Cloud-based computing and virtualization have been one of the hottest areas in communications over the past decade, and as the technology matures, it is being used by many corporations worldwide. In order to build a world-class technology, one must examine lessons learned from previous introductions of technology insertion into the marketplace in order to optimize cost, capability and cycle time across the business. Below is my personal top 10 of some of do or dies in the cloud computing world, which can be used by network engineers in order to design and provide best in class products across the globe.
10. Network Virtualization: Due to the size of DC capacity using layer 4 and layer 7 switches, and given everything is moving toward Ethernet Based services, I believe brand new architecture for Data Centers is needed for all Service Providers and Over the Top Players to succeed.
None of the capabilities that we just talked about would be of value if, at the end of the day, users do not trust the security and privacy of their data in the cloud. Moreover, depending on the industry vertical, there are stringent requirements of security and availability for some mission-critical applications that should not be underestimated in the process of migrating to the cloud.
9. Platform Virtualization: Today, vendors measure themselves and present to ISPs by capacity of backbone, i.e., have 8 Tbps. The new world cares about scalability and reliability but we need to move to the cheapest hardware commodity while increasing software value. This is the key to succeed in the 21st versus the 20th century for vendors, SPs and OTT Players.
Platform Virtualization is really part of the Network Virtualization process and, simply put, refers to the virtualization of hardware and its components. Hardware-independency provides an extra degree of freedom and flexibility for operators to be able to build elastic networks that can scale up and down as required.
8. Infrastructure Value: SPs understand the unique value of network assets they have, like policy changing, but what they are lacking is user awareness, high level QoS per user per session. This is a great area for SPs to enable their network to become 21st century IP compatible and support many services such as cloud and others.
Service providers should focus on unique network-based assets and capabilities that can add value to Cloud and OTT applications and expose (and possibly monetize) such capabilities through Web APIs for win-win use-cases where both end-users and cloud-based applications can benefit from such capabilities. In the absence of such unique capabilities, OTTs will continue to hurt operators’ revenues. One example: OTT’s will burn 40% of SMS revenue by 2015!
7. Converged Services: Today, SPs charge around $8 per month, but no one is offering HD quality with much more flexible charging with QoS and authentication. One example is within a session working with OTT players like NFLX and charging an extra $1 for HD streaming and then sharing that with SPs. I think this is a fair use of SP’s network assets, but this new thinking needs to be paid much more attention.
Users are increasingly using multiple devices and multiple network technologies to access their data and applications in the cloud. Operators should focus on differentiating capabilities for treating each user, and in turn, each session and content type differently and provide personalized QoE, with capabilities for flexible charging between devices, networks, etc. One example is zero-rating for certain apps, e.g. Facebook, etc. – or free (out-of-dataplan-quota) HD-quality for certain OTT sites.
6. Data Analytics: 1980s: Data Mining, 1990s: Information Mining, and now we have the Knowledge Mining model for the 21st century. There is a huge amount of unexploited data in the network which can be used for generating new services. Given the massive amount of data, including data retention, one needs to think about heavy duty compression technologies to help solve this problem. For example, in the US alone, 30% of devices are returned for a number of reasons, including unhappy customers, RAN network, Packet Core, CSP and Network – and no one is looking at user-generated data in knowledge mining factors.
There’s probably a mind-blowing amount of data that is generated on the network today that is not being stored or exploited. This includes various interactions between network resources as well as interactions between users, devices and the network. When you look at the extent and reach of an SP’s end-to-end network, you can imagine how much unique network data and user data analytics can be processed to add value to the cloud applications. This of course does not have to be intrusive in any way – one example of predictive analytics and actionable intelligence is to dynamically and automatically provision network capability based on previous user traffic.
5. Cloud-Based Services: If operators want to be part of the cloud ecosystem, they should start building their new offerings in the cloud, maybe initially in their private clouds and data centers, before a more comprehensive model where they would have offerings in hybrid clouds (more suitable for enterprise) and even public clouds. These services can include both consumer services (e.g. entertainment, communications, data management (backup/iCloud-like)) and enterprise services, as well as their own internal Telco-applications (including OSS/BSS applications).
The main message is to implement your own innovations given the massive networks SPs have in the industry – Web Conferencing is one in which video will consume over 91% of traffic in 2015 and an even higher percentage by the 2020 timeframe.
4. Cloud Orchestration: The operational complexity of advanced cloud services increases exponentially, especially when you deploy large-scale applications with thousands of virtual machines across multiple data centers where SLAs need to be enforced and software usage (license) and faults needs to be monitored and managed.
Operators are in a unique position to add value to cloud orchestration through network awareness (e.g. orchestrate according to network status) and integrating cloud orchestration with their network management platforms. If we look at cloud services today, there are also huge challenges we need to look into quickly. For example, with NOCs and GNOCs, how do SPs and OTT players with cloud services manage the services across? Today, there is no way to orchestrate down to network management and problems will even be more complex due to legacy systems and applications that exist across the planet. So, cloud management and orchestration needs to be aligned with various products and services.
3. Enterprise Applications: It’s no secret that one of the biggest revenue-generating opportunities for the cloud providers are the enterprise applications. Operators need to start working with their enterprise customers to identify unique requirements for each application type and industry sector, and reengineer the multi-tier enterprise architecture (e.g. data, business logic, web/presentation) to design the most optimal hybrid model and to determine what the actually benefits are from moving to the cloud and what needs to stay within the enterprise.
Today, we have a huge complex model for enterprise customers with companies like IBM, SAP and others, selling architecture with middleware, data plane, customer plane and other network assets. What I believe is needed is to “cloudify,” which is practically not possible today. For example, in industries like biotech, pharmacy, manufacturing and auto, one needs to tier every enterprise customer via cloud by looking at their data, MW and applications. Some data can stay within enterprise but some can easily move into cloud for better efficiency and productivity. Like unified communications, we need to reengineer the entire enterprise model for all services, which I used to call XoIP (SoIP). Google has attacked consumers, and the next path for them is enterprise and building a reliable and scalable platform which will also eat SP lunch.
2. Unified Communications: Most of us today probably use one form or another of free cloud-based tools for unified communications, e.g. video conference and chat (at a minimum) – e.g. Skype, or equivalent. With the advent of HTML5 and WebRTC, cloud-based real-time communications have the potential of being the next big wave and a significant threat to traditional unified communications businesses.
Operators need to start migrating their communications services to the cloud and start working with OTTs for unique value-added capabilities, e.g. enforced SLAs and cloud-federation. Companies like Avaya and Polycom will have to start thinking about how to manage cloud services for unified communications across different SPs. For instance, if we have a video call from the UK to the US and go through 1-3 cloud services, how does one manage this complexity? IMS had some capability, but not to the level it needs for the 21st century. For example, how do you enforce policies with a data service? Cloud services have had some major outages with violations of data privacy. One needs to get the users on board and trust the cloud services they buy and agree to. New work will for sure real time with HTTP and browser capability so everyone needs to think about unbundling and virtualize quickly.
1. Security & Availability: 1980s, Dumb Devices, Intelligent Networks, 1990s, IP bigots in SV said we need to have dumb networks and intelligent devices. I said wait a minute — How do you scale billions of IP end points with policies, encryption and authentication? Then they called and said you’re right. We cannot be on this side of the pendulum, so I said the proper model for the 21st century is smart devices with intelligent network is where the world is heading in 10 years or less, which is exactly what we have been experiencing with smart phones like the iPhone. None of the capabilities that we just talked about would be of any value if, at the end of the day, users do not trust the security and privacy of their data in the cloud.
Moreover, depending on the industry vertical, there are stringent requirements of security and availability for some mission-critical applications that should not underestimate the process of migrating to the cloud. So, if we do not pay attention to cyber security around cloud services, we will have problems of biblical proportions in this industry very soon.
Dr. Eslambolchi
