Internet(work) Philosophical Musings in the Cloud Era
Jump to TL;DR
Fair warning, this will likely be a long post; there is a lot going on, so I hope I can make it relatively coherent!
With cloud service providers (CSPs) all building out their own private backbones, I've been thinking a lot about the nature of the internet, open source, and the implications to organisations that are focusing on the cloud. More specifically I've been trying to describe the business value of the network, which is proving to be more difficult as cloud adoption becomes mainstream.
The internet has been a great leveller. Large and powerful organisations of old are struggling to be agile, transform and innovate whilst tech start-ups are disrupting entire industries with an app. It has been key to the economic growth in many developing parts of the world.
However now there are disruptive forces changing the shape of the internet itself, as the world and society become ever more digitally dependent. I’m only focusing on the business aspects, and not implications for private and personal use as various social and cultural controversies emerge… maybe in another post.
My contention is that a robust multi-cloud strategy gives organisations the commercial levers to counter the business challenges of going "all in" to cloud. We'll explore some of these ideas throughout this post.
I believe we are now approaching a period of heightened anxiety across tech industries, particularly networking, for traditional IT services; perhaps across many other industries too as the realities of AI and automation become more present. CSPs have a very powerful value proposition: greater agility; lower operational costs; reduced capex; simplify business processes (i.e. purchasing, logistics, vendor management, change management) to name a few. It is becoming more important for IT to show its true value, but at the same time harder to articulate and, in the face of “cloud first” strategies, even have the conversation in the first place.
At the risk of "stating the bleeding obvious", the key to truly realising the benefits cloud is doing it right. That, however, is a hard thing to. Organisations that are transforming, rather than born in the cloud, can struggle with the significant organisational change, complexity (despite promises of simplicity… another post required here too) and implementing governance at scale. Processes throughout the business, not just technology teams, need to be completely reworked to achieve this. Additionally, technology teams need to reskill, applications need to be refactored, and cloud-based solutions need to be constantly developed to take advantage of the myriad new features and enhancements.
Given the difficulties of delivering though, why is there this heightened anxiety? It’s my view that traditional IT teams are having a tough time demonstrating real business value, as in a Return on Investment (RoI). Traditionally, big enterprise has treated IT as a cost centre only and changing that stereotype in the midst of the cloud revolution is no simple task.
I’m wondering off topic here though, so let’s park that for now and move on…
I thought I would start with a few (fairly) recent announcements, to help set the context of this post. The developing technical capabilities of CSPs, their focus on networking, and the pace of change in an ever more complex landscape have led me to contemplate this topic much more of late.
At AWS re:Invent in 2018 an announcement that caught my attention, and created a mental picture of service provider CEOs around the world head in hands, was Global Accelerator. It allows the advertisement the IP address of an endpoint hosted in AWS from all edge locations globally using a methodology called anycast, irrespective of the region it is hosted in. An end users data path to this service enters the AWS network at the nearest edge location globally and is backhauled across the AWS network rather than across the traditional internet.
The benefit, AWS tell us, is that there will be an improvement in end-user experience due to improved network performance. The pros and cons of this and technical strengths and weaknesses versus CDN type services is not something I'll be addressing, suffice to say there are some genuine benefits (and some great marketing promises too!).
The diagram below describes the path traffic would take between the end user and an application over both a standard internet transit and the AWS Global Accelerator service:
It's certainly not just Amazon though, Microsoft’s Global Network has received a significant amount of investment, and that is only continuing. There are a number of services that fit the context of this post, but ExpressRoute Global Reach and Virtual WAN which I first noticed in the Azure Networking fall 2018 update is a good example.
An Azure customer can integrate their SD-WAN service if it is a connectivity partner, or alternatively connect via IPsec VPN to a gateway via their local ISP to the Azure Virtual WAN service in their region, then take advantage of the Microsoft Global Network for backhaul. Additionally, customers who are using ExpressRoute services can also take advantage of the same network for global reach.
In essence, Azure customers with existing ExpressRoute services or SD-WAN have options to use the Azure Global Network to backhaul traffic around the world. It is very clear CSPs are tackling the tricky subject of enterprise WANs, and networking more generally; perhaps seeing these as another way to encourage customers into their ecosystems... they are certainly not short of vendors chomping at the bit to partner with them.
The following diagram is taken from the Azure Global Network pages, and gives a sense of the size and scale of the investment Microsoft are making:
This is a little bit of an aside but is a contributing factor to my central point. The internet as we know it is changing as governments attempt to regulate it. Broadly speaking I understand the concerns that things such as the net neutrality repeal and things like Article 13 in the EU are trying to address, but I do not think it is being handled well. In these examples, the focus appears to be on only one side of the argument and going against what I believe should be a free and open medium for sharing information. I am against any legislation that negatively impacts the user experience of the internet as we know (or knew) it. I'm not going to go into any detail on the politics/cultural/societal side of things here; to restate I am focused on technical and commercial impact only.
As with everything there are downsides but stifling all the consumer benefits of the internet with regulation is not the right approach in my view - we must look for new and innovative ways to tackle the downsides. But it’s not just the consumer that will ultimately be disadvantaged.
So… What's Going On Here?
With respect to business value, In the short to medium term businesses (particularly large organisations) that cede competitive advantage in favour of promised simplification and efficiency will fail to deliver and struggle to innovate. Predominantly this will be due to the size and scale of the transformation and weight of subsequent restructuring required, along with a lack of clear strategy since they will self-evidently be breaking new ground.
What is interesting to me is to consider why the CSPs want all the traffic on their private networks, and what the potential impact is to global ISPs businesses. As the CSPs grow and hosts more and more of the internet in their data centres, it makes perfect sense; through businesses consolidating onto the cloud, they have the economies of scale that mean it is more profitable to effectively grow their own “internet”.
But what about the businesses hosting on those platforms? Getting end-user traffic into their network as quickly as possible, CSPs may gain a competitive advantage or opportunities for additional commercial benefits through agreements and partnerships to enhance favourable consumer decisions for other customers hosting on their platforms.
Imagine a platform or website with a business model that relies on advertising. Displaying adverts for other products or services from companies that host within the same CSP would be a commercial (and technical) benefit to them. CSPs might, for example, discount advertising on a platform that is hosted within their environment or prioritise the data. This is all getting a little convoluted, but I hope you understand the point; getting customer traffic on their platform and keeping it there creates a kind of captive market/ecosystem that you might have to pay a premium to access (or leave...).
A further idea that I’ve been contemplating is the possible consequences of businesses integrating their networks with CSPs, or perhaps, in the case of Google fiber for instance, directly consume internet services through them. Services hosted within the CSPs ecosystem may be treated preferentially, or perhaps services that are not within it may have a tariff applied to it. So as a customer of a given ISP for my last mile, it may cost me more to buy a service that isn't hosted within that ecosystem. The implication to the business being that it may need to consider hosting its services within multiple clouds to attract customers and thus will have higher operating costs.
Hybrid & Multi-Cloud
Firstly, I will define the terms as I understand them:
Hybrid-Cloud: Infrastructure is deployed in traditional on-premise data centre or colocation centre and a single CSP. A more contemporary definition is a platform with workloads across a combination of public and private cloud.
Multi-Cloud: Infrastructure is deployed in multiple CSPs and may also be contained within traditional on-premise data centre or colocation centre.
I’m currently considering implications of vendor lock-in in the cloud and mitigation or “commercial levers” that an enterprise should perhaps consider and have a reference architecture for. It seems that the 3 main clouds are associated with specific capabilities that they are recognised or seen as the leader for, which I would summarise as:
- AWS: Web/App
- Azure: End-User
- Google: Big data/Machine Learning
I’m not sure if this is a good or a bad thing but it does offer an opportunity for, and value to, a multi-cloud proposition. The diagram below describes at a very high level, some of the architectural opportunities available with a well-planned multi-cloud strategy:
It should be noted “cloud-native” or “cloud first” strategies themselves are hard to get right as part of a transformation as I mentioned earlier in this post. I think this topic warrants a further post, rather than veering off on another tangent again, however. What I am suggesting here is that there is value in developing a network strategy that uses the internet natively and delivers value over the top in the “services layer”. During the transformation these might be cloud enabling services, showing value to the business in being a way of moving services to the cloud more efficiently. Later these become value added services, using the ubiquity of the network as a key strategic asset in understanding data flows, analytics, customer experience, efficiency and so on. What I’m really trying to do here is articulate what the network has always done, take advantage of new “software-defined” ideas and architectures, and describe the value and importance of the network in a way that makes sense for a business focused on cloud transformation.
It is easy to disregard the value of traditional IT when embarking on a cloud transformation. At the simplest level everything is available over the internet so the network itself is questioned as to how valuable it is to an enterprise, and indeed to cost of having specialists that understand it in the face of this cloud revolution. The problem then becomes one of translation; without getting into technical detail (at which point you may lose the attention of the senior business stakeholders) it is difficult to articulate the value of the network clearly. To start to change this takes perseverance, and it may even require problems to arise in the cloud migration that necessitates a rethink of the business strategy. Some applications may be more difficult or costly than first anticipated to move to the cloud, this would be a good time to start to discuss the benefits of hybrid architectures.
I remain convinced of the necessity of hybrid-cloud particularly with respect to digital/cloud transformation, and it appears Microsoft and AWS agree with me given the development of Azure Stack and AWS Outposts respectively. This is another strong argument to support the development of a multi-cloud strategy early. It’s also logical from a technical and business perspective; as you start on the journey to the cloud from on-premise to my mind you are at least hybrid, and therefore supporting and operating cloud-based and traditional on-premise infrastructure.
For years AWS would barely even recognise the idea of hybrid-cloud, but with the advent of VMware Cloud on AWS and now AWS Outposts shows the reality. There are many reasons that businesses would want to proceed with these solutions, and that again is the topic for another post, but we need to consider what this means from a long-term cloud vendor lock-in perspective.
In the context of some of the ideas I raised earlier there are additional further strategic considerations with regards to data gravity and the cost of data transit as the organisation becomes more hybrid. Additionally, if your organisation is global, and is exploring growth markets, it is likely you’ll need to consider multi-cloud much more closely, with key vendors such as Alibaba Cloud offering more comprehensive services in these regions.
I have to bring this post to a close, but I hope my ramblings here have provided some food for thought. To try to briefly summarise, I’ve looked at how cloud providers are building out their own networks, and what potential impact that has on traditional enterprise networking as we know it.
There are a lot of subjects I have only touched on, and I will look to write more frequently this year to try to clarify and expand upon some of them. The purpose of this post was to outline some of my thoughts and areas of “concern” as many organisations leap headlong into a cloud strategy that serves several immediate benefits. The trouble as I see it is that it lead them on a journey that locks organisations into significant cost, without much of a backout strategy.
If you are making the move to the cloud, and I think that is inevitable for most companies, ensure you seriously consider how you are going to approach using multiple cloud vendors, and create a strategy that best suits your business. Even if you only begin your journey with one strategic vendor, understanding multi-cloud and the strategic benefits will help decision making in terms of vendor comparison if nothing else. Importantly it may also help define key skills and even the culture you need to develop in your business that will be of benefit in the future.
- Cloud Service Providers are disrupting everything
- There is a lot of investment into private global networks and submarine fibre
- Beware cloud vendor lock-in
- Start developing your multi-cloud strategy now