Engineering Cazena’s Big Data as a Service: A Year in Perspective

[field_blogimage]

January 15, 2019

By: John Piekos, VP Engineering

When you deliver product incrementally, in an agile fashion, it is often hard to see, week to week, sprint to sprint, all the great progress made throughout the year. That’s why, at the end of every year, I enjoy looking back at what my teams and company have accomplished the past twelve months. For Cazena it was an exceptional year with triple-digit growth of our customer subscriber base.  Behind the curtain, the data paints an interesting picture of what it means to provide this growing customer base with a fully-managed, secure, highly-available big data platform.

Cazena offers a single-tenant Big Data as a Service, which embeds best-of-breed software from Cloudera, Azure, AWS and others. Enterprises use Cazena solutions for fully-managed data lakes, analytics, data science/ML sandboxes and data engineering. Cazena’s model means that each customer has their own provisioned set of cloud resources (compute and storage), with no co-mingling of one customer’s data with another.  Our customers are located all around the world, with production deployments in two (2) cloud providers, AWS and Azure, across seven (7) geographical regions.

Looking back at Cazena in 2018 here are some interesting stats on how enterprises use Cazena’s fully-managed Big Data as a Service.

Analytics Workloads Run

Cazena is accessed with a variety of tools and methods, depending on the dataset and use case. Some use embedded SaaS interfaces for SQL, R, Python via Cloudera Hue, others connect custom analytics applications or their preferred data science and data engineering tools.

~1 Million / Month - The number of big data and analytics jobs, on average, executed by Cazena customers, every month of 2018.

Each of our customer’s data environments contains hundreds of big data and analytics workloads, consisting of Spark applications, Spark streaming jobs, SQL or Hive workloads, R jobs and more. Cazena ensures that SLAs are met for each workload -- and helps teams deliver value every day by ensuring consistent access to historical and operational analytics, data processing and real-time decisioning.

Customer Support and Data Operations

Cazena customers enjoy a fully-managed “white glove” service, relating to every aspect of our cloud solutions (data lakes, data science sandboxes, etc.). Part of our offering includes a comprehensive support function, including architecture, big data job advice, user management, security and permissions, performance and resource utilization and monitoring. 

1,100 - The number of solved Cazena support requests in 2018.

Breaking this number down, 35% were typical helpdesk requests, 28% were “how to” inquiries for data lake expertise, 32% required assistance with big data job problems, system performance and resource utilization optimizations, and the occasional cloud-provider induced machine failure or network issue.

Cloud DevOps, Service Health & Availability

Cazena’s robust monitoring and alerting watches over all of our production systems. Our monitoring tracks the availability of the system, including cloud infrastructure, network, virtual machines, and big data services, for each of our customers. Resource utilization such as CPU, disk and job performance are also monitored and can trigger alerts.  Most of these alerts are handled automatically by Cazena’s systems without human intervention.

4,750 - The number of production system alerts addressed in 2018.

Cazena’s alerting system goes one step further, however. We recognize that each customer’s business relies on more than just system health: if their important jobs aren’t running for any reason their business could come to a standstill. As such, we implement custom monitoring, specific to customer requirements, to alert both us and the customer if something is amiss with key jobs. One example is an alert for a specific Kafka real-time ingestion job stall or failure. If this job fails, the data lake, though running and operational, won’t be providing real time analytics. Cazena is often the first to learn of these failures and can alert our customer and work with them to remedy the situation.   

Production Operations & Hadoop Platform Administration

Cazena’s cloud data and analytic solutions power our customer’s businesses, so any software upgrade must not disrupt any enterprise workflows.  Each new version of Cazena includes security patches, defect fixes, new capabilities, performance and resource usage optimizations and often, new versions of software, including Cloudera CDH. 

39 - The number of software upgrades performed on our customers’ data environments in 2018.

There are a lot of moving pieces. Cazena Engineering tests and validates new releases on multiple cloud providers across multiple cloud regions. Because our customers view analytics as a mission-critical capability, we ensure that all big data jobs continue to run successfully after an upgrade.

End-to-End Security (Across Cloud, Cluster, Data, etc.)

Cazena deploys end-to-end, fully-secured data solutions. The network is locked down. Data is encrypted in motion and at rest. Users have authenticated and authorized access to the data.  Security is considered in every aspect of the architecture, evident in all solutions. For example, in addition to anti-malware and antivirus software, each machine in a Cazena data lake runs intrusion and anomaly detection software. Security scans are run weekly on every customer deployment, and every new release of Cazena delivers the latest security patches to ensure the security of these systems.

1,749 - The number of software components and packages updated in 2018 to address security vulnerabilities identified in 329 CentOS Errata and Security Advisories (CESAs) applied to our production systems.

Every day new security flaws are detected in existing software and systems throughout the world. Cazena diligently monitors and addresses identified security vulnerabilities of the underlying operating systems, kernels and software packages. We also have an extensive security issue monitoring system, which means that our customers don’t have to develop and maintain their own. 

350,483,925 – In 2018, over 350 million security events were captured, analyzed and stored for auditing, diagnostics, and compliance alerting.

Saying we’re secure and proving it are two different things - that’s why we capture a full set of security events for full traceability and auditability of each of our customer’s production data environments.

Cazena Engineering   

Behind the curtain at Cazena is a passionate and dedicated agile engineering team with decades of data and devops expertise. The team is spread across three states in the US and two countries.

81 - The number of sprints completed by the Cazena engineering team in 2018.

13 - In 2018 we released 13 versions of the foundational Cazena data platform for data lakes, data science/ML and analytics. Two of our 13 releases in 2018 featured a new version of Cloudera. The Cazena engineering team fully vets and tests each upgraded version of Cloudera, including performance in the cloud, before rolling it out to our customers, so customer teams don’t have to. 

605 - The number of JIRA tickets closed by the Cazena engineering team in 2018.

The Cazena engineering team tests our offering in three (3) cloud providers across seven (7) regions. Cazena tests in more cloud providers than we have in production. Additionally, you’d be surprised (at least I was) at the difference in behavior between regions within the same cloud provider.

Thank You!

Much of Cazena’s success this past year is due to the commitment and hard work of the Cazena Engineering and Support teams. We are also grateful for the support and collaboration with all of our technology service providers, analytics partners and our enterprise customers. Thank you for an amazing year!

While 2018 was a great year, 2019 promises to be even better. As we enter the New Year, the Cazena team wishes you and your teams health and happiness -- and no wasted time or failure on big data projects, of course!

If we can help you in any way with your big data initiatives in 2019, don’t hesitate to reach out to us.

Back ›