For quite some time VMware has been letting it known that it intends to be a management software vendor on top of its own virtualization platform. In fact VMware has already taken a series of steps that prove that management is not just an intention, but will be a strategy fulfilled with product deliveries. These actions include buying B-hive and delivering the resulting AppSpeed product, buying the four management products from EMC/Ionix which resulted in Configuration Manager, Application Discovery Manager, and Service Manager (Fastscale seems to have gotten lost in the shuffle), developing and delivering CapacityIQ, buying SpringSource with which came Hyperic and integrating it into vCenter,Â andÂ delivering Chargeback.
Management Strategy Articulated at VMworld
At VMworld 2010 in San Francisco, VMware worked hard to flesh out its management strategy and also made an importantÂ acquisitionÂ announcement that we will get to in a moment. The presentation of the strategy was started by the sharing of some survey results in which VMware listed some key management areas where VMware said that their customers were asking VMware to step up and improve the situation:
|Virt. Â Management Objective||Percent Who Responded “Important”||VMW Strategy|
|Simplify IT infrastructure management to reduce OPEX||> 50%||Monitoring server performance and availability|
|Streamline application management for the cloud||> 35%||Managing applications performance for virtualized applications|
|Enable agile, self service IT||>30%||Improving VM provisioning speed and control|
In other words, VMware started its articulation of its management strategy by saying that their customers were asking them to get into the server monitoring, application monitoring and self-service/IT-as-a-Service businesses. VMware went on to further justify these new efforts with the following observations:
- Management is too complex and too expensive in the physical world, with unique and overlapping management processes and management products for various horizontal and vertical silos in theÂ environment
- That the rigid processes (read this as ITIL) are inappropriate for new dynamic and cloud based environments
- ThatÂ insufficientÂ automation exists which in turn means that manual actions by administrators cannot keep up with changes in theÂ environment
The IntegrienÂ Acquisition
The last point above lead directly into the announcement that VMware was buying Intregrien, one of the two standalone vendors of self-learning performance analytics (Netuitive is the other vendor in this space and is by far the market leader in terms of installed customers and revenue). This acquisition by VMware is extremely significant because it says that VMware now recognizes a key point that we made in our Root Cause and Tier 1 Applications post which is that in a dynamic data center it is impossible to deterministically trace root cause, and that therefore a real time and self learning approach to root cause will be needed.
So VMware has come out and said that you need to monitor your servers for performance andÂ availability, you need to monitor your virtualized applications for performance, you need a real time learning overlay on top of all of this to link applications performance issues with root causes in the underlyingÂ infrastructure, and you need an IT as a Service management solution glued onto the side of all of this. This vision is depicted in the diagram on the left side of the table below. We believe that VMware’s diagram is missing a layer on the monitoring side, which is Infrastructure Performance Management. The reason for this belief is that you can monitor servers (physical and virtual) to your heart’s content but that doing so will not tell you if your entire infrastructure (end-to-end from the user to the guest to the spindle in the array) is responding in a manner adequate to the requirements of your applications. For this you need an Infrastructure Performance Management solution that actually measures Infrastructure Response Time across your entire virtual and physical infrastructure. The solutions that meet this need are highlighted in Infrastructure Performance Management Heats Up.
|VMware’s Vision||Virtualization Practice Vision|
The Reality Today
The reality today is that organizations who are actually virtualizing business critical applications, or are considering placing such applications in public clouds there are significant gaps between either vision above and what you can buy from VMware to fulfill the needs presented. These gaps include:
- VMware does not have a Server Performance andÂ AvailabilityÂ Monitoring Solution. Hyperic was such a solution when VMware purchased SpringSource but it appears that Hyperic has been focused upon managing applications running on vFabric as it has now been labeled as the vFabric Hyperic and is listed in the vFabric Cloud Applications Platform section of the VMware product catalog. VMware does have CapacityIQ which meets part of the need (the capacity management part, but not the performance management part). Therefore addressing this need should start with finding a broad scale Resource and AvailabilityÂ MonitoringÂ solution that supports both the physical and the virtual environment. Some good choices to start your search are listed in the Resource andÂ AvailabilityÂ Management column below.
- We have made it very clear Â that in virtualized and cloud based systems, one cannot infer the actual performance of theÂ environmentÂ from resource utilization statistics. This gives rise to a new category – Infrastructure Performance Management – solution that actually calculate Infrastructure Response Time. VMware has no product in this category at all, so enterprises looking for a solution in this area should evaluate the five products listed in this category below.
- In the Applications Performance Management category, VMware has three products. It appears that vFabric Hyperic is the APM solution of choice if you are running your application on vFabric (Spring TC Server with the recent messaging and data cachingÂ acquisitionsÂ integrated in). If you are running a web-Java/.Net-database application in your internal environment AppSpeed appears to be the recommendation, although VMware was noticeably silent on AppSpeed at the conference (there was not a single session on AppSpeed at VMworld 2010). Finally VMware has taken the N-Layers product acquired from Ionix and re-branded it as Application Discovery Manager. However there appears to be no integration between the ADM product and the APM products that would presumably benefit from know the topology of the applications they are monitoring. There are many APM solutions that have been around for a long time (CA/Wily, HP Diagnostics, IBM ITCAM, and Quest Foglight), with HP Quest and CA listed by Gartner as leaders in this space. New Relic and AppDynamics have next generation APM solutions that are specifically built for virtualization and cloud deployment scenarios. It is also the case that the selection of these solutions is usually made by the team that owns the application, not the team that owns the virtualÂ infrastructure which negates any inherent advantages VMware might have in this space.
- On the It as a Service front we have VMware’s newest product vCloud Director. This is the product that enterprises who want to put up private clouds and service providers who use vCloud should use as the self-service and provisioning Â engine. While vCloud Director is an important step forward for VMware, it is the first release of this product which enters a category that has quite a number of third party solutions that are in their third or fourth major releases. The third party solutions are at this point like to be more mature, will likely interface with more complementary solutions to handle physical provisioning scenarios, and are more likely to span multiple virtualization platforms.
- Lastly with theÂ acquisitionÂ of Integrien, VMware has bought one of the two vendors that provide automated self learning performance analysis systems. We will write an entire post on how to understand and evaluate these solutions. For right now, let’s just start that process with stating that it is very important that these solutions have a broad range of connectors into various monitoring solutions (they typically do not collect their own data), and that these solutions be as fully automated and self-learning as possible.
VMware and Third Party Solutions By Category
While VMware has articulated the need for a management strategy and has provided some building blocks for its management stack, there are currently and will be for the near term future significant gaps in the VMware management offerings even when the domain of the problem is constrained to the management of the VMware platform. Once the domain is expanded to include the physical infrastructure which underlies the virtualization platform, and once it is again expanded to include multiple virtualization platforms the use case is outside of what VMware intends to provide. For these reasons, third party solutions should be considered for each component of the diagram above when evaluating solutions from VMware.