• No se han encontrado resultados

Análisis de la evolución del discurso de los líderes del PSOE

2. Análisis de la evolución de los discursos de una misma lengua

2.2. Análisis de la evolución del discurso de los líderes del PSOE

OpenStack isn’t a virtual switch, but it does manage several physical and virtual net- work devices and virtual overlay networks. Unlike the OpenStack control of hypervi- sors, which is limited to the services provided by the hypervisor alone, OpenStack directly provides network services like DHCP, routing, and so on. But much like hyper- visor management, OpenStack is agnostic to the underlying vendor technology, which could be a commercial or open source technology.

More importantly, a backend technology change like moving from one type of network or vendor to another doesn’t necessitate client configuration changes. Given the great deal of proprietary hardware, software, and user interfaces involved in net- working, it’s often not trivial to switch from one vendor or technology to another. With OpenStack, those interfaces are abstracted by the OpenStack API, as shown in figure 1.6.

OpenStack can manage many types of network technology (mechanisms), includ- ing those provided by Arista Networks, Cisco Nexus, Linux bridging, Open vSwitch (OVS), and others. Throughout this book we use networking services provided by

Linux containers

Recently, there has been great interest in the use of operating-system-level virtualiza- tion in place of infrastructure-level visualization for providing OpenStack instances. Operating-system-level virtualization provides the ability to run multiple isolated OS instances (containers) on a single server. But it isn’t a hypervisor technology—it oper- ates on the system level where containers share the same kernel. You can think of containers as providing virtual separation where needed, without the emulation over- head of full virtualization.

The two most popular operating-system-level virtualization projects are Docker (https://www.docker.com/) and Rocket (https://github.com/coreos/rkt). Although it can be argued that containers are better suited for application runtime delivery than OS-level instances, technologies based on containers will undoubtedly be widely adopted for building clouds.

OpenStack and OVS. OVS is a common switch choice for OpenStack that you can easily obtain and replicate in your environ- ment without specific hardware require- ments. Along with the network mechanisms, there are several network types (VLANs, tunnels, and so on) sup- ported by OpenStack, and those are cov- ered in chapter 6.

1.3.3 OpenStack and storage

OpenStack isn’t a storage array, at least not in the way you’d generally think of storage. OpenStack doesn’t physically provide the storage used by virtual machines.

If you’ve ever used a file share (NFS, CIFS, and the like), you were using “file- based” storage. This type of storage is easy for humans to navigate and computers to access, but it’s generally an abstraction of another type of storage: block storage. You can think of operating systems or filesys- tems as being the primary users of block storage.

There’s also another type of storage that people in systems roles might not be familiar with: object-based storage. This type of storage is generally accessed through software APIs (for example, GET/obj=xxx). Object-based storage is a further abstrac- tion of file and block storage, but without the restrictions of either. Object-based stor- age can be easily distributed and replicated between many participating nodes. Unlike block storage, which must quickly be accessed by a VM, distributed object storage is much more latent and wouldn’t be used for things like VM volumes (which are actively attached to an instance). It’s common to use object storage to store backups of vol- umes and images (containing operating systems) to be applied to volumes when they’re created.

Let’s first address how OpenStack works with block storage, and then we’ll talk about object storage.

BLOCKSTORAGE

OpenStack doesn’t currently manage file-based storage for end users. In figure 1.7, you can see that OpenStack manages block (VM) storage much like it manages the hypervisor and network.

This figure shows a complete picture from a basic VM resource-management pro- spective. OpenStack can manage many vendor-provided storage solutions, including solutions from Ceph, Dell, EMC, HP, IBM, NetApp, and others. As it does with hypervisor

OpenStack Vendor network VM network VM compute Networking Server hardware

and network components, OpenStack pro- vides you with the flexibility to switch between storage vendors and technologies without changing the client configuration. OBJECTSTORAGE

Although OpenStack isn’t a storage array for block storage (used to boot VMs), it does have the native ability to provide object storage. Other than physical hard- ware running a supported version of Linux, no other software is required for OpenStack to provide a distributed object storage cluster. This type of storage can be used to store volume backups, and it’s also common to use object storage for large amounts of data that can be easily isolated into binary objects. Figure 1.8 shows a basic object server deployment that’s all contained in the OpenStack environment.

Vendor network VM network VM compute Vendor storage system VM volume Server hardware OpenStack Storage Networking

Figure 1.7 OpenStack manages block (VM) storage. OpenStack Proxy node Storage node Storage node Storage node

Figure 1.8 OpenStack provides object-based (API) storage.

Object storage doesn’t have to be in a single location. In fact, nodes (proxy and stor- age) could be in several locations that replicate between each other.

The traditional use case for object storage is to store data that will be accessed by an application, such as a document or file that will be used by an application on a user’s behalf. There are several use cases for object storage in OpenStack environ- ments. For example, it’s common to use this type of storage as a repository for VM images. This isn’t to say that the VMs use this storage directly, but they’re provisioned from data maintained in this system. This is reasonable because the provisioning pro- cess won’t need low-latency access to random data. Object storage might also be used to back up a snapshot of an existing VM for a long-term backup.

Documento similar