Skip to main content

Posts

Showing posts from 2010

Cloud Computing SOU

Interesting findings from Zenoss study on CC. Flexibility is now the leading driver for virtualization ahead of capex savings. Even though 90% of startups are working on products with a value prop that is a variation on "Managing VM Sprawl", only 26% of customer have deployed a tool to manage virtual infrastructure. So apparently, the customer is not driven to virtualization based on capex or opex savings. What gives?

Actually this is kind of good news. The whole cloud thing was hijacked by "Opex Savings" value prop i.e. management automation etc. around late 2008. The original cloud vision of internet as a platform (like OS today) was put on the back burner. May be we should get back to the roots of the cloud.

Framework for Network Svcs Deployment

If we refer to the bubble years as the "siliconization" of the network services, then today we are experiencing "containerization" of the nework services i.e. all network services need to be ported to a container that abstracts the oddities of the underlying platform and manages all non-configurable parameters. In addition, it offers the run-time for the control plane of the network services and leverages PCI architecture's VF to leverage HW acceleration.

Should this architecture win in the market place, it will have a profound impact on the networking industry.

HTML 5 Demo

The site http://www.thewildernessdowntown.com/ shows what media made specific for a browser will look like. It uses HTML5. HTML 5 has every thing (but double buffering) that a media designer wants. 3D engine for rendering, texture mapping, color correction, sequencing, audio and video support.

Changing Traffic Pattern on Internet

The absolute traffic on internet is rising but the nature of the traffic is changing. Another way to look at this is 77% of traffic on internet is social networking related and supported by advertising based business models.If you are a proxy device for the web and support FTP, DNS, and HTTP then this data point should get you thinking.

Content as a Service

IDC says the industry spent $16B in 2009 on public clouds. The spending includes HW and SW.


Besides the fact that they don't have a network spend category, what stands out is the increase in relative spending in the App Dev category. Increase in tooling related spending can means that atleast until 2014, IaaS continues to be the dominant service model. End user spending on tooling suggests an increasing ecosystem of cloud enabling tools that enable hosting and interconnections on various available IaaS clouds. Another stand out is decrease in application spending. If we are to believe the three service models I/P/SaaS then we should be seeing an increase in application spending (also PaaS should decrease the tooling costs).

Another interesting data point is incremental spending is still leaning in favor of traditional IT. If in 5 years we are still spendig 2/3rd of incremental dollars on traditional IT, that is not good news for Cloud Apps.





Yet another point which was not surprising…

VM Density

Cloud economics are most sensitive to VM density and less to automated virtual infrastructure management. If the VM density decreases as self-service portals are introduced or SLAs enforced, it will undermine the cloud economics.

The way to increase the VM density is to offload functionality from the VM that creates headroom or "idleness" in CPU, memory and IO. Creating virtual versions of network services as VAs is exactly what you don't want. Having a footprint on the virtual server backed up with HW acceleration outside the VI is what will increase the density. In other words, virtualization of network services is not a P2V activity but more a distributed computing exercise