Data migration has long been the bane of IT, particularly as the cloud introduced entirely new classes of infrastructure and helped push data volumes to new heights. But now that on-premises infrastructure has become more cloud-like and automation/orchestration tools are pushing fully software defined data center (SDDC) architectures across distributed IT environments, migration is not nearly the chore it once was. In fact, much of the difficulty surround cloud migration these days lies in the planning and preparation stages, primarily in choosing what to migrate and exactly how you want the resulting cloud environment to function. So in that vein, here are some of the leading tips for achieving a successful migration:
The Many Faces of Migration
First off, it helps to understand what type of migration you are planning to undertake, and for that you need to determine what kind of cloud you are planning to implement. Will it be an all-cloud environment or a hybrid? Will it support customer-facing, mission-critical apps, back-office functions like business intelligence or data analysis, or long-term operations like archiving and disaster recovery? Will it require a one-time migration, infrequent data transfers or steady rapid bursting? Each of these factors will pose different challenges in terms of data access, continuity and transfer rates during and after the migration process.
If this is your first migration, it’s probably wise to start with a small load of non-mission-critical data just to get the IT team familiar with migration tools and procedures, and even gain a better understanding of what a successful migration actually looks like. Quite often, it isn’t easy to tell if everything has gone smoothly until the new environment is brought online, particularly for technicians with little or no cloud experience. (More on that in a moment.)
Make a Plan
Once the initial objectives have been defined, it’s time to construct a working plan for migration.
A good place to start is to conduct a thorough assessment of legacy infrastructure and applications to determine exactly what you want to migrate and why. The cloud is not the optimal environment for every workload, so IT executives should take the time to determine what kind of performance they expect to achieve, what security precautions to take, how governance and policy management will be enforced and a host of other factors.
As well, the plan should incorporate any and all regulatory factors that govern data, ranging from residency and availability to disaster recovery, data integration and risk management. Not all clouds are the same, and in fact many providers are starting to tailor their offerings to the regulatory requirements of key industry verticals like healthcare and finance. General purpose clouds can certainly be architected for regulatory compliance, but this requires a fair amount of in-house expertise that may end up costing more than a pre-configured environment.
And it goes without saying that the cost-benefit of any migration should figure high in the planning phase. Not only should the cloud provide a lower cost basis on which to support data and applications, but the migration itself should be designed to provide the quickest, easiest and least disruptive manner of moving workloads from the enterprise data center to a third-party facility.
A Seamless Migration
It is unlikely that a cloud migration will not impose some downtime for affected applications, so you should think about how much is acceptable to your end users. As a general rule of thumb, the less critical the app the greater tolerance for downtime, but this can vary widely from application to application and from enterprise to enterprise. According to Tom Andrulis, president of U.S. tech management firm Intelligent Technical Solutions, downtime can stretch from a few minutes to several days, so organizations should have a clear idea of what is at stake before the migration begins.
Once underway, the migration will require continual monitoring to ensure not only that data is moving as expected but that data relationships, network pathways, replication settings and multiple other elements remain in sync. This is particularly crucial if the ultimate goal is a hybrid cloud in which data and applications will be expected to traverse third-party and on-premises infrastructure on a continual basis.
Unfortunately, there are no universal guidelines to avoid the headaches that arise during the typical migration, as every enterprise has unique requirements for data replication, accessibility and myriad other factors. As tech consultant Kurt Marko notes, simply choosing what data to replicate, how often and for how long can be a lengthy process, and it can often depend on multiple other decisions, such as whether all application layers are being migrated or just the user interface and business logic.
Test, Test and Test Again
The seldom-spoken truth is that the migration doesn’t end once the last bit has made it to the cloud. Indeed, in many ways the testing phase to determine the success, or failure, of the operation can be just as complicated as the actual data movement.
As mentioned above, the biggest challenge here is the fact that in-house technical staff rarely has the right skillsets to ascertain what is working in the cloud and what is not, or to identify and correct problems in the latter case. Multiple tools are available for this phase, most of which are available from the cloud provider. If not, it’s probably best to look for another provider before you start.
In general, however, the same basic principal applies to the cloud as for newly provisioned local infrastructure: don’t go live until you are sure the application is providing acceptable performance. Again, this will vary for the application, with some requiring low latency and others needing broad access to multiple storage volumes and compute resources.
Another item on the post-migration checklist is the establishment of broad visibility into the new environment. Few cloud providers will allow clients to peer directly into their infrastructure, but there are various third-party tools on the market that can peer into data movement, resource utilization and cost allocation to at least allow you to determine for yourself that you are getting what you paid for.
Keep it Secure
Another critical element at this point is the establishment of an adequate security regime. As Radware’s Ron Winward points out, just because you’re on the cloud does not mean you are limited to your provider’s security. Many providers will allow third-party security systems access to DNS changes and BGP redirects so clients can perform their own attack detection and mitigation functions.
You should also have a plan to separate normal traffic from compromised traffic, preferably keeping data decryption to a minimum. This can be done using behavioral threat algorithms and other advanced tools that not only enhance user privacy but speed up the containment and damage mitigation processes as well.
Find a Good Partner
Since few organizations have the in-house expertise in cloud migration, selection of a strong partner is essential. This can be the cloud provider itself, the developer of cloud management software or even a third-party consultant.
No matter who your guide is, you should insist upon a clear and concise service level agreement (SLA) for the life of the cloud deployment to cover things like data availability, uptime, performance levels and the steps to remediate disputes. As well, all partners in the process should provide a straightforward billing statement that delineates cost schedules, additional fees, services and anything else that may arise.
And it goes without saying that a good provider will be attentive to your needs not their own, particularly if you are paying for enterprise-class service rather than simple bulk storage. Many small and medium providers, in fact, are eager to differentiate themselves from the hyperscale players like Google and Amazon by providing highly customized services to enterprise clients.
More than any other cloud-related function, data migration adheres most closely to the old garbage in/garbage out (GIGO) paradigm: the less care you take upfront, the poorer your results on the back end.
Enterprises that want to implement a productive cloud environment smoothly and efficiently would do well to take the migration process seriously, if only to prevent it from become more of a chore than it has to be.