Some ideas about Apache Mesos

[Editor's Note] This article briefly introduces Mesos and Myriad projects that blend Mesos and YARN. The text has a nice metaphor to explain Mesos 'resource handling, and Mesos' resource processing is like parenting a child's birthday party …

I am concerned about Apache Mesos for a long time. Apache Mesos started from the research paper , becoming the Apache incubation project in 2010, later "graduating from ASF" and establishing commercial entities in 2013, Mesosphere .

Over the past few months, there have been many things, so I thought it was a good opportunity to write articles about Mesos and its ecology .

There have been a lot of discussion about Mesos and YARN. I have also seen comments such as " Mesos' resource request model is very backward " and also notes that Mesos has become more popular in the past few years. One of the key factors here may be Docker's hype and their needs for collaboration . At the end of this chapter, we will return to Mesos vs. YARN again.

I admit that I did not fully understand the real potential of Mesos until I sat on that day and finished the Mesos research paper , which included design philosophy, resource allocation, isolation assurance and fault tolerance.

The core challenge of Mesos is that it does not understand the framework of a constraint on how to satisfy the constraints of the framework , which is the most difficult to understand the distribution of resources. Mesos handle the way things are like parenting a child's birthday party: like you have to provide food for the 15 children (==), and you can not know their preferences (== placement tendencies). But you can offer them a piece of pizza or a bowl of sesame, and they can be free to accept (now or after a) or refuse. Moreover, the father who gets off the bus is likely to tell you that the child is a vegetarian, and it is unreasonable to provide beef burger (== filter) to the child.
There is an interesting fact (although I think it is well known), Mesos and Spark have one thing in common: Matei Zaharia – from a small town near Ontario, Canada – he is a student at the AMP laboratory at the University of California at Berkeley. Room for Mesos and Spark have made a great contribution. Recently, he served as Databricks CTO, Databricks is a guide to Spark's business entity company.

Back to Mesos vs. YARN – Fortunately, this is no longer a matter of choice: recently using the Myriad project (co-developed by eBay, Mesosphere and MapR, now hanged by ASF), you can make them Cluster coexistence and scheduling them. In short, a Mesos framework is used to dynamically extend YARN clusters and supports running Hadoop applications such as Spark and non-Hadoop applications such as Node.js, Memcached, RoR, and so on. Exciting moments!

This is my personal view of Apache Mesos, written in mid-February 2015. I will continue to focus on Myriad, as a beginner if you have not tried, maybe you can try to test drive Mesos .

Source: Thoughts on Apache Mesos (Translator: Sun Ke School: Li Yingjie)

Translator introduces a student, line code site maintainer, focusing on Linux system programming and compilation techniques. Keen on learning a variety of languages ​​and technologies. Is currently developing a Java Web framework (BI) and a web server (D language).

    Heads up! This alert needs your attention, but it's not super important.