Author

Alex
Network Specialist

On the death of the Corporate WAN Strategy Moat

Sunday 20 September, 2020

When the Castle's gone, the lay of the Land looks different

In the recent past, the way IT generally worked could be compared more to a that of a Feudal System from the Middle Ages constructing a castle than to the current situation we find ourselves in mid-pandemic. Previously what we built, and have done so since the late nineties, looked something like:

  • Build a Corporate WAN out of MPLS, Leased Lines and IPsec VPNs (the "Castle")
  • Build some Firewalls, IDS and IPS around it (the "Moat")
  • Build an isolated VPN Concentrator into this for the rare times your Serfs needed Remote Access (the "Drawbridge")
  • Host all your important Websites and Apps from within it, using Reverse Proxies and the like (the "Battlements")
  • Force people to work in Office and Branch Environments, using certain tools, in a prescriptive way (the "Feudal System")
  • Conform to the whims of the Architecture and Strategy Departments when modifying this (the "King")

I could probably go on, a lifetime spent in IT is indeed a lifetime learning to analogise; and in some ways, there truly is nothing new under the sun, just reimagining of concepts and events that have already occurred; but that's a blog post for another day. The point I'm trying to make here is that, much as in the fall of the Middle Ages, it takes factors outside of the current system to change into a new one; and that's where we believe the IT and associated Networking and Telco fields are currently at - the divergence of a new "age" of working, if you will.

The Cloud Era was too early

To understand this a bit more, lets step back a bit more in recent times than historical ones - to the beginnings of the Cloud Era, or around 2006-2009, when notably:

  1. 2006 - Amazon Web Services (AWS) launches with three Infrastructure as a Service (IaaS) offerings - Amazon S3 Storage, Amazon SQS Messaging and Amazon EC2 Compute
  2. 2008 - Google App Engine launches (the beginnings of a Google Cloud Platform, GCP)
  3. 2010 - "Windows" Azure launches (but doesn't yet really know what it is, or wants to do)

For most Enterprises ("Big E") and Small to Medium-size Businesses (SME/SMB), these aren't too relevant; probably roughly where we were a few years ago with Enterprise adoption, or "pondering" over Kubernetes and Containers - at worst, they are background noise; at best, a few people maybe dip their toe in and deploy some throwaway Proof of Concept (PoC) into bits of them. Nobody is really calling these out and differentiating based on terms like IaaS vs PaaS vs SaaS yet, as we're not thinking in an Cloudy mode; heck, "Cloud" isn't really a term that resonates with anyone at this point - apart from on TV Weather Broadcasts. Indeed, nobody really uses or understands the terms "On-Premises" and "Off-Premises" at this point, because there isn't really a world outside the Castle (the "On-Premises") anyway; when you live in a Castle, there isn't really a word for the outside world or alternative, because it doesn't enter your consciousness.

For context, let's look at the on-premises landscape during this same time period, where we're dabbling with:

  1. 2009 - Fibre Channel over Ethernet (FCoE) drafted and begins to be adopted by some vendors, such as the Cisco N5596UP in 2011
  2. 2012 - VMware ESXi 5.1 released and adoption of the concept of Virtual Machines becomes more prevalent in "Big E"
  3. 2013 - Cisco Prime LMS (which replaced CiscoWorks NMS) now becomes Cisco Prime Infrastructure, to consolidate the previously-separate Cisco LMS LAN/WAN Network Management and Cisco WCS WLAN Network Management product lines

What really stands out here is that, roughly during the time of the advent of Cloud, the rest of the world is focussing on one sole thing - convergence (of concepts and tooling that already exists). We're so busy dealing with multiple siloes of Infrastructure (that was the big point of FCoE vs FC + IP/Ethernet, it just never really got widespread traction), and trying to simplify the management of the then current state-of-the-art that we're not looking outside the Castle Walls. From our earlier Feudal System analogy, we're playing an introverted Game of Thrones inside the Castle Walls - some Barons will win (FCoE, for a very short time period); some Lords will lose (notion of one Physical Server per App) - but none of us are interested in what the world outside our Moat looks like, nor too much what the next Castle across the way is doing (maybe they never bothered to do FCoE, and jumped straight to iSCSI/NFS).

Shadow IT begins

Meanwhile, outside the Castle Walls, the Cloud Service Providers (CSPs) are getting their act together, and have realised that to break through they need to focus on the same person the IT Department serves - the End User - rather than just acting as an enabler for the IT Department itself. We're nearing the end of the Middle Ages, the weaponry is much better, and people no longer need to care for the whims of the King; they're leaving the Castle Walls, and building Settlements and Cities beyond the Castle Wall - beyond the (direct) influence of the King. Those Serfs, like our End Users, are doing this for much the same reasons - the Feudal System, within the Castle Walls, isn't serving their interests any more; they want more than to just abide by the rules of, and in the interests of, the King.

It's notable that now, things are fast-changing; a 2014 Release of the O'Reilly "Google Compute Engine" Book barely mentions any other Google Cloud Service, because few existed at it's release; not too long afterwards, rather than the thinking being "Cloud is just Compute", we exit the macrocosm of the this thinking, and create terms to describe other offerings that Cloud can fulfil:

  • IaaS (Infrastructure as a Service)
  • PaaS (Platform as a Service)
  • SaaS (Software as a Service)

Now we have a compass and framework to categorise - and more importantly, to understand and consume against, we start to see the "Shadow IT" trend begin; it's good to note the negative undertone to this statement is because it's the IT Department that coins this term, and it's chiefly a negative one because the IT Department, much like the King, knows the significance of the pull of the Castle Walls are coming to an end; if you ask the Serf how they fell about life outside the Castle Walls, it's more likely to be a positive term they'll use - not so much for the King at the fall of his dominance.

People run off, and much like Devolution within politics, effectively start doing their own thing - for mostly no reason other than just because they can now do their own thing - well kind of, but then they hit the brick wall the Castle had already solved: centralised control.

SAML becomes the fiat currency

If we go back within the Castle Walls, the currency you use to transact goods and services isn't fiat (Government-backed); and it doesn't really matter too much, as you're stuck within the confines of the Castle Walls anyway; if your Castle uses Guineas and the Castle over the road uses Florins, it doesn't matter to you, unless you have the privilege to both leave your Castle and enter theirs. If you're the King, however, it probably does matter to you, and keeping peaceful relations with your neighbouring counterpart. Stepping back into IT, it's not a true analogy to say we've got one "King", even within the times of the Castle - we all know IT is a silo of siloes, so we've probably got:

  • Microsoft Active Directory (AD) for universal End User Identity Management
    • It's likely everyone from the Janitor to the CEO has an AD Account, "just in case"
  • Oracle Access Manager (OAM) for certain End User Identity Management
    • It's unlikely John from Engineering Division has an OAM Account as well; it's more likely Jane from Finance has the unfortunate "fun" of having an OAM login as well as her AD login

We've probably cobbled together some OAM-to-AD bridge, or worse-yet just used the same 8-character Username for both OAM and AD logins and told people to use differing passwords for both; but if we don't, we can kind of lie to ourselves and consider it a Single Sign On (SSO) solution. Maybe. Anyway, we've got a fiat currency here - the AD Account. But it's only fiat within our Castle Walls; if we do a Merger or Acquisition, this becomes blatantly clear that our AD Florins do not match in value to the other Company's AD Guineas, and we go through some kind of painful process you might enliken to the introduction of the European Single Currency.

Shadow IT, however, breaks this; we're not in control of the SaaS or PaaS, or cobbled-together App the End User has created, and it's outside our Castle Walls anyway - so it might not honour the inner-Castle currency of our AD install. We as the IT Department would still like some control though, and the End User is now sick of doing all the boring moves-adds-changes and CRUD/ACID-like management - much like the Feudal Castles, we need a Bank of England to aribter between the disparate fiats. We need SSO. We need SAML.

Monarchy and the Republic

Security Assertion Markup Language (SAML)/SSO is a fantastic response to the IT Department at the death of the Castle; if we follow the principles of SaaS/PaaS well, we can use SAML/SSO to do our part in the devolved xaaS/Cloudy world - we can disaggregate some of the functionalities we'd previously have mandated (Compute/Platform/Storage/Network), and impart the skills we have honed over many, many years (CRUD/ACID/Security/Account Management) to best-advantage for the supposed "Shadow IT" our End Users have procured; we can enact power, influence and security over something we're happy is devolved to someone else. Indeed, it's telling that it's taken until fairly recent times for the likes of my logins to Cisco, Juniper, BT and so on to start using some form of SSO - just as "Big E" struggles with the journey, so too do "our" IT/ISP Providers struggle with the same "in house, everything is AD, those are the rules" versus "we enable via SAML, you do the rest" struggles; the current "Big E" version of the monolith versus microservice debate, if you will.

Wrapping back to Strategy, the important thing that SAML/SSO allows us to do is exert our best skillsets in the interest of End Users; prior to more-widespread adoption of SSO, the narrative ran that "Cloud will take our jobs!" and prior to that, the narrative ran that "Cloud is just someone else's Computer anyway!". But as our understanding and maturity evolves, so too does our understanding that there are some areas of specialism and complexity that it doesn't serve us well to even try and do - as AWS' Werner Vogels often states, there is little point in undertaking "undifferentiated heavy lifting" of aspects, skills or specialisms that we, as a larger Enterprise or Company, don't have - and frankly isn't one of our core competences. Much like many workplaces outsource the cafeteria to an outside concern that specialises in food preparation; why should we be concerned if an End User consumes WidgitCo Ltd's SaaS that does the job they need - as long as we can still exert some influence and control on the use of this, via a technology such as SAML or SSO.

In some ways, this is similar to the concept of Monarchy and Government, each acting as a balance to the other - but each far-enough removed from the daily running of the other so as to not interfere and conflict. Really, this is where "IT Strategy" needs to go - we need to care less about exerting direct control and management over the Compute/Storage/Networking building blocks, and step back and consider what the End User intent is, and why the lower elements of the stack need to exist to service the End User's needs.

ZTNA to the Future

Allow me to take my Enterprise Architecture hat off for a second, and apply this to the area I'm more au fait with - Network Engineering. In our world, Remote Access ("Dial-in VPN") has looked relatively samey in the past forever years:

  1. Nail up a Direct Internet Access (DIA) or two into the Enterprise WAN at various Office or Data Centre Points of Presence (PoPs)
  2. Implement a big ol' stack of expensive "Black Magic" Security Boxen (IDS/IPS/Firewall/<list goes on>)
  3. Implement a number of fleets of IPsec or SSL-VPN Concentrator Boxes
  4. Allocate the End User an OSI Layer 3 Address (IP Address) to connect them into the Network
  5. Implement a whole bag of distributed Access Control Lists (ACLs) and related to undo the Network-wide L3 Access we just gave them

Much like those who believed the earth was flat, before anyone had any chance to explore - or feedback through stories thereof - in yesteryear, this looks fine if it's what you're used to, and you're too busy doing other things to question it (a life in Network Engineering is one spent atoning - with time and suffering - for your alter ego's sins, I'm sure). However, it takes a change in perspective to make you realise the daftness of this model - step in Zero Trust Network Architecture (ZTNA), where instead we:

  1. (Either) Nail up a Direct Internet Access (DIA) or two into the Enterprise WAN at various Office or Data Centre Points of Presence (PoPs)
  2. (Or) Exploit our Cloud On-Ramps (Microsoft ExpressRoute/AWS Direct Connect/Oracle FastConnect/Google Cloud Interconnect) into the Enterprise WAN to act as PoPs
  3. Implement some virtualised "Black Magic" SASE Boxen
  4. SASE Boxen act at OSI Layer 4 and above, so End Users don't have an L3 Address (IP Address) that's natively on the network
  5. ...so we don't have to worry about ACLs and undoing the L3-wide access we (didn't) just give them

When you see this architecture, you quickly realise the floors of the former; why indeed would I give an End User complete L3-wide IPv4 Access to the Network, at the OSI Network Layer, when all they need are access to some Applications, which reside at the OSI Transport Layers and beyond? this is the SASE/Remote Access VPN-equivalent of using Stretch L2 VLANs in the Data Centre space "just in case it's ever needed"; a problem waiting for a solution.

Cloud and the role of Strategy

All of which is why the role of the "Strategy Department" needs to change it's thinking and remit from where it is today; the Castle's Walls are down, the Drawbridge is no longer the only way in; your edicts can't enforce ways of working on people; "Shadow IT" might not be as bad as you think, because chances are, we weren't doing the greatest job doing it to begin with. Much as in the concept of Devolution within politics, "with great power comes great responsibility" - and End Users are indeed learning from, and suffering because of, this very notion. Sure, they spun up some App somewhere and it was working fine, but now the adoption has grown to many more Users, and people are asking them questions about Compliance, and Remote Access and this "IP Addressing?" thing they don't know about has led to them appreciating, in some small respect, what the IT Department had to deal with.

Now is your time to either step up, and help them out - which is really what the concept of Strategy is all about (imparted learnings onto the masses, based on prior expertise) - or sneer loudly at them, which has two chief negative byproducts:

  1. They'll not like you for it (and likely go off and do the same thing again, with something else, if only to spite you)
  2. You'll have no role left to fulfil (if they've got the power due to xaaS and you're not helping, "why" are you still in existence?)

If this resonates with you, I'd suggest now would be the time to "Cloudify" your thinking and go with the flow, and use your knowledge and expertise for the good of everyone, and act less as a dictator of how to build low-level Compute/Storage/Network stacks, and more as an advocator for how to best-utilise the "Global Computer" that is Cloud and xaaS. Your role is no longer to tell people the exact model of car to drive; it's to educate them in how best to drive and maintain whatever they choose.

Otherwise, I can tell you about what happened to the Jester in the Court from the Middle Ages...