Introducing Jazz: T-Mobile’s Serverless Development Platform
Ship production code with confidence to FaaS platforms & more!
Serverless is fascinating! No server setup, no failover management, no scaling issues, no infrastructure operations — Just the code! We’ll not get into the definitions, benefits, serverless trends etc — lots of great speakers, authors and writers have shared their excitement around serverless and a simple google search can answer most of the basic questions around serverless. Instead, we’ll focus on the serverless story in our organization, observations, learnings & the work we have been doing in this space.
We have been thinking serverless at T-Mobile too! Our serverless journey began after re:Invent ’15 showcased few exciting use cases with AWS Lambda and other serverless services in AWS. We, cloud team at T-Mobile, identified couple of use cases that fit very well with event-driven computing model to understand Lambda and find the hidden gems. As expected, implementations for both these use cases (just code!) are working perfectly well till date and costing us under $10 compared to thousands if we have used EC2 instances to host them instead. We have started observing patterns where developers are taking serverless first approach; all new services are being built in serverless fashion unless there is an obvious reason not to go serverless. This is paving the way towards building next generation infrastructure that is application first rather than building applications that conform to the infrastructure that was built already.
Not just the cost factor, the idea of not managing servers running these applications, built-in scaling, high availability, resiliency (and more) excited us and we have started building complex event-driven applications. We have seen developers writing more and more of these ‘Lambda’ functions, creating ‘API Gateway’ endpoints, ‘DynamoDB’ tables etc. which is when we have started seeing ‘service sprawl’. Service discovery became harder. There was no auditability. Cost control and security guardrails were not in place. Developers weren’t able to monitor their services easily, getting to their service logs was getting difficult, lack of continuous integration started increasing the time to market for their features. Overall, gaps within the existing serverless tooling was clearly felt and understood.
Let’s dive into few of these developer pain points in more detail; this is important because some of these led to teams questioning the readiness of serverless services for enterprise adoption.
- New Tools: While serverless promised the notion of ‘writing code and forget everything else’, using these serverless services is not straightforward. Code once ready has to be deployed into serverless infrastructure which should be provisioned using cli/api or new serverless deployment tools. Each of these tools requires developers to understand how they work and sometimes require changes to their application which is not appreciated by the application developers.
- CI/CD: Most of the available serverless deployment tools (open-source/commercial) addresses major concerns related to easier/faster development and deployment. However, integrating those with existing enterprise CI/CD tooling is required to comply with the enterprise standards. This resulted in more platform integration work for developers instead of focusing on their application.
- Monitoring & Logging: Cloud providers have their own implementations to support monitoring & logging. Developers always wanted an easier and quicker access to their application logs preferably at a location that they are already comfortable with (typically Splunk, Elasticsearch etc. within enterprises). While it is not impossible, developers have to spend good amount of time to understand and configure these details themselves. This is not considering the fact that cloud providers do change (frequently) lower level details on how they implement logging (log formats for example).
- Security: While building APIs in 2 minutes appear exciting, today’s serverless service implementations like AWS API Gateway may leave us with a greater exposed attack surface. Adding the security controls is required before we build a production grade serverless application using these services. Continuous compliance should be enforced to ensure services are always secure. Leaving these details to developers might leave the surface open for attacks.
- Multi Cloud: Most of these pain points multiply when developers want to use more than one cloud providers to host their applications built using serverless architectures. In fact, these are not easy problems to solve even when they try to use multiple accounts within a single cloud provider alone (in case of AWS).
- DRY: All in all, when developers started writing these smallest components of their applications — now called functions — sharing best practices within teams implementing decoupled services just got more difficult which led to duplication.
We spoke with various serverless enthusiasts across the organization (and outside) to discuss these gaps. What developers needed was an easier way for them to create, deploy, manage (serverless) services with monitoring, centralizing logging, stricter security guardrails and complete visibility on the spend. We wanted application developers to truly focus on their application only— their code — and leave all other issues discussed above to a platform that can solve these pain points for them.
Jazz to the rescue!
Jazz allows developers to seamlessly create, deploy, manage, monitor & debug cloud native applications leveraging serverless services. Jazz helps developers focus on code only and solves most of the operational issues around deployment/management. Jazz provides cleaner reusable code templates with all the best practices embedded into them providing the required quick start for developers. With code commits triggering workflows to do rest of the action, a suite of operational tools built into Jazz helps developers securely manage code, configurations and deployments for their cloud native applications. Jazz comes with CI/CD by default so that all the manual, error-prone build and deployment work is completely automated resulting in high quality software with continuous integration, automated tests, code metrics etc. It helps developers focus on continuous code quality & find vulnerabilities in the code before the code goes live! We have started identifying patterns for type of cloud native applications that developers started building and created reusable code templates for these service patterns (APIs, functions, static websites, workflows etc.) so that they can get going in seconds. Pay-per-use makes serverless applications appear insanely cheap. Jazz helps developers to see their spending trend on a per service and environment basis so that they avoid billing mishaps that are not uncommon. This immensely helps developers with not much experience in server management and administration to focus on their code and deliver value to their business. More about Jazz’s architecture will be discussed in the future posts.
Jazz comes with a beautiful user interface for developers to easily create & manage their services (Jazz supports APIs, functions & static websites as of now). Today, Jazz is helping our developers within T-Mobile build production ready applications based on event driven architectures in minutes!
Watch the video preview here!