Enterprise applications typically represent a significant on-premises investment, with years of integration, upgrades and honed processes. Even with a “Lift and Shift” methodology, organizations have struggled to untangle the inherent interdependencies of these long-running, IT-centric business applications in order to move them – and their data – to the public cloud. Consequently, when considering public cloud adoption, one of the CIO’s highest priorities must be to minimize migration risk.
In this article, we’ll consider two migration alternatives: the traditional replication approach vs an innovative streaming-based approach. We will examine how each approach addresses the migration challenges associated with data-intensive workloads and supports your enterprise’s vision for cost-efficiency and business agility in the cloud.
Considering the Two Cloud Migration Alternatives
When moving enterprise applications to the public cloud, CIOs typically have two main drivers: a) cost-efficiency – leveraging the public cloud’s pay-as-you go model, in lieu of building out on-premises infrastructure; b) business agility – capitalizing on the cloud’s endless capacity on-demand.
Enterprises have two primary options for migrating workloads to public cloud Infrastructure as a Service (Iaas):
- Replication-based Migration – the approach used in the majority of migration solutions, where workloads are moved “as-is” using a traditional migration tool. An entire replica of the existing on-premises data is moved to the cloud, and then active workload and users are switched to the replica (“cutover”).
- Streaming-based Migration – a new approach, where workloads instantly start in the cloud, and data is moved by “streaming”. Only the data necessary to run the application is migrated first, allowing for immediate productivity. Optimizations ensure full application performance over the WAN. The remaining data is streamed in the background.
Both replication and streaming-based approaches require initial discovery and assessment to determine dependencies and cloud fit using manual and tool-based approaches, such as the AWS Discovery tool or Microsoft Assessment and Planning (MAP) Toolkit for Windows Azure Platform. Both also require analysis and planning to create the “runbook” containing the workloads, sequencing and scheduling plan.
Once the discovery, assessment and planning is complete, the migration process begins. Here are 5 considerations to evaluate when choosing your migration technology.
How long does it take to get your application running in the cloud? How long does it take to complete the total migration process and cutover? How much time will be added for testing before and after the migration?
The traditional replication based approach presents many challenges when migrating large datasets to the cloud. For instance, moving 1 TB over a 50 Mbps link takes at least 2 days (47 hours). Even with a faster link like 100 Mbps, it takes 1 day. And what if 2, 4, 10 or 100 TB applications need to be moved? The impact on time is particularly acute with apps that frequently create new data. By the time the initial set of data is replicated to the cloud, it is no longer up to date, and there is more data on-premises that must be synced, which can substantially increase the overall replication time.
With streaming-based approaches, the actual migration (cutover) happens first. Caching, WAN optimization, deduplication and other technologies allow the application to run with performance that is comparable to on-premises immediately—well before all of the data is present in the cloud. Workloads can be live in the cloud in minutes, while the rest of the data is migrated in the background. New changes are made to the live cloud instance and kept in the cloud, so there’s no need for complex syncing or risk of data loss during cutover.
In a recent survey conducted by Morgan Stanley, over 90% of CIOs reported that migration fell short of expectations, mostly due to “complexity.”
Migration for enterprise applications has many moving parts. When considering migration tools, try to avoid any additional complexity they may add to the process, such as:
- Replication agents that need be installed, consume resources, and potentially introduce a security vulnerability
- Does the system require learning a new UI or management interface or does it integrate with your existing systems and training investments?
- Does the system migrate only individual VMs or can you script and automate multiple VM migrations and configurations?
Mistakes also happen in the cloud. Despite the best planning, sometimes things don’t work out as you thought they would. How can you ensure that an application you have on-premises is going to function well in the cloud? What do you do if you don’t get the expected results? Or you want to switch cloud service providers?
An important tip when considering migration tools is to first validate if you can easily ‘test before you migrate,’ avoiding unexpected surprises that can cost you time and money. Another thing to check is what you need to do if you want to revert back after the migration “cutover.”
With replication-based migration, you’ll need to follow the same lengthy, complex migration process to move the entire data for a simple test in cloud, and the same in order to move back to on-premises, or to another cloud provider. For most organizations, this is so distasteful that they are essentially locked-into their cloud provider.
With streaming-based migration, full mobility is inherent in the architecture. You can take a snapshot of a workload state and test it in minutes in the cloud with all its data – quickly uncovering any issues, performance bottlenecks and cost implications. If you’ve migrated an application but decide another cloud service provider has better terms, you can easily move it to the new provider in minutes. This minimizes your migration risk.
How can you ensure cloud migration will bring the expected performance (and cost)? What happens when your cloud provider changes services or pricing?
Just replicating an application “as-is” to the cloud doesn’t mean you’ll get better cost/performance, but it may take some time before you have a proven cost model. Replication-based approaches require the same effort to move back out of the cloud or two a different cloud, so you are essentially stuck with your provider. Using streaming-based migration keeps the IT environment flexible – making it possible to move workloads in, out and between clouds gradually – all in real-time.
Is your data safe, protected and compliant? According to a LinkedIn Information Security Community survey, 49% of CIOs and CSOs feel that “one of the major barriers to cloud adoption is the fear of data loss and leakage… 59% believe that traditional network security tools/appliances worked only somewhat or not at all” in the cloud.
Especially for enterprise application migration projects, concern rises due to data sensitivity. So, examine migration tools for their security requirements and capabilities, such as encryption (in motion and at rest in the cloud). For some organizations and use cases, streaming-based migration may be favorable for sensitive data in the cloud, as it is possible to retain authoritative data on-premises and only move a small, encrypted subset is cached, reducing cloud footprint.
A Comparison Table
Here’s a comparison table summarizing the 2 approaches based on the 5 main challenges of IT teams in migrating enterprise applications to the cloud:
|Cloud Workflow Migration Challenge||Replication-based Approach||Streaming-based Approach|
|1. Time to cloud||Days or weeks||Minutes|
|2. Complexity||Complex process with agents and new management systems||Simple, agentless process using existing tools and procedures|
|3. Risk||Lock-in (effectively)||Mobility|
|4. Cost/Performance||Rigid operational environment with potential for waste||Flexible workload control for optimization by migrating in/out of cloud|
|5. Security||All data in cloud with manual/complex encryption||Choice in data location (keep on-premises or move to cloud) with built-in encryption|
Many factors present barriers to cloud adoption in enterprises. In this article, we’ve focused on one of the major challenges for CIOs today who want to capitalize on the public cloud for cost-savings and business agility – moving data-intensive, enterprise workloads into the cloud.
InfoWorld asks: Can CIOs “achieve workload mobility through intelligent streaming, not heavy lifting?” While replication has been doing the job of moving workloads to the cloud, the majority of enterprises are still hesitant to fully embrace it. Complexity, implementation time and performance issues remain major stumbling blocks. Can the streaming based migration architecture possibly offer a new paradigm for bridging this gap by adding mobility and flexibility?