In my last post I discussed the pervasive issue of relational modeling as a (poor) substitute for proper domain modeling, the reasons why (almost) everyone continues to build software that way, and the resulting problems that arise for those with ambitions to move their relationally-modeled software system to the cloud.
Given the prevalence of RDBMS-backed enterprise software and the accelerating pace of public cloud adoption, chances are pretty good you’re faced with this scenario today. Let’s discuss what you can do about it.
Scenario: We Built Our App Against A Relational Model… But Now We Want It In The Cloud!
My first piece of advice is make sure you really need public cloud. The cloud has well-documented scalability, elasticity, and agility benefits, but most of those arise only for software designed intentionally to take advantage of them. They also tend to prove most cost-effective when amortized across a relatively long application life span, or across a relatively large (and, ideally, increasing) number of transaction requests. If yours is a legacy app with modest and relatively static resource needs, the cost justification for a public cloud deployment may not be so obvious.
It may, in the final analysis, not exist at all. Sure, you can leverage IaaS capabilities to spin up VMs and perhaps buy yourself a modest additional level of deployment flexibility. For some, that might be a reasonable and justifiable decision; perhaps capital expense constraints preclude reinvesting in updated hardware for the private data center. But let’s face it, a “modest improvement” in deployment flexibility is hardly the stuff of IT legend.
Instead of diving blindly into the cloud and/or immediately devoting resources toward a cloud-focused refactoring effort, first consider your reasons for wanting to adopt public cloud. If your application performs poorly on existing private data center hardware, the cloud isn’t pixie dust to wave over it and make it better. In fact, for some legacy codebases cloud hosting could degrade performance even further! You must understand what you’re dealing with; find a trusted software architect to assess existing issues with your application, determine root causes, and recommend mitigation strategies. Know where your big resource-hogging transactions are. Understand which tables in your current relational database are hotspots, and what you can do to fix that. Understand what happens front-to-back-to-front as users click through your web UI and wait several seconds or more for page navigation, data refresh, etc. In many cases, additional (if non-trivial) refactoring might be enough to squeeze better performance from your existing infrastructure. If yours is a multi-tenant architecture, consider sharding your database layer by customer to better accommodate heavy load. This may also put you in a better position to eventually migrate to the cloud at a more appropriate time.
Perhaps your legacy app runs well enough for current user demands, but you anticipate additional capacity needs in the near future. This isn’t a bad reason to consider public cloud; but are you sure your app will scale well horizontally? Recall that scale-out (horizontal scaling) is far and away the preferred scalability mechanism in the cloud; scale-up (vertical scaling) is also possible but tends to be very expensive in the cloud and is only a good solution in relatively narrow cases. If scale-up is your preferred (or only?) strategy, it’s entirely possible you can achieve your goals more cheaply by migrating to bigger iron in your private data center than by paying hourly costs for super-sized cloud VMs. Maybe… maybe not. Again, you may need to find a trusted resource to help you do TCO analysis of cloud and private data center options. Be informed. Don’t make knee-jerk decisions, in either direction.
So far it sounds as though I’m recommending against legacy app migration to the cloud. Far from it… done correctly and for the right reasons it can easily justify the needed time and resource commitments. If your due diligence indicates that a move to the cloud makes sense for your legacy application, your next step is to determine whether the IaaS or PaaS hosting model makes sense for you (I’ll skip over the choice of public cloud provider, for now anyway).
We can keep this short: you’re almost certainly going to want to start with IaaS virtual machines. PaaS hosting is a great model for applications intentionally built to leverage the cloud and comfortable with the constraints imposed by the PaaS provider of choice (usually things like choice of technology stack, deployment options, etc). PaaS can have a superior TCO story relative to the IaaS hosting model. But as you’re starting with a legacy application that wasn’t built for the cloud, PaaS may not be a deploy-and-forget scenario for you. To leverage PaaS you’ll need a horizontally scalable application that minimizes dependencies on custom OS configuration and file system access (custom Windows registry keys and reliance on specific or deep file system hierarchies are all red flags). You’ll need to understand the configuration, monitoring, and authentication/authorization subsystems of your PaaS provider and ensure your software plays nicely with those. You’ll also need to enable access from your PaaS-hosted app to your database; many PaaS solutions provide such behavior but only for a limited subset of technologies and configurations. Non-standard setups either require manual configuration or might simply be disallowed. In summary, a PaaS-hosted application is a cloud-ready application; getting your legacy app to that state is going to involve some effort. Thus IaaS will usually be a better starting point for existing apps.
Of course, VM hosting in the cloud can help you more easily scale up your existing relational database to the limits of your data size and query patterns (and budget ) and help you correspondingly scale up/out your application tier. To the extent this gives you more flexibility than you would have otherwise had in your private data center, this is a good thing. But if your resource needs continue to grow, you’ll eventually end up needing to refactor your database and your code to scale horizontally. Where to start?
Here’s where proper domain modeling is needed. There are a number of techniques for this; my personal favorite is Domain-Driven Design. Set aside your existing software for the moment and build a conceptual model of the business problem its solving... something a user of the system would understand. Know that this is no easy task; prepare to spend time and money to do it right. But the overarching goal is two-fold. First, develop this conceptual picture of your business problem independent of any storage-centric model used to represent data. This will give you technology-agnostic building blocks through which you can sort through challenging aspects of the system. Second, subdivide that large conceptual model into smaller, easier-to-digest sub-models that describe discrete processes within the larger problem domain. In DDD these sub-models are known as bounded contexts; they provide ambient, point-in-time meaning for actions and state changes within the system (you’re not just adding line items to a purchase order, you’re building the user’s shopping cart and preparing it for processing) and they’re bounded by a primary reason for existence (sort of a modeler’s take on the Single Responsibility Principle). A single conceptual model will consist of many such bounded contexts.
Bounded contexts become the unit of transactional consistency within a larger system; that is, state changes within a bounded context should be encapsulated within a single database transaction. Thus the use of bounded contexts tends to largely eliminate a big problem of monolithic relationally-modeled applications: the tendency for state changes to cascade across large swaths of the relational model, which results in very large (and slow!) transactions. Of course, the need will arise to propagate changes from one context to another; in DDD this often occurs as a formally modeled state change event (“item X added to shopping cart”) which can be handled by any other context which previously registered interest in such an occurrence. The handling of these events occurs beyond the scope of the initiating transaction, however. For more information on how this works, read up on eventual consistency (or drop me a line!).
The use of bounded contexts has an interesting side effect for cloud-based systems. By subdividing your application model into smaller, self-consistent sub-models with small transactions, you make it possible to deploy storage independently for each context. In other words, you get horizontal scale of your database as a fundamental part of your design, which is the fundamental building block of a cloud-native application.
In practice, multiple bounded contexts will store their state on the same machine (even the same database instance); but there’s nothing stopping you from using multiple database machines or even multiple database products across multiple machines, if the need arises from a scalability perspective. In fact, that’s precisely how cloud-hosted data stores scale to handle request levels far beyond what was considered “big” even five years ago. In the cloud, you don’t need (or even want) One Big Database.
So back to your existing application. Once you’ve done some domain modeling and subdivided that into manageable sub-models, consider the work you did to identify performance bottlenecks and database hotspots, and map that list of issues to the sub-models you created. Granted the lines won’t always be neat and clean, but the goal is to prioritize sub-models that you can migrate from the existing relational model to their own self-encapsulated ones. Start with the highest priority items and work your way down the list. With time and patience you can convert that legacy, monolithic enterprise application into a horizontally-scalable, cloud-ready fire breather. Just realize that there is no silver bullet here; application migration to the cloud is a process that rewards steady persistence above all else.
A minor disclaimer: clearly I’m glossing over the details of this process, and focusing principally on the database layer. This is intentional, but it does omit other interesting and relevant topics like security, integration, deployment mechanics, legal and economic considerations of the cloud, etc. I haven’t even scratched the surface of all DDD has to offer. Those are fun discussions too . Ping me if you’d like to talk specifics of your situation.
Next time, I’ll talk about ways to avoid this whole mess if you’re writing a new cloud-targeted enterprise app… we have the technology!
For CTOs looking to squeeze new life out of legacy enterprise applications, the cloud offers tantalizing prospects. Pushbutton scalability, reduced capital costs, outsourcing of non-core IT functions, potentially greater monitoring and health management capabilities and almost certainly greater overall uptime… even with the potential downsides, its no wonder senior management is tempted.
And yet those downsides are more than nagging problems; in many cases they pose significant barriers to a successful move from private data center to public cloud. An existing enterprise app might work fine running on internal hardware, with a modest user base… but move it blindly to a VM in Azure or AWS and suddenly that clever little accounting app grinds to a halt (as does your business). But why is that, exactly?
Where’s The Rub?
There are many potential difficulties to overcome when migrating an existing enterprise app to the cloud: reliance on past-generation (read: potentially unsafe) database or file-system drivers that might be ill-suited (or incompatible with) your chosen cloud stack, legal or regulatory requirements that mandate where the data lives, preconceptions about ambient hardware or network infrastructure baked (inadvertently, or otherwise) into your software, security contexts or sandbox privileges required for successful operation that may not be recommended best practices in the cloud, etc. Any of these (and many more) can trip up a migration effort. But there’s one incompatibility so pervasive that its worth discussing further, on its own. Its origins largely pre-date cloud computing itself, in fact. But the negative effects haunt us now, and we’ll likely continue to deal with them for years to come.
It’s your application’s data model.
It’s not the data itself… even if you’ve got a lot of that, there’s plenty of room to store it all in the cloud, if you want. And it’s not the application code per se, though it’s likely that you’ll need to change at least some of that to maximize the full potential of your cloud-hosted application.
No, what I’m talking about is the original conceptual model used to define the database underpinning your application. This model was probably created a long time ago… perhaps you paid a lot of money to a database architect who studied your requirements and used tools like Erwin or ER/Studio to make complex graphical depictions of tables and relationships, or maybe the model was defined by developers in code using APIs like Entity Framework or NHibernate. In either case, you could likely sit down with a developer on your team and have them walk you through the model, and you’d see elements of your business domain that you recognize… a Customer table, defined relationships between an Address and a Warehouse, etc. And this would seem logical and reasonable to you… the application performs some vital function for your business, as part of that it manipulates data, that data needs to live somewhere… voila! Here it is… in the database, created from this model.
A Minor Assumption, With Major Implications
The problem is that this model almost certainly has one very big assumption baked into it… it assumes there will be one physical database created from the model, and that all the data will live there. It is by definition a relational model… the concepts modeled within and their relationships (their “referential integrity”) can only be reliably maintained if the data is reasonably co-located, such that the database process can enforce transaction boundaries, data staleness and visibility rules, update query indexes as data changes render them obsolete, process complex joins across multiple tables, etc. Relational databases do not (cannot) reliably do these things across multiple machines. For a more detailed, nerdy explanation of why this is so, see CAP theorem.
In short, a relational model is predicated on the existence of One Giant Database. And unfortunately, sooner or later that single machine is doing as much work as it can do, but you need more. And now you have One Giant Problem.
To be clear, this isn’t really a cloud-specific issue. Any computational- or data-intensive resource (cloud-based or not) will eventually saturate. At that point, you have two options: scale up (buy a bigger server) or scale out (buy more servers). If the resource in question is an application server, either option (assuming your application architect is competent and anticipated scale-out scenarios) can work. But if the resource is a traditional relational database, you really only have two options: scale up and hope it’s good enough, or re-architect for scale-out. Sharding is sometimes a possible third option, sometimes a manifestation of the second… it has it’s place, but also enough drawbacks to make it unsuitable for the general case.
Scale up… or out… or ?
So for relational database scalability issues, scale up is usually the first consideration. More memory, more processors, more and faster disks… these will help your application serve more requests and handle more users, for a time. But you eventually bump up against the laws of physics. There is only so much RAM, CPU, and disk I/O you can bake into a database machine (physical or virtual, cloud-based or not). And even if your data access needs are within reach of the current technological state of the art, they may not be within reach of your budget (a quick perusal of AWS hosted database pricing shows a range from less than 2 cents/hour to over $7.50/hour… cha-ching!).
And so you’re left with the option of re-architecting for scale out. Scale out has two significant advantage over scale up. First, it’s theoretically unbounded; you can keep adding more servers forever. In practice this isn’t even remotely true, as life and software architecture inevitably intrudes and poses some actual upper bound. But still, it’s reasonable to say that a properly architected enterprise application can scale out much, much further than it can ever scale up. The second advantage is cost; scaled out solutions can be done incrementally, and with commodity hardware. This affords you the opportunity to purchase as much scalability as you need at any given moment in time. Scaled up solutions require the use of ever-more-expensive hardware resources, and perhaps worse, necessitate that existing resources must be retired or repurposed… with scale up, you can’t aggregate hardware to work cooperatively (which is exactly what happens in scale out).
But the big disadvantage of scale out is that you have to plan for it, architect for it, and choose technologies that enable it. And there’s the core issue with relational models and scale out; a relational model, and a database created from it, and likely the code written to work with it, are all fundamentally incompatible with any plan to scale out arbitrarily (darn that referential integrity!). Something will have to change, and that something will cost you time and money. There are limited options in products like SQL Server and Oracle for clustering a handful of machines together, but these tend to be used more in service to failover/reliability than pure scalability/availability needs.
A Storage Model Is Not a Domain Model
So, fine then… relational databases are incompatible with the preferred means of scaling cloud-based software (meaning, scale out). Relational models are poor but frequently used tools for modeling business domains, with significant negative implications for future scalability of the affected applications. But how did this happen? Didn’t we see this coming?
Sure we did. For years, smart people have implored us to stop using (relational) database models as the blueprint for software implementing non-trivial business processes. We just didn’t listen. Our tools (cough Entity Framework cough) make it easy to go from database to code, and while things like EF code first provide us with other modeling alternatives, many applications are still constructed bottom-up from a relational database model. Guess what? If you start with a monolithic relational model and auto-generate EF code to talk to that model, your EF code isn’t any more cloud-ready than your database is (to be clear, I like EF and think it’s entirely appropriate for use on constrained subsets of an otherwise large model, even in the cloud… it’s the naïve use of huge, monolithic EF models that I object to).
“But we’ve always done it this way.” Sure we have. In fairness, that’s not entirely our fault… the skills and tools needed to create a proper domain model independent of a dedicated storage model have for various reasons not yet gained broad traction. The path of least (initial) resistance is to start with a database and build upwards. Legions of enterprise developers have written code like this for years (decades?) and still do. I like to think we’re slowly moving beyond this, and I have high hopes for things like Domain-Driven Design, microservices architectures, and polyglot persistence as some of the practices and patterns that will help us break the cycle. More on that in my next post. But for now, we’re still a long way from industry-wide enlightenment.
Your Technical Debt Is Now Past Due
We’ve kicked this relational modeling can down the road for a long time, because we could. In a world of small private data centers, modest departmental application needs and manual, Excel-driven business processes, relational databases with relational CRUD-style code on top and built from relational models are not always great but are often good enough. It’s when our ambitions grow, and our anticipated use of these creaky enterprise apps grows along with them, that our best laid plans face the harsh reality of the technical debt we’ve incurred.
You want to move your IT infrastructure to the cloud? You want a more elastic, robust, flexible, agile infrastructure upon which to run your business? That’s a valuable goal. The cloud can give you that, and more. But make plans to retire that technical debt first.
In my next post, I’ll explore ways to do just that… we’ll talk about migration strategies for existing applications, and also touch on ways to minimize that technical debt in the first place.
I just got home from Devlink (unfortunately I had to bail out a day early) but I wanted to take a moment to say how impressed I am with the event, facilities, staff, and most important… the content! There were some excellent sessions throughout the week and my only regret in giving two talks of my own is that it left less time to soak up knowledge from everyone else. This was my first Devlink… it definitely won’t be my last. Kudos and sincere thanks to John Kellar and the Devlink board for putting on a great conference.
I had the pleasure of delivering two talks… “Node.js for .NET Developers” and “AWS vs. Microsoft Azure”. Both had great audience engagement and were lots of fun to deliver. I also did tag-team delivery of the all-day Microsoft Azure Pre-Con session with fellow Wintellectual John Garland, himself a fountain of Azure knowledge and all-around smart dude. It’s almost enough for me to forgive the fact that he’s a Florida Gator. Almost.
If you’re interested in the slide deck for my Node.js for .NET Developers talk, it can be found here. Likewise, the deck for my AWS vs. Azure talk is here. If you enjoy reading through them or have questions/comments/feedback, drop me a line at firstname.lastname@example.org. Always happy to talk Node, cloud, and other fun stuff.