal Cloud. I suspect that no ROI for External Cloud in Intel's case is related to efficient resources usage including implementation of Internal Clouds. As far as Intel is concerned, my opinion is supported by Processor usage of 80% and 90% presented by Gregory Touretskey.
These numbers are highly above the average usage numbers in enterprises (Industry average of 15% usage was presented by IBM in the same conference). The conclusion may be generalized to other users of Internal Clouds (whatever definition of that ambiguous concept is used. See Zapthink's Zapflash Private Clouds a Valuable Concept or  Buzzword Bingo ):  
expect significantly less dramatic savings for enterprises which moved from Internal Clouds to External Clouds.

The Case of Microsoft
Microsoft is not IBM. Providing Cloud Computing services does not change significantly IBM's Business model, because IBM is already providing services (Hosting, IGS, and Outsourcing).
Microsoft is a software only vendor. Building Data Centers for providing Azure Cloud Computing Services is a significant business model change.
I can recall only one change in Microsoft's history, which is as dramatic as this model change: In 1995: Microsoft replaced the MSN centric vision by Web model.
In 2009 as well as in 1995 Microsoft is entering the new market after its competitors. A major risk of the new Cloud Computing model is in cannibalizing traditional software sales revenues.
It is too early for analyzing the consequences of that model shift. Expect for coexistence of traditional software selling model with emerging Cloud Computing model.

The Case of Netapp
The presentation of John Rollason, Product Solutions and alliances Marketing Manager EMA, was a down to earth presentation. He defined four types of Cloud Computing implementation types: IaaS, PaaS, SaaS and STaaS.
The ST in the last acronym is Storage. No wonder that Storage vendor like Netapp defines the fourth unique type in addition to the three well known types of Cloud Computing.
He described IaaS as replacing of Hardware Resources Silos for different applications by shared virtual pool of hardware resources.
In my opinion it is a result of the replacement of Silo applications (named also as stovepipes or monolithic applications) by SOA services. In a new world of services hardware is no longer owned by an application and services share the same hardware.
According to John Rollason changing the infrastructure architecture to a shared Cloud model requires changing of Storage, Servers and Communication equipment simultaneously. Therefore cooperation between Storage, Servers, Network and Virtualization vendors is crucial.
Netapp partnered with Cisco and VMware.
 I doubt if that partnership is the right partnership for Netapp and if that partnership is a long term strategic partnership.
EMC, VMware (owned by EMC) partnership is tighter partnership including the product development phase. EMC and VMware also partnered with Cisco as a Servers and Network provider.
Netapp could continue its partnership with Cisco, but probably will create a strategic partnership with another Virtualization vendor, who is not a Storage vendor (e.g. Citrix).

Among Netapp IaaS Case Studies he mentioned: British Telecom (BT) and T-Mobile.
Is it a coincidence that I mentioned the same companies in my SOA presentations as examples of successful SOA implementation (BT) and successful Web Services implementation (T-Mobile)?
I think that it is not. SOA together with Web 2.0 are the basis for Cloud Computing.
Once your Enterprise architecture is SOA based you can expand the Services deployment to the Virtual Enterprise including Clouds.

1 comment:

Avi Rosenthal said...

I received the following comment in a Linkedin Group

Thank you for your post, very interesting.
Unfortunately, it were two cloud-related seminars Dec, 2nd and I went to another one - of VMWARE/EMC/Cisco.
As for clouds, it seems that most of medium-big companies will implement some kind of private cloud when small companies will go to external clouds as part of outsourcing strategies.
Posted by Vitaly Karasik

Public Cloud Core Banking: Hype or Reality? - Revisited

  More than 4 years ago I was asked if Public Cloud Core Banking is a Hype or a Short Term Reality? If you had read the post, you would prob...