As mentioned in a previous aricle on virtualization I use SaaS (Software as a Service) and PaaS (Platform as a Service) interchangeably as I remain unconvinced of the difference from a customer’s perspective. The cloud computing market size, depending upon how it’s defined, is estimated to be about $50B by 2011, of which 70% is in PaaS/SaaS, 30% in IaaS (Infrastructure as a Service).
Within PaaS we must define the focus of the platform, e.g. web applications, CRM applications, billing applications or SDP. Google, through its App Engine offer, enables businesses to host their websites and data in the Cloud and enable them to use services like BigTable. The two things are important to note: you can only host python applications as of today and you pay per bandwidth, storage and CPU used for hosting this web application; and you can easily integrate with other Google Services and Google Accounts.
Amazon does not offer a way to host web applications on the Cloud, but simply provides virtualized hardware (IaaS). The main IaaS providers I see in the market are: Amazon, Joyent, Flexiscale, Reckspace, f5 and GoGrid (plus there are lots of brokers and niche providers). The savings come through economies of scale and statistical multiplexing to give a higher average processor utilization (>50%) compared to a typical 10-20% untilization within an enterprise, generally a factor of between 5-7 in savings is possible, which is significant.
So the choice a customer makes between IaaS and PaaS needs to analyze:
- Vendor Lock-in: if you deploy on Google or Microsoft Clouds, you make a choice on both technologies and with which services you’d like to integrate;
- Ease of Use: deploying a web application is easier to do than deploying and managing a complete infrastructure; and
- PaaS or IaaS: do you want the Cloud Provider to offer you a way to host your applications (if you can accommodate with their technical restrictions) or an infrastructure allowing you to host your applications (without restrictions) the way you want?
Operators pour money into their data centres and custom systems integration; Vodafone Live as a great example of a vast custom CDM (Content Delivery Management) development with little to show in service differentiation or lower operational costs. Today, suppliers such as Ericsson and IMIMobile (see this previous article) offer content delivery management as a PaaS, smaller operators have widely adopted this approach.
I think this trend will naturally extend to the SDP as operators look to minimize investment in large speculative SI projects. Hence, there is an immediate business opportunity for the NEPs (Network Equipment Provider) to deliver a PaaS SDP, especially to small and medium sized operators or operators undergoing rapid growth. A PaaS approach allows the operator to flexibly customize their platform to react to local market conditions, while taking advantage of the 5-7 times cost advantage of using a cloud based approach.
However, for both IaaS and PaaS the critical issue is SLAs (Service Level Agreement), this is going to be a critical barrier for many operators. SLAs within the data center are easy; the challenge comes for the end-to-end SLA (which includes the WAN, and enterprise network availability.) Here the need for remote infrastructure to monitor and manage link availability in partnership with operators is essential, as well as integrate into the existing network infrastructure. SLAs are one of the reasons Google is investing in undersea optical capacity and other global optical infrastructure, to enable it to get preferential access to global connectivity to offer its cloud services with a global SLA. Some operators are in an advantageous position, they can provide the WAN SLA, hence why I think PaaS in telco is an interesting opportunity especially for SDP.