Most Global 2000 IT Infrastructure and Operations Managers feel like the wheels have come off their data center modernization project. The goal was to create an environment that can respond to the needs of thousands of applications simultaneously and automatically. Users, applications and data were supposed to flow seamlessly through the infrastructure. Instead, all too often, they keep hitting performance brick walls – forced to wait while IT scrambles to work-around environmental inconsistencies. The problem is IT lacks the end-to-end view of the data center that allows it to proactively manage the infrastructure instead of continually reacting to it.
IT’s goal should be to provide users and application owners with the exact level of performance they need at any given point in time. No more, no less. Delivering more performance than users need is wasting IT and organizational budgets. Failure to deliver enough performance means users aren’t able to work up to their full potential or worse, lost revenue. That leads to dissatisfaction and lack of confidence in IT. The challenge is delivering “just the right amount of resources” is walking a fine line. That line needs monitoring, management, and forecasting to make sure IT is able to meet its SLA commitments.
There are various performance management tools to aid IT in meeting these commitments. The problem is that they are also themselves silo’ed. They are either application focused, application performance management (APM) or infrastructure focused, infrastructure performance management (IPM). Each of these are then silo’ed again into various types of infrastructure: server virtualization, network, storage, etc..
The State of Data Center Modernization
Most organizations are pushing IT to modernize their data center, with making it more cloud-like as the eventual goal. They are looking for something that is self-service and automatically adjusts to user and application demands on-the-fly.
As it stands today, the typical data center has scaled beyond human comprehension, exacerbated by virtualization. Virtualization has both helped and hurt IT in its effort to tame the data center beast. Virtualization allows IT to quickly deploy compute for a new application by abstracting the applications from the server hardware. The problem is deploying a virtual machine still requires specific steps and knowledge regarding networking and storage. Virtualization’s abstraction also makes it difficult to determine the cause of any performance problem or inconsistency.
In an effort not to get caught off-guard, IT will typically over deploy compute, networking and storage resources. But even with seemingly more than enough resources, one user request or application process can trigger a noisy neighbor situation. Noisy neighbors occur when a virtual machine suddenly increases processor consumption or storage IO that it might starve out other VMs from accessing those resources. The result is a ripple effect of inconsistent performance throughout the enterprise, with no real indication of the original request that caused the problem in the first place. Without that information IT is forced, once again, to throw even more hardware at the problem wasting even more IT budget.
Silo’ed Management for an Enterprise Problem
IT, instead of spending time on modernization efforts is drowning in a sea of day-to-day task and interrupt driven fire-fighting drills. To help they have enlisted the aid of multiple management tools to provide them with the insight into the performance problems they are encountering. The problem is most of these solutions are silo’ed. While they often can provide excellent details on the particular component they monitor, the tools can’t correlate that information across the other infrastructure components and have no understanding of the applications they are affecting.
The lack of end-to-end management is especially a challenge in the highly virtualized data center of today. And, if not corrected, will be even worse as organizations move from highly virtualized environments to high containerized environments. The reality is most virtualization management tools only manage out to the physical server, and most infrastructure tools manage up to the physical server. IT lacks a solution that can cross the chasm and provide high-quality monitoring and analysis of the entire infrastructure.
Without the end-to-end view of the environment, IT must take a brute force route to reducing performance inconsistencies by doing what it has always done, throw hardware at the problem and try to manually correlate data across systems. As a result, data centers end up with too many servers, that are too large, networks with too much bandwidth and all-flash storage systems that sit idle most of the time – in addition to the significant staff time that is wasted in troubleshooting problems.
What IT Needs – App-centric Infrastructure Performance Monitoring
An application is the main entry point into the IT infrastructure for most users. It makes sense that an end-to-end management strategy starts at this point. Starting at the application is a finer level of granularity than the VM and can provide application-specific insight. The next point in the infrastructure that users will notice is the other end of the data center stack, the storage infrastructure. Understanding how the storage architecture is responding to application requests especially when under load from multiple applications, is key to delivering predictable, consistent performance to users. Of course, all the components in-between need management and monitoring as well. IT needs the best of both. It needs an Infrastructure Performance Management solution that understands where applications live and how they are interacting with the infrastructure at any given moment – an App-centric IPM solution.
Another key need for IT is for the solution to provide a true real-time perspective instead of an occasional polling of the end-to-end infrastructure. The speed at which IO requests occur and are processed today means too much can happen in-between one to five minute polling requests and IT may miss the key piece of information that would have alerted them to the problem.
IT also needs the solution to provide decision assistance. Most IT professionals are too busy, and there’s simply too much data to gather for IT to process all the information and make decisions as quickly as may be required. Instead, the solution needs to employ machine learning techniques, so the tool can move beyond collecting data but make a recommendation based on previous best practices knowledge bases. Decision assistance allows troubleshooting by the operations team instead of a “guru” who may or may not be readily available.
There are of course other tools in the environment like IT Service Management (ITSM) tools, IT Operations Analytics tools (ITOA) and other application and infrastructure management tools. The solution needs to integrate with and complement these tools to further break down silos.
Finally, the solution also needs to provide IT with the ability to engage persons outside of IT like CxO’s, line of business managers and application teams. Providing real-time visibility on infrastructure health, utilization and performance will give these groups greater confidence in IT’s ability to execute, as well as a keener understanding of what current data center limits are.
Virtual Instruments VirtualWisdom: An Architecture to Get Modernization Back on Track
VirtualWisdom, whose roots started in storage infrastructure performance management has evolved into an end-to-end infrastructure performance management solution. There are two key inflection points required for an infrastructure performance management (IPM) solution to capture the needed data. The first is at the software layer. The IPM solution needs to assess the physical server at the virtual machine layer and capture data related to the application, the virtual machine and the physical hosts. VirtualWisdom is able to capture, correlate, and analyze data at all of these layers through its family of Probes and advanced Analytics.
The other key inflection point is “on-the-wire.” The modern data center is still made up of a lot of physical hardware and that hardware at some point (or multiple points) connects via a network. The mistake most solutions make is, if they collect this data at all, they do so via polling intervals, which as explained above may miss critical moments in managing performance. The capture of this data needs to be on-the-wire, either through specific hardware based non-intrusive taps or modern switches with analytics engines that the IPM tool can leverage.
VirtualWisdom can leverage taps that collect wire data in real-time across a variety of infrastructures, including both IP and fibre channel (FC). It can also interface with modern switch architectures that have their own metric capture capabilities built into them. The real-time analysis means that every IO transaction is captured and can be correlated with other activities going on elsewhere in the infrastructure.
VirtualWisdom is more than just a data capture tool. It leverages machine learning, workload intelligence and a human knowledge base built from a decade of experience troubleshooting infrastructure issues. That information, instead of presenting IT with data to plow through, leverages application-centric analytics to alert IT to potential future problems and provides guided methods to resolve those problems. VirtualWisdom is the first IPM solution to enable IT managers to visualize their infrastructure in the context of the applications being supported. It’s unique ability to discover and map applications to infrastructure and analyze application workload behavior in real-time gives IT teams a fast-track to resolving issues collaboratively.
To move forward to data center modernization, IT needs to make sure it is not bogged down in day-to-day tasks and reacting to problems only after they reach critical mass, a.k.a firefighting. It needs an infrastructure performance management solution that is app-centric so an end-to-end view of the data center can be created in the context of the applications. Companies like Virtual Instruments are moving in the end-to-end direction enabling IT to respond and be more proactive in its management of the infrastructure, and enable it to do more with less, which will drive down costs while increasing user satisfaction.