What’s Wrong with a Cloud First Strategy – Velostrata Briefing Note

The problem with a cloud first strategy is organizations assume clouds are the ultimate data centers, that the organization can’t do it better themselves. The reality is that there are some functions where the cloud is more optimal. Then there are functions where either an on-premises data center or a different cloud provider is a better alternative. And the variables impacting where an application or workload should be hosted change as the organization and cloud providers evolve. Mobility is essential to make sure the organization’s workloads are running in the right data center at the right time.

The challenge is that storage and compute are tightly coupled, making movement based on the changes in business variables very difficult. Most organizations are glad to get their first cloud migration done, they don’t want to have to do it again anytime soon.

The Cloud As a Resource

The cloud is a resource at IT’s disposal to enable it to better meet service levels. The tight coupling of compute and storage makes it difficult to use only the parts of the cloud that the organization wants to. This tight coupling is not just a legacy problem but also a modern day cloud problem. Most providers expect the organization to put both its applications and associated data in the cloud and to never leave.

The tight coupling of compute and storage limits application mobility because when an application is moved from on-premises to the cloud or from one cloud to another data has to precede its movement. If the related data is of any size at all then that means a significant amount of delay when starting the application.

The reason for the tight coupling of compute and storage is to reduce or eliminate latency. This concern is especially valid in the cloud. Imagine how slow, if at all, a cloud migrated application would run if it had to wait for each I/O request to be serviced across an internet connection.

What if part of the decoupling also included caching and only the active data and then updated data followed the application? Assume that the cache could then be anywhere, on-premises or in one of multiple cloud destinations. Then the application could rapidly move between these locations almost at will.

A de-coupled compute and storage tier combined with intelligent caching would need one more ingredient, transformation. Each cloud is different, it has a different architecture and uses a different hypervisor. For the application to run in the alternative cloud requires that adaptations be made and drivers injected.

Velostrata Decouples to Enable Mobility

Velostrata’s agentless software first moves critical boot data to a cloud-based cache, enabling an application to begin running in the cloud in minutes, and then streams additional data in the background as needed. This process is similar to NetFlix – you can begin watching a movie before all the data is transferred. Additionally, Velostrata auto-adapts each workload on the fly so it can run natively on the cloud target.

With Velostrata the bulk of the data can stay on-premises, it becomes easy for an organization to quickly test how applications will perform in the cloud. Velostrata’s WAN optimizations and caching ensure normal application performance, despite the WAN. If the application works as expected in the cloud, data can then be migrated in the background while the application runs normally.

Unlike other migration solutions, “running in the cloud” does not mean all the application’s data needs to move to the cloud. It can remain on-premises and only active data is cached to the cloud. Alternatively data can be 100% in the cloud but have changes synchronized back to the on-premises data center so that if there is a cloud outage the data center can be the DR site.

Velostrata Architecture

Velostrata integrates into the organization’s on-premises VMware infrastructure as a virtual appliance. There is no need for agents or drivers within the virtual machine. Velostrata also supports mixed environments with virtual and physical machines. For the cloud, Velostrata’s on-premises management virtual appliance will automate the deployment of the Velostrata CloudExtension in the cloud provider of choice. Today Velostrata supports Amazon Web Services and Microsoft Azure.The CloudExtension virtual appliance is dual node and active-active for performance and scale. The write journal is mirrored across the two nodes to enable a seamless failover in case of a single node failure.

After the software’s installation, the Velostrata vCenter plug-in manages the migration process. Velostrata also provides a RESTful API to support integration with other management tools. Most applications are not made up of one virtual machine, but dozens all with their own inter-dependencies. Velostrata also provides an Automation Runbook tool that can insure the VMs are logically grouped, all the components are available to the application, and that they start in the right order when migrated.

To ensure performance during streaming and even after migration, the CloudExtension virtual appliance node is multi-tiered consisting of RAM, SSD and Object Storage. Initially data is streamed into RAM and then destages over time to SSD and Object Storage (assuming that a cloud data presence is acceptable). All writes are cached locally and persisted across two nodes in different availability zones so as not to impact write performance, and then persisted in the durable object store. Per user choice, writes can also be compacted and efficiently transferred to the primary data center asynchronously.

Migration is not limited to VMware, the solution supports physical, non-virtualized, servers. While most data centers are over 70 percent virtualized, typically the portion of the environment that is not virtualized are mission critical systems. The ability to include these systems in the organization’s cloud strategy is vital to creating a comprehensive plan.

Cloud to Cloud to On-Premises

As organizations include the cloud in their strategies a new request is developing. They now want to be able to move their workloads between regions in a cloud and between cloud providers. The primary motivation is maintaining continuity if a cloud provider has an outage. The same caching and streaming technology that Velostrata uses to move workloads between on-premises and a primary cloud, also works between cloud regions and even cloud providers. Data can be cached and streamed, allowing an almost instant movement of an application.

StorageSwiss Take

Velostrata enables an organization to use the cloud as a real extension of the data center. The organization can start its cloud journey by leveraging the cloud for disaster, creating its own disaster recovery as a service solution. Then it can test applications to see if they would run well or even better in the cloud. Those applications can then be migrated, but the organization can decide where the data associated with those applications should reside. They may decide to leverage the compute of the cloud but keep data secure on-premises.

Velostrata allows an organization to create a “cloud when it makes sense” strategy instead of a “cloud first” strategy, which is really the better way to leverage cloud resources.

Eight years ago George Crump, founded Storage Switzerland with one simple goal. To educate IT professionals about all aspects of data center storage. He is the primary contributor to Storage Switzerland and is and a heavily sought after public speaker. With 25 years of experience designing storage solutions for data centers across the US, he has seen the birth of such technologies as RAID, NAS and SAN, Virtualization, Cloud and Enterprise Flash. Prior to founding Storage Switzerland he was CTO at one the nation's largest storage integrators where he was in charge of technology testing, integration and product selection.

Tagged with: , , , , , , , , ,
Posted in Briefing Note

Enter your email address to follow this blog and receive notifications of new posts by email.

Join 21,781 other followers

Blog Stats
%d bloggers like this: