Cloud-washing, a practice of renaming or repackaging existing products and dubbing them cloud-ready, has become an art form. Every software package, service and now even hardware is described as cloud-enabled. Seemingly, we are all headed into the clouds never to return. Sounds like heaven, right? Maybe, maybe not.
While the atmosphere is everywhere, the cloud decidedly is not. It is by definition remote. That means if your applications require the highest in performance, cloud may not be the answer. For example, when you rent movies that are delivered over the cloud in real-time to your home, periodically, everything freezes and you see the infamous “buffering” message. While this may enhance the mystery of what is going on in the “Lost” episode you’re watching, it would be a disaster for algorithmic trading. In that environment milliseconds can determine profitability or loss.
Not all applications can be completely move to the cloud. In some cases this is done by design and the type of application, while in others the impact of migration is too high. For those business-critical applications that require the highest performance and have the most sensitive information and complexity of process, a hybrid model of cloud services to bridge the datacenter, branch offices and trading partners is best.
In this slideshow, Charley Rich, vice president of product management and marketing at Nastel Technologies, covers the top five best practices for ensuring optimal performance and uptime when using the cloud for your most demanding applications.
Click through for five tips on ensuring optimal cloud performance for your most demanding applications, as identified by Charley Rich, vice president of product management and marketing at Nastel Technologies.
Determine what applications can be safely moved to the cloud and which ones cannot. Consider issues of sensitivity where it may deemed too risky to store information in the cloud, such as credit card data. This decision will obviously differ by organization — an auction-type payment provider will certainly utilize the cloud, while a large B2B bank will most likely not risk the potential exposure of its customer account data. Also, consider issues associated with latency constraints which would deeply impact applications such as algorithmic trading. Many of these applications are “co-located” in the stock exchange in order to circumvent the lag of any external network. However, there are many applications where latency in the low seconds is acceptable, and these may be good candidates for cloud residency.
The cloud is elastic. Are your applications? One of the big drivers for using the cloud is that you only pay for the resources you utilize. When it determines that you need more resources, the cloud automatically provisions them for you. But not all applications can immediately make use of additional CPUs and memory. There may be serialization in the applications, problems with threading and even memory leaks. Don’t assume that any application will transparently expand to handle rapid peaks in demand merely by virtue of being in the cloud. The same basic IT architecture principles for handling bottlenecks and designing in performance still apply.
Remember the transaction. Monitoring is often forgotten when talking cloud; however, it is essential. Rather than merely monitoring the individual components such as the network, servers, J2EE Web apps and message operations, take a step back and monitor the entire transaction path.
If you just monitor the components it is like reading the cast of characters at the beginning of a play, but not reading the actual dialog. You will know who is in the play, but not what they said or did. Transactions are conversations, if you don’t monitor them, just like in the play example, you will have no clear idea of what is going on. Each component could be quite happy, but an order may still not be processed. Instead, focus on transaction performance: the speed with which an application completes an activity that meets business requirements and expectations.
Start early and save. When your enterprise applications begin communicating with other applications in the cloud you are embarking on a new integration scenario and performance testing end-to-end is critical. Enterprise applications often communicate with other applications using middleware, so it is important to begin testing enterprise apps immediately after unit testing is completed. This is especially important with transactions that will span the datacenter and the cloud. Besides performance bottlenecks, there may be logic errors uncovered. This is also a great time to put in place policies that will prevent known problems from reoccurring, and these policies should be consistent across both the datacenter and the cloud. Think of the lifecycle of an application working with the cloud as a continuum. The farther to the left you begin testing the less expensive it will be to resolve problems in availability, logic and performance.
Self-service in the cloud drives business performance. In a hybrid cloud environment, there can be many stakeholders who need access to application data – from the enterprise architects, to development/user acceptance testing to IT operations and the shared services team. Therefore, it’s important to make access to data as easy as going to an ATM. Self-service access enables named users to immediately get to the data they need without installing any software and without disrupting the productivity of others to get it.
