Primary storage: what transition from on-premise to the cloud?

Most enterprise applications today rely on SAN and NAS storage arrays to store their data. These applications rely on traditional architectures and do not have the resiliency attributes of modern applications such as native cloud applications.

When migrating to the cloud, care must be taken to replicate the characteristics of the infrastructures on which they depend (in the case of a “lift and shift” approach) or to find alternative services in the cloud in order to reproduce their operation.

Another important issue is that of the more or less transparent organization of data migration.

Hybrid cloud homogeneous: the easy way

The simplest solution for migrating an on-premise cloud application is to transport it to a target whose technology is identical to that of the source platform. For companies that are operating a virtualized infrastructure with VMware, for example, it is possible to opt for a VMware Cloud Foundation powered cloud service, such as the one offered by IBM or the one that VMware markets on the Amazon AWS platform.

That of IBM has the advantage of being already widely available internationally, while VMware on AWS is only available in the United States (but the service should expand its coverage this spring). OVH is also expected to further develop its VMware private cloud offering to VMware Cloud Foundation. This offering builds on VSAN hyperconveraged storage and therefore has most of the traditional storage attributes of advanced services, while having scalability attributes similar to those of a cloud service.

The advantage of a homogeneous approach is that it allows hybrid cloud deployments driven from a hybrid console. It is even possible – provided you have a sufficiently efficient connection – to migrate applications “live”, that is to say without production stop, between the infrastructure on premise and the cloud.

With Nutanix Xi and Azure Stack, Nutanix and Microsoft have the ambition to offer the same kind of simplicity.

This is also the case of Oracle, which allows transparent migration of its Cloud @ customer offer to its cloud. Or French OutScale for customers who deploy their offer both private mode on-premise in public mode.

The benefit of such an approach is that it can migrate its applications quickly in a public or private cloud without fundamentally changing the way one operates its applications, or reforming its staff. The company no longer has to manage the infrastructure of migrated applications in the cloud and can convert all or part of its spend into OPEX mode.

The only downside, by choosing this approach, we do not really benefit from all the benefits of a real cloud especially in terms of flexibility and elasticity.

Lift and Shift approach

The other way to migrate an application to the cloud is to adopt a “lift and shift” approach. The goal is to recreate an infrastructure in the cloud that has similar characteristics to on-premise applications, then migrate application-by-application data to re-instantiate progressively over a different infrastructure than the infrastructure. origin.

For applications virtualized with VMware, multiple solutions exist to migrate to the most common clouds. While migrations can be done manually, Amazon offers a vCenter plug-in that orchestrates the migration of VMs and their associated storage to Amazon AWS.

This plug-in lets you recreate EBS volumes on the target instead of the virtual disks of the original machines. However, migration needs to be accurately planned (especially for network management and addressing plans) and requires some time. It also requires a scheduled shutdown of the application.

To minimize migration times, it is possible to create the target instances with high-performance EBS SSD volumes, before gradually switching these volumes to more appropriate service classes.

The migration to Google Cloud can also be done manually or in an automated way, via the services of CloudEndure, an Israeli Google Cloud partner who has developed a solution for migration of water environments on premise to the cloud infrastructure of the American giant.

Like AWS, Google Cloud recommends first to make sure you set the network configuration before migrating VMs and their associated disks. For NAS service migration, it is not possible (just like at AWS) to organize data copying and synchronization using NFS – Google Cloud connections that do not offer a NAS service.

But as Bastien Legras explains, in charge of the vendor’s pre-sales teams in France, it is possible to combine the Google Cloud Storage object storage offer with Avere’s virtualized services (vFXT platform) for Deliver powerful NAS storage services and combine them with the same on-premises tools to seamlessly migrate NAS services to its cloud.

Avere’s technology can be deployed in the same way on most market clouds for a similar purpose.

At Microsoft, several migration scenarios are possible, but the simplest is of course the migration of a Hyper-V environment to Azure.

But Microsoft also offers the Azure Migrate service to help plan and orchestrate the migration of virtualized environments with VMware to its public cloud. The tool allows you to inventory resources to migrate, analyze dependencies, and then optimally match Azure storage resources to on-premises storage resources. Another tool – Site Recovery – then automates a large part of the migration process.

A last alternative in the case of approach type “lift and shift” is to not only migrate applications, but to make a “lift and shift” of the storage infrastructure.

This is an approach that NetApp allows, for example, by providing a software appliance version of its storage arrays that can be deployed over the block services of most cloud services. With the Ontap Cloud appliance, it is possible to have the equivalent of a real NetApp array with all its services (snapshots, synchronous and asynchronous replication, high availability) but in the cloud.

The generalization of Software Defined Storage approaches also allows companies using these technologies to extend them to the cloud. Distributed file systems like Dell EMC Scale IO or Elastifile, to name a few, can be deployed in hybrid mode and serve as a storage base for a distributed infrastructure between private data centers and public cloud infrastructures.

Beyond Lift and Shift

In some cases, it may be interesting in the migration phase to go beyond the simple “Lift and Shift”.

For databases, for example, it may be interesting to switch from a traditional database model, or you need to provision storage, VMs and database instances (with the impact that this can have. software licensing) to a Database as a Service model.

This model is for example encouraged by Oracle for its DBMS. It is also at AWS and Microsoft. For its part, Google offers its CloudSQL service for the open source database environment switch such as MySQL or PostGreSQL. It also puts forward its Google Spanner home services (massively distributed SQL database) and Google BigTable (distributed NoSQL database) as alternatives to equivalent on-site databases.

The advantage of such an approach is that the company no longer has to manage the underlying underlying storage infrastructure and is assured that performance will evolve according to its needs (within limits, of course, the capabilities of the managed database service).

Migration to cloud-based managed services can also be of interest to companies operating Hadoop clusters. Most public cloud providers offer a managed Hadoop / Spark service.

Finally, there is today a large ecosystem of tools that facilitates the migration of applications and their data to the cloud. For example, providers of replication and disaster recovery tools such as Double Take, Zerto or Cloud Endure, or data backup and management actors such as Veeam or Rubrik can be mentioned.

These solutions can greatly simplify the gradual migration of data and, ultimately, the switch applications to the cloud.

Leave a Reply

Your email address will not be published. Required fields are marked *