What is actually under the hood of your cloud? Here you’ll read some in-depth technology considerations to cover when you are evaluating your cloud hardware.
Physical resources are the hardware devices that are used to create virtual resources. Physical resources include compute, storage nodes, and the network the resources use to communicate between themselves and the Internet.
- Node – A node is a physical machine dedicated to either running virtual machines (a compute node), storing data (a storage nodes) or, like in the case of Flexiant Cloud Orchestrator, routing network traffic (a router node). You’ll need to consider if you are using a centralized or localized storage model as this will impact your required disk space vs. each node.
- Cluster – Normally a cluster consists of a collection of one or more nodes, where all the nodes within the cluster share the same physical location, network and hypervisor type. If you want to have two different hypervisors types or have presence in different geographic locations, you will probably want to break your nodesinto more than one cluster. You’ll need to consider how you manage multiple clusters.
- Hypervisor – Your hypervisor choice will affect the types of network and storage you can use. Will you opt for an open source hypervisor like KVM or Xen4 or commercial ones like VMware, Hyper-V or Parallels Cloud Server (PCS) for example? With commercial hypervisors, don’t forget to think about the licensing and costs for these. To learn more about how to select a hypervisor, you can read our paper on the topic here.
- Network – The method the virtual resources in a cluster use to communicate with each other and with the Internet. This is different from the physical network that connects your compute nodes to each other within a cluster. You’ll need to consider what networking modes for virtual resources you want to use. Two examples are:
- Single shared VLAN. Remove the complexity of layer-2 network management and make all your customers share the same VLAN. If it sounds unsecure, there are technologies that solve this problem by moving the layer-2 isolation at the compute node level. Flexiant Cloud Orchestrator with its Public Virtual Internet Protocol (PVIP) gives the ability to assign a customer a single public IP at a time, while retaining the benefits of full isolation on a per server basis.
- One or more VLANs per customer – With traditional VLANs, customers can self-serve real layer-2 networks that have a finite number per cluster and require a block of dedicated IPs (subnet) to interconnect to other networks. Use of this networking mode requires routing to be setup to correctly determine where the network traffic for running virtual machines should be sent.
- You can read our paper on ensuring your network can scale to support your cloud business.
- Storage – If storage is persistent, data stored by a virtual machine is retained even when the virtual machine shuts down or loses power. The storage resource can be either a central resource or located on compute nodes as localized storage. Each of these options has its own advantages and disadvantages which you’ll need to consider:
- Centralized storage gives your platform increased resiliency and reduces the disruption to customers in the event that a compute node loses power, but hardware can be expensive and this may create scaling bottlenecks.
- Localized storage is less resilient per se as, if a compute node loses power, the storage device must be removed from the node and placed in a working node before customers can access their virtual machines again. Localized storage has the advantage of being cheaper and can be more performant in some use cases where high throughput is required.
- Distributed storage is an attempt to offer the advantages of centralized storage with the scalability and cost base of local storage. Distributed block storage is a technique for distributing the storage requirements of your virtual machines across a large number of commodity storage nodes, that in some cases can be compute nodes themselves equipped with local storage devices. For example, Ceph is an open source distributed software-defined storage technology that is beginning to be used in a variety of cloud orchestration systems. It has the most market and technology traction of any of the software-based distributed storage technologies with a good community of support.
You can also download our ultimate guide to cloud – Zero to Cloud – a one-stop guide to helping you whatever stage you might be at on your journey to or in the cloud. It includes more considerations for the business, technology, cloud management software, cloud deployment and customer acquisition. Download our Zero to Cloud checklist now.