What enterprise workloads are right for the cloud? [Q&A]
The question many businesses are asking themselves is, 'should we be moving to the cloud?' The public cloud is clearly a success, as shown by the significant adoption of Amazon Web Services (AWS). Companies are benefiting from the pay-as-go nature of the cloud, as well as from the ability to turn up services as needed without the traditional hardware spend. Gartner recently reported that cloud computing will affect more than $1 trillion in IT spending by 2020.
But even with this growing popularity, corporate IT departments are still struggling with how to integrate public clouds into their data center initiatives. One of the main reasons for this hesitation is the uncertainty around maintaining workload performance once the data gets to the cloud. This is a valid concern, but one that can be overcome. We spoke to Len Rosenthal, chief marketing officer for infrastructure performance specialist Virtual Instruments, to discuss how workload analysis and modeling is the first step for any cloud migration initiative.
BN: What's holding up cloud adoption at large enterprises?
LR: IT teams have been considering the cloud and its benefits for over a decade. SaaS solutions like CRM, cloud storage and project management, to name a few, are already entrenched departmentally. Moving workloads to the cloud is another activity altogether. While security and control concerns have lessened as cloud adoption continues, IT departments are worried about another critical factor -- guaranteeing performance off premise. And compounding the issue is that IT departments cannot afford to spend long periods of time planning, testing and experimenting. Cloud innovations and developments are happening so rapidly that in a year’s time you will witness a vast change in options, vendors and approaches. You have to work intelligently and with purpose when initiating a cloud project.
Decision makers need to consider a range of options and be realistic about what it will take to actually migrate each potential workload. What's the process? How will I manage them? What if I need to migrate my workloads back into my data center? These questions and more must be addressed. This may sound like common sense, but you’d be surprised how many companies jump first and ask questions later.
BN: How can understanding a workload's performance profile help?
LR: You need to gain an understanding of exactly how each workload will perform under different conditions and access variations. This can't be generic test data. It has to be your organization's real workload profile data, put through the paces to create an accurate reality of how your workload deployed in the cloud will perform. Only then, can you evaluate its feasibility in a cloud environment.
One technology that helps mitigate some of the risk is performance profiling, which you’ll also hear called things like multi-dimensional benchmarking or performance corners testing. No matter what it's called, performance profiling provides a sort of 'performance map' that lays out how each tested workload will perform under literally hundreds, or even thousands, of different scenarios. With this data in hand, companies can more accurately define their workload deployment strategy.
BN: How can companies maintain performance once data moves to the cloud?
LR: Just getting workloads to the cloud isn't enough. That's only the start. Maintaining workload performance is a concern we hear daily. Once the workloads are up, they need to perform just as well as they did beforehand. And to focus on this you need to be able to act and react in real-time with the help of monitoring tools.
This means measuring critical metrics, like I/O performance, and using those measurements to assure that each workload is performing up to where it should be. A performance monitoring and assurance solution like this enables IT to troubleshoot faster by discovering potential problems and solving them before they affect production performance. Make sure your cloud-hosting provider utilizes a real-time monitoring solution that covers their entire IT infrastructure before you deploy in the cloud.
BN: What should people keep in mind when considering performance profiling and assurance options?
LR: As with any important piece of technology, you really have to do your research before deciding. There is a wide array of key features and concepts to consider when looking at evaluating solutions, but these are the most important considerations:
The first is to make sure the solution is vendor neutral. You have to be able to accurately, almost instantaneously, receive and analyze this data no matter where or which device it's coming from.
Next, the solution has to use your data so you can weigh it against the cloud vendor’s SLAs and guarantees, as well as the cost needed to support the cloud instances. Cloud providers will give you performance statistics, and those are helpful, but only marginally because it’s their data and it's not a direct representation of what you have. It's a best guess scenario and there's too much at stake to just leave it all up to chance. You have to be able to conduct performance profiling on your own data, using simulations that reflect your actual environment.
Finally, once you're in the cloud, the importance of being able to analyze real-time data can't be overstated. And you need real-time access to this data. Business moves so fast these days that even data that's a few minutes old might not be valid.
BN: In general, which types of profiles perform best in the cloud?
LR: This will, naturally, vary from business to business. Your performance profiling data will provide specific information about which workloads are best for migration for your particular situation. That being said, there are a couple common use cases that make particularly good business sense for the cloud:
- Disaster recovery: The single most common use case for cloud storage right now is disaster recovery. When you think about it, it makes perfect sense: replicating your critical data, workloads and applications to the cloud provides another layer of security if anything should happen to your physical infrastructure. With disaster recovery, performance issues like latency aren't really a problem, and performance limitations are usually limited to bandwidth.
- Secondary data: The cloud is all about maintaining reliable infrastructure while optimizing costs. With this in mind, it’s easy to see how storing secondary or Tier 2 data in the cloud can be beneficial. It allows your company to save money by partially cutting on-premise storage needs. This reduces the constant physical management required, of course, while also eliminating the need to replace hardware when it wears out. Many non-mission-critical file-based workloads will likely be good candidates.
Again, every company's mileage will vary. The keys to determining what will work best for you is to understand each of your individual workloads before deciding what to move to the cloud, and then monitoring those workloads closely and continuously once they're there.