A holistic guide for edge Services hosting and commercial deployment models for a digital Telco
Market Segment: <Telco , Applications and Enterprise domain>
In the latest Edge computing conference in London the palpable notion was clear that developers clearly see a potential of applications that can exploit the potential of Edge services , consider the fact major T1 operators made initial 5G rollouts it seems inevitably true that CSP’s in the new 5G era either have to speedup or simply lose the game to Cloud vehement something that already happened in 4G era when OTT’s built there future on Telco’s dollars of investments . This is why I thought to share some insights on this direction as things look more clearer to me in this direction now compared to my thoughts shared earlier.
This paper is the part-2 and succession about MEC and edge deployments I shared before, if you have not read that paper, I suggest to please read it first MEC as enabler of Telco’s Digital Transformation ~An ETSI ISG Perspective
I think a new paper is needed as a lot of progress in industry has happened since I shared my thoughts on first version of paper . Specifically, with 5G early deployments taking off more and more requirements are coming that necessitates an Edge deployment. Similarly, as we are becoming clearer about Edge services it looks like the boundary between clouds, applications, end users will become more diluted whereby finally customer expects to develop anything as per requirements, SLA’s, regulatory limitations, compute and edge resource needs etc
Edge deployments is not a new concept as those familiar with Web scales and Enterprise will confirm it however it is also a reality that MEC or Edge in Telco’s is still a Stanza among all transformation and cloud based solutions used by Telco’s but this is going to change specifically with startup’s and hyper scale cloud vendors joining the edge race as example of this is Vodafone using Saguna MEC solutions , Google and Aarna partnership going to Carriers Edge is another proof of a changing game in the Edge transformation .
Since the core idea of this paper is to show my view point on practice and less theory so as to negate idea that MEC solutions are not production ready or still they depict some sort of elusive solutions. So, let’s start with the understanding of production ready and market hungry MEC use cases first as below.
Killer Use cases of MEC
The potential list of MEC and Edge solutions may vary but what as Telco or industry we should focus more is the quick win. As of now what I understand all Use cases considering current limitations of 4G and 5G NSA are to be focused on
- Latency reduction like AR/VR , Gaming
- Data Reduction like analytics
- Off Load at the Edge like CDN , video streaming
From technology commercialization point of view and consider experience of projects deployments following seems like top use cases.
- Top1: Video and AR
- Top2: CDN and Video Caching
- Top3: Video analytics
- Top4: Gaming solutions using VR
- Top5: Enterprise connectivity at the Edge
- Top6: Remote Factory – DT MobileX key use case in EU market
Who should own the Edge solutions?
For this the best source I found recently is interview of CNBC with Verizon and IBM CEO which explains Edge will offer an opportunity to Telco’s which they offered to OTT in 3G/4G era . Nokia. Chris Jones says that mobile operators in particular have the advantage when “serving mission-critical industrial automation applications requiring high performance wireless connectivity, where tight optimization of the entire data plane is essential.”
Telecom operators also have the motivation to use edge clouds to operate their own networks more efficiently
However, it is not enough as vendor capabilities required for Edge will be totally different than those of legacy. As an example, relying on Open API’s for the Edge applications and use cases more stitched to Cloud and Data Lakes like AI, ML, analytics will mean entrance of entirely new breed of vendors for example Saguna, Google, IBM with solutions deployed in 5G edge sites.
This is definitely something we could not imagine in 3G/4G era with one vendor dominating the wireless side provided both the access nodes like eNodeB’s and their related Core Infrastructure at least the RNC level.
The Hyperscale and Public Cloud giants have already shown muscle in this domain and we have seen active participation of AWS , Google , Azure in Edge deployments of Telco’s. As an example Azure Edge , AWS Outposts and Google Anthos
Enterprise opportunity for Edge solutions?
Las year we did a comprehensive white paper in ETSI for MEC application in Enterprise MEC in an Enterprise Setting: A Solution Outline (September 2018) and since then the requirements and focus of MEC use cases for enterprises has only increased however when it comes to business and enterprise market still we believe biggest challenge to MEC and Edge commercialization is embodiment of use cases that fit exactly the enterprise requirements without looking too difficult and complex .
This is an area that requires a careful analysis of applications and platforms and their packaging to support real world deployments , for this in 2019 ETSI MEC has done lot of work together with partners and vendors to define MDT’s (MEC deployment trials) which focus on commercial use cases and actual deployments of MEC in the Edge . The full list of ETSI MEC MDT’s can be found here. MEC Deployment Framework.
A Special node on Hyperscale providers
Continuum to who Own’s the Edge it is more important to build capability using Hyper scales primarily as their edge infrastructure offerings are already mature and Telco’s do not come to a situation where they build something can not be integrated with their networks or will make them reach to customers sell their own solutions irrespective of access media .
There is a huge risk of cannibalization because skills, capability and customers outreach of cloud CSP’s is too mature compared to Telco’s.
Running the RFP of Edge Deployments
It is true that everyone recognizes importance of Edge however $ dollarization of this is another issue. This is why it is very importance for a Telco to evaluate RFP and use cases that meet certain characteristics of which scaling is the most important. It is because in Edge we will be talking about 1000’s of nodes compared to 10’s of nodes previously in Core sites.
Next and most important is how to develop an Edge platform that can meet diverse applications in a common manner.Such applications will mostly come from nimble players and 3rd party
For RFP solution selection thus, I find below are top requirements to be considered.
- The Edge solution should seamlessly local breakout irrespective of access
- ZTM provisioning of all Edge services
- Remote and automatic O&M services for agile operations
- Operational efficiency (Only domain truly exist now VNF Day1 , FCAPS , RCA , Monitoring
- Intelligent automation (Predictive analysis, Day1 /Day2 automation)
- Should capitalize all infrastructure options like Public Cloud, Private Cloud, Hybrid cloud etc.
- PaaS and TaaS capabilities must be provided to developers and it should be provided as developers use tools in SDK environments of public cloud as of today. A friendly ecosystem for developer in terms of API abstraction and Testing using Swagger and OPENAPI2.0
- Total Infrastructure abstraction of infra
Finally, I think still Edge SDO’s are fragmented for example we can see ETSI MEC, OSF Edge computing, Akraino, Airship and we believe it is in the best interest of the industry as a whole to coalesce efforts and resources into a common way forward.
@ airship as application platform specially Containers seems deliver more promise
@StarlingX fully integrated and tested platform for Edge including FM/PM, where deployment is only one of the challenges to address? That needs alignment with App management
For Long term success it is very important to align vendors MVP with SDO efforts to have clear mapping with MEC applications and platform capability to ensure maximum monetization of MEC use cases.
Automation of Edge Deployments
One thing I specifically want to talk about is the automation at the edge. It is because it is different than Core sites orchestration is unique way that edge sites need more intelligence and automation along with scaling measures. It is infact therefore most of T1 Operators are building Edge Orchestration independently along above perimeters.
The other important issue of this automation is how these services will be consumed as ecosystem will not be infrastructure and VNF’s but also 3rd party, OTT’s and infact any developer.
The third and most important issues which Edge need to solve is Workload placements and optimization. Imagine a world where you want to host a Analytics application based on specific customer segment and use case so how you can decide where to deploy it , in 5G era it will not be something based on Site and coverage as we believe it today . This is something that has to be decided based on network real data in real time.
Then finally for the Edge we need to solve issues of workload portability , with cheap infrastructure at the edge how to migrate workloads across sites using same experiences as in hyper scales Cloud providers . This once achieved will certainly be a game changer for the Telco’s and will make them both famous and rich like Mark Zuckerberg
Source: CISCO intelligent Edge
NFVI Models for the Edge
Lot of discussion happening in many communities to define and standardize the NFVI and Cloud offering at the edge , as a CSP will definitely look to reap advantages of scaling hence what we must look out is a pre-built and pre-integrated model that can be deployed in the far site in the most automated manner . The second key selection criteria must be data protection and security specially consider the fact that at the Edge we will run many mission critical service like Healthcare IoT, smart homes, automated cars etc. Finally Edge computing allows the transfer of data to occur in real-time. The predicted surge in data transfers from networking requires a robust system for data gathering and analyzing. All of this means HCI should be the target model for investing in Infrastructure at the Edge .
The most important use cases will be around MEC and 5G as an example which a CSP should target in a commercial environment .
- Healthcare: AR/VR for pain management, AI/ML for imaging, IoT identification/segregation, remote home diagnostics
- Manufacturing: AR/VR for training, AI/ML for IoT, IoT management, Private 5G
- Retail: Smart mirror, AI/ML for consumer behavior, Private 5G, SD-WAN
Recently lot of work has done in Akraino community as per official website of Akraino states
“Launched in 2018, Akraino Edge Stack aims to create an open source software stack that supports high-availability cloud services optimized for edge computing systems and applications. The Akraino Edge Stack is designed to improve the state of edge cloud infrastructure for enterprise edge, OTT edge, and carrier edge networks.”
The best direction for Edge commercialization is to start of small and scale fast using one of the following configuration models.
Config-0 = singe servers with out H/A (use case on 5G sites)
Config-1 = 2 Xsingle servers with H/A (use case on 5G Hub sites)
Config-3 = unicycle PoD
This looks simple however it looks clear to me the Big Three RAN Infrastructure vendors lacks the ecosystem and hence we must start to partner with the vehement of Web scale Cloud providers to look this part as an example a Reference architecture from Google Anthos can look something like below .
Finally be noted the right ecosystem for Edge is the containers and not virtualization primarily because of application requirements and its alignment with API’s .
I will wrap this paper with the following take always as follows
- Edge platforms must be ready one year ahead of 5G SA deployments if an operator has to fully monetize its advantages.
- Use cases selection of MEC should target quick Win and at the same type support to increase customers appetite for Edge applications
- Intelligence for Edge is more important than of core sites and Telco’s must select the solutions carefully considering this long-term direction.
- Edge deployments and RFP’s should give more weight to building platforms that can support onboarding and LCM of any type of 3rd party applications
- Finally, the time for the Edge is now.
ETSI MEC docs
Cisco Intelligent edge – public document
Ericsson Edge –public document
Ericsson Show case in #GITEX2019
5G Cloud and Edge Party – Heavy Reading Public Document
PoC Akraino courtesy of Aarna and Google team
Nokia White paper on the Edge – Public white papers