For decades, workload mobility has been considered an important feature of every enterprise data center. It’s required any time IT needs to move an application and its data between different and distant environments – including cases such as switching colocation centers or establishing a new data center to support global expansion.
But, even though workload mobility is not a new concept, the rapid adoption of public cloud has brought this idea front and center within many enterprises. With the multitude of cloud service providers (CSPs) ranging from online hyper-cloud providers such as Amazon, Microsoft and Google to dozens of public cloud vendors, today more than ever, enterprise IT teams need workload mobility to unlock their on-premises environments and extend their data centers to the cloud. At the same time, they don’t want to end up in cloud jail, tied indefinitely to one vendor with no easy alternative to move back on premises or to another CSP.
“Everyone wishes they could go back in time and tell themselves to use the cloud to grow, but not get tied to any one provider.”
– Avi Freedman, CEO, Kentic
The Challenges of Moving Workloads
In general, moving workloads is a challenging process for IT teams for a few reasons:
- The application must be taken down in one location, adapted, and then completely restored in the new target location
- Application SLAs must be met in the cloud environment, including performance (e.g., acceptable response time, IOPS etc.), high availability, fault tolerance and more.
- Data needs to be migrated which can include databases that must remain consistent and complete
- Network configuration must be updated, including IP addresses, subnets/vlans and general network security and isolation policies
- Management of the environment, including aspects such as security, backup, and monitoring, need to reconfigured
- Staff need to be trained on how to use, manage, and support the new environment using new management tools and operational models
For these reasons (and others), workload migration has been slow, complicated, and risky, which is why IT mostly just tried to avoid it. However, as the agility and cost efficiency benefits afforded by public cloud become undeniable, IT organizations find that they really do need to find a workload mobility solution that can cross these barriers.
As migration and mobility platforms become easier, faster, and risk-free, IT can begin to capitalize on some or all of the following cloud workload mobility use cases.
- Running stateful or stateless Dev/Test
With the prevalence of comprehensive, modern web applications such as social sites and mobile device apps, users have come to expect rapidly evolving applications with fast, flawless performance. These same users expect a similar experience at work. Therefore, IT organizations are even more pressured to be innovative, deliver new applications and services at an accelerated rate, while maintaining high quality.
Depending on what IT needs, there are actually two unique dev/test scenarios which might occur. The first is sandbox testing, where IT has to clone a workload into the cloud and run tests but the cloud instances and data will not remain persistent afterwards. In fact, the production workload (and data) remains on-prem while this cloud clone runs. Essentially, it’s disposable testing so that IT can get a feel for how certain workloads perform under certain conditions and then fine tune accordingly.
The second use case, however, which has posed more challenging for IT is stateful dev/test. In this scenario, IT migrates an actual workload into the cloud with production data that must persist throughout the testing. This scenario is often a dry-run where actual users are using their production apps (and data) so IT can see how the new environment stacks up. If results are poor, IT will migrate the workload (and data) back on-premises, and if results are strong then IT will keep the workload in the cloud. Either way, it is crucial for production data to remain 100% persistent through any migrations and/or usage.
Moving development and testing (dev/test) to the cloud helps significantly speed up the delivery process. IT can create an “as-is” production environment including both application and data for dev/test, parallelize testing, and easily decommission resources when they are no longer needed – saving money. In the past, it’s taken too long to migrate data into the cloud for quick dev/test use cases, but modern workload mobility solutions incorporate streaming-based cloud dev/test, which creates a thin clone replica in the cloud, while key data is accessed on-demand or streamed in the background, allowing the workload to start in the cloud within minutes (instead of hours or days).
- Ensuring Multi-Cloud Flexibility
Workload mobility is important for preventing cloud vendor lock-in, and it gives IT the freedom of choice to build out a multi-cloud environment which best meets their needs. With cloud provider choice, you can benefit from the strengths of each offering. For example, using the strengths of both AWS and Azure in parallel, by being able to switch between them for feature and cost optimization across your hybrid cloud footprint.
However, as in the dev/test use case, moving stored data remained the challenge. A phenomenon referred to as ‘data gravity’ by blogger Dave McCrory: “as Data accumulates (builds mass) there is a greater likelihood that additional services and applications will be attracted to this data… data, if large enough, can be virtually impossible to move.”
Essentially, the more data you have in a specific location, the harder it is to move due to the number of applications that depend on the data. To properly mitigate the challenges of cloud lock-in and data gravity, it is often ideal to move the application compute workloads separately from the data, another scenario where streaming the required workload data in the background for workload mobility is most successful.
Another important challenge in moving workloads across cloud providers is with the conversion of the images to fit the format of the required target cloud, as well as the type of hypervisor, device drivers, network configurations, security and all aspects related to moving into a cloud environment.
- Supporting a High Performing Global Enterprise Network
Another main challenge of global enterprises is connecting global offices with a high performing network. The high performing global network is important to facilitate for things like information sharing and collaboration. However, one of the difficulties is getting the data closer to its consumer at varying points of time. This often requires a complex and costly mechanism for dynamically syncing data between regions.
To combat this challenge, advanced workload mobility solutions such as Velostrata empower global IT teams to keep all of their storage in one central, secure site, while only moving an application’s compute workloads between sites. Data caching and WAN optimization technologies help deliver high performance which alleviates end user latency issues, and prevents having to move huge data sets, too. Decoupling the data residency and compute considerations from the specific IT environment’s configuration gives IT the flexibility to choose between the portability of specific workloads versus the entire application stack.
- Migrating Bidirectionally between clouds and on-prem
With the public cloud market growth, CIOs today are always looking for ways to migrate their entire application stack to clouds like Amazon or Azure. However, veteran enterprise IT leaders want to mitigate the risks involved with leading large technology projects like data center deployments and/or migrations. Therefore, gradual movement of applications accompanied by pilot projects and tests can help IT evaluate the risks of moving to the public cloud. With each iteration, optimizations can be made for things like performance and capacity.
To empower quick, iterative testing, fast workload movement is needed in both directions: to the cloud, and back on-prem (if needed). In addition, the migration solution must be able to maintain data on-premises or in the cloud, and then synchronize back to the other as needed. This state-persistent, data-persistent, bidirectional migration support eliminates the risks involved in running a full-blown migration because if a test fails, its quick and easy to get that application back on-premises without any negative impact (to users, performance, or data). With bi-directional workload mobility, IT teams can understand the limitations with their workloads before a mass migration and architect accordingly. Moreover, they’ll always have a safety net, via quick migration back on-prem if needed.
- Bursting to Relieve a Bottleneck
Joe Weinman, a prominent cloud computing researcher, in his book “Cloudonomics: The Business Value of Cloud Computing” discusses the economics of the hybrid cloud. His basic theory is that hybrid clouds are more beneficial, especially when the unit cost of a public cloud’s resources are higher than the dedicated on-premises resources. This basic benefit is demonstrated very well in an environment with a variable demand. So, while the average steady load is kept on-premises, the public cloud can be used for bursts.
This is a perfect use case for workload mobility because it enables IT to satisfy their end user’s demand with optimal performance, but at optimized cost levels. The trick here, however, is to have a solution that can get additional capacity into the cloud within minutes, while leveraging the production data set on-premises. Ideally, a solution will have built-in support for the most affordable instances, like Amazon’s spot instances, to further reduce your capacity bursting costs.
- Leveraging special instance types on-demand
One of the more unique benefits of the public cloud is being able to use systems you wouldn’t ordinarily be able to take advantage of. A few of these systems include Amazon’s X1 with 128 CPUs and 2 Terabytes of RAM, or the Azure N Series with hefty GPU power. These systems are often too expensive, too big, or require too much special skill to manage for organizations to deploy on-premises, so it’s a big advantage when an organization is equipped to rent them on-demand.
To successfully leverage these unique computing resources in the cloud, IT teams need an easy way to migrate their workloads back and forth. Once again, with the right workload mobility platform, you can quickly get your workloads computing on these unique instances while your data remains on-premises. This empowers IT to maximize the time and cost of using these unique instances for things like high-performance batch processing or monte carlo simulations.
The adoption of public cloud is already happening within a rapidly growing market. At the same time, however, decades of investments in on-premises environments will not (and cannot) simply vanish. This is why hybrid clouds and workload mobility are such a key component of any public cloud discussion.
Ultimately, decoupling things like compute power from data helps bring give IT organizations the freedom of choice and flexibility with regards to their implementation. It empowers IT to leverage the best combinations of the best options within the market. As these combinations continue to prove most powerful, workload mobility will continue to be a necessity for maintaining a modern and a dynamic enterprise IT environment.