Get Real with Hadoop: Rock Around the Clock!

Previous posts in this series provided insights about nine good reasons why customers choose MapR: from security to performance and TCO to ease of integration. All this is due to the superior architecture of the platform and the time and diligence MapR engineers have invested in the past five years, naturally by standing on the shoulders of giants.

Not very surprisingly, for more than 80% of our customers one of the chief reasons to choose MapR is that we offer business continuity out-of-the-box, rather than treating it as an nice-to-have add-on feature. Why is that so? Well, we’ll find out in the rest of the post …

What Does Business Continuity Mean (To You)?

In our always-on society, with the expectation of (online) services being available 24/7, it shouldn’t come as a surprise that continuing the operation no matter what is a fundamental requirement for the core place your data lives in. A disruption of the service you’re providing to your internal or external customers not only is an annoyance for your ops team (poor Fred got paged the second time this week at 3am) but really has deep impact on your business:

  • Short term: almost always means a loss of revenue; can be life-threatening (think: health care or automotive sector); customer satisfaction goes south, etc.
  • Long term: brand is damaged; customers lose trust in your capabilities to provide the service and churn increases; if you’re a startup your VCs get nervous and if you’re a public company you might not look forward to your quarterly earnings call, etc.

OK. I think we can agree that business continuity is essential. But what exactly is business continuity? To understand that, we have to take a step back. Generally speaking, there are two types of outages that one has to tackle in order to provide business continuity-related Service Level Agreements (SLAs):

  • Minor outages: a single node or component of a node such as a disk failing, network partitions, upgrades of services and/or apps, firmware bugs in routers, etc.
  • Major outages also known as ‘disaster strikes’: fire, tornado, flood, power grid down, etc.

Based on above classification—borrowed from Klaus Schmidt’s seminal book High Availability and Disaster Recovery—we can now appreciate High Availability (HA) as the capability of the data platform to protect and recover from minor outages, while we understand Disaster Recovery (DR) as the ability of the platform to recover from and resume business after major outages as quickly and cost-efficiently as possible.

Can Business Continuity Be Optional?

The basic question we asked ourselves at MapR was: in terms of business continuity, why expect less from a big data platform than you do from the system that currently powers your operation (such as a single-node RDBMS)?

We believe that business continuity SLAs are critical for Hadoop production deployments and this is why we provide 99.999% uptime. With the no-NameNode architecture, true snapshots and mirroring capabilities–as well as rolling upgrades–our platform is built to uniquely provide self-healing from multiple cluster failures along with HA for YARN jobs as well as MapReduce v1 jobs. Also, the business continuity extends to NoSQL database operations as well as integration aspects, i.e., NFS access.

How do we do this? In a nutshell, the main reason we can provide business continuity features out-of-the-box is due to the unique architecture of the MapR platform: distributing file system metadata in shards across the cluster, co-residing with the data with which it is associated, allowing instant recovery, with files and tables available rapidly after node failures or cluster restarts. Further, if nodes fail while MapReduce batch jobs are running, MapR lets the jobs run to completion, rather than requiring a complete restart from scratch (meaning: you can meet your SLAs).

If you want to learn more about the business continuity capabilities of the MapR platform, I suggest you have a closer look at the HA tech brief as well as read the blog post my colleague Bruce Penn wrote, earlier in this series: Complete Data Protection and Disaster Recovery.

Don’t Take Our Word For It …

Vendors are of course free to make claims about the business continuity features of their respective offerings, however, I personally find customer testimonials the most convincing way of communicating the true capabilities—so, let’s have a look at a few:

In summary, business continuity is not a nice-to-have add-on but a critical characteristic of a big data processing platform. MapR is the only Hadoop distribution that provides you with both HA and DR capabilities out-of-the-box and that is also the #1 reason customers decide to run their applications on our platform.

Be sure to check out the complete top 10 list here.

no

Streaming Data Architecture:

New Designs Using Apache Kafka and MapR Streams

 

 

 

Download for free