DCaaS - new service model
Good day, $ habrauser!
A conference of DevConf 2014 professional developers will take place very soon in Moscow. Within the framework of this conference, I will be happy to tell you our vision of a new DCaaS model - DataCenter as a Service. How much we have changed the approach to the infrastructure is up to you, we just did it in the way that is convenient for us and our clients. We do not pretend to new standards, just this implementation fits this model very much.
')
What we [considered | studied | used] when building a model:
1. At the equipment level, Canonical MaaS , part of the Mirantis Fuel package , which manages the input to the “battle” —the actual hardware deployment
2. At the level of building infrastructure - kickstart, cobbler , the model of the same Mirantis Fuel
3. At the level of infrastructure management - OpenStack and OpenNebula
What is the purpose:
To give the client the ability to independently manage resources, ranging from a “bare” server installed in the data center to a fully out-of-the-box combat infrastructure with full control over the web. Divide the roles that the client can assign to the equipment — from a dedicated server to a virtual infrastructure element — a compute node, a CEPH storage ... and incorporate the server into an existing infrastructure or create a new one. Give the opportunity to create your little Amazon, well, or several, within the framework of this model based on OpenStack.
What we did:
1. At the first stage, the client receives a list of “bare” physical servers of various configurations that stand and wait for the client to perform one or another installation script, disk partitioning and installation.
2. Here we divide into 2 main steps of the script - or a dedicated server - everything is simple - you choose the OS and after a few minutes the server is ready for use with the certificates and password hash (which you previously installed in the panel) or the second script, which much more interesting - you choose a model for creating your own virtualization infrastructure. In this case, of course, one physical server will not be enough for you.
3. Own virtualization infrastructure: in this scenario, the minimum number of physical servers = 2. The first server is the management server, the second is the compute node. This is the minimum set. There are options with HA - in this case - it is at least 2 management node and 2 compute node. In the case of dedicated data storage CEPH = 6.
4. Installation of the virtual infrastructure takes from 25 to 40 minutes. After completing all the installation scenarios, you get a full-fledged control panel of your own small Amazon with storing images, snapshots, managing internal users, allocating resources, editing network infrastructure and other features.
At DevConf 2014, I will talk in great detail about the development of this model, technical aspects, hardware requirements, and also about the development plans of the project.
Looking ahead a bit, I’ll say that we are actively working on the issue of migrating a virtual machine image to a physical server, which, in my opinion, is very much in demand by startups that go to a good production.
See you at DevConf 2014 !
A conference of DevConf 2014 professional developers will take place very soon in Moscow. Within the framework of this conference, I will be happy to tell you our vision of a new DCaaS model - DataCenter as a Service. How much we have changed the approach to the infrastructure is up to you, we just did it in the way that is convenient for us and our clients. We do not pretend to new standards, just this implementation fits this model very much.
')
What we [considered | studied | used] when building a model:
1. At the equipment level, Canonical MaaS , part of the Mirantis Fuel package , which manages the input to the “battle” —the actual hardware deployment
2. At the level of building infrastructure - kickstart, cobbler , the model of the same Mirantis Fuel
3. At the level of infrastructure management - OpenStack and OpenNebula
What is the purpose:
To give the client the ability to independently manage resources, ranging from a “bare” server installed in the data center to a fully out-of-the-box combat infrastructure with full control over the web. Divide the roles that the client can assign to the equipment — from a dedicated server to a virtual infrastructure element — a compute node, a CEPH storage ... and incorporate the server into an existing infrastructure or create a new one. Give the opportunity to create your little Amazon, well, or several, within the framework of this model based on OpenStack.
What we did:
1. At the first stage, the client receives a list of “bare” physical servers of various configurations that stand and wait for the client to perform one or another installation script, disk partitioning and installation.
2. Here we divide into 2 main steps of the script - or a dedicated server - everything is simple - you choose the OS and after a few minutes the server is ready for use with the certificates and password hash (which you previously installed in the panel) or the second script, which much more interesting - you choose a model for creating your own virtualization infrastructure. In this case, of course, one physical server will not be enough for you.
3. Own virtualization infrastructure: in this scenario, the minimum number of physical servers = 2. The first server is the management server, the second is the compute node. This is the minimum set. There are options with HA - in this case - it is at least 2 management node and 2 compute node. In the case of dedicated data storage CEPH = 6.
4. Installation of the virtual infrastructure takes from 25 to 40 minutes. After completing all the installation scenarios, you get a full-fledged control panel of your own small Amazon with storing images, snapshots, managing internal users, allocating resources, editing network infrastructure and other features.
At DevConf 2014, I will talk in great detail about the development of this model, technical aspects, hardware requirements, and also about the development plans of the project.
Looking ahead a bit, I’ll say that we are actively working on the issue of migrating a virtual machine image to a physical server, which, in my opinion, is very much in demand by startups that go to a good production.
See you at DevConf 2014 !
Source: https://habr.com/ru/post/225587/
All Articles