Multiple container management platforms (Kubernetes, Mesosphere DC/OS, and Amazon ECS) have been adopted across the industry during the last two years, driving different benefits to a wide class of use cases. Additionally, a handful of web-scale companies have developed solutions on top of Apache Mesos to meet the unique needs of their organizations.
可以尝试看下apache mesos？
How is Titus different from other container platforms
-
tight integration between Titus and both Amazon and Netflix infrastructure
Titus has advanced ENI and security group management support spanning not only our networking fabric but also our scheduling logic. This allows us to handle ENIs and IPs as resources and ensure safe large scale deployments that consider EC2 VPC API call rate limits
同亚马逊的结合
We also worked with AWS on the design of IP target groups for Application Load Balancers, which brings support for full IP stack containers and AWS load balancing
-
We choose this path to ensure a common developer and operational approach between VMs and containers. This is evident through our Spinnakerenablement, support in our service discovery (Eureka), changes in our telemetry system (Atlas), and performance insight technologies.
-
scale
-
we run over a thousand different applications, with some being very compute heavy (media encoding), some being critical Netflix customer facing services, some memory and GPU heavy (algorithm training), some being network bound (stream processing), some that are happy with resource over commitment (big data jobs) and some that are not
we do not believe there are off-the-shelf solutions that can take on each of these scale challenges
-
We always try to maintain a philosophy of “just enough” vs “just in case” with the goal of keeping things as simple and maintainable as possible.
In the scheduling layer, we support advanced concepts such as capacity management, agent management, and dynamic scheduling profiles
-
-