Rackspace a leading hosting and public cloud vendor has acquired Cloudkick, a Monitoring as a Service vendor focused upon monitoring the software infrastructure layer in Infrastructure as a Service (IaaS) clouds. Rackspace will now be able to offer basic monitoring of the operating system layer of the IaaS stack as a service to its customers, and also use the very same dashboards that the customers use for their own support staff to assist in troubleshooting. The interesting question raised by this acquisition is what does this mean for both the business of monitoring systems and applications in the cloud, and what does this mean for the 2000+ hosting vendors who are in the process of becoming public cloud vendors through VMware’s vCloud partner program?
Clearly the answer is not that every hosting/cloud vendor is going to go out and acquire a monitoring vendor. Keeping a monitoring product competitive and current is as software product management, development, testing, and marketing exercise that is outside of the core competencies of 99% of the prospective cloud vendors. Furthermore, since any cloud vendor that wants their own monitoring solution is much more likely to license one from an existing vendor than build one from scratch, one simply has to look at the set of vendors in this business to realize that there are not enough monitoring vendors to go around.
Therefore it is likely that this acquisition of Cloudkick by Rackspace is an exception rather than the start of a trend. The far more likely scenario is that vendors who offer Monitoring as a Service (profiled in this post) will start to find a very receptive audience at the cloud vendors when it comes to resale and partnership conversations.
What Should Cloud Vendors Do?
If cloud vendors want to attract more than the most tactical/development oriented workloads they need to offer two levels of performance assurance that they do not for the most part offer today. The first task is to be able to ensure the performance of the infrastructure that they provide to their customer. By performance in this case, it is not sufficient to republish the resource utilization statistics gathered from the guest OS or the hypervisor to the customer – rather what is meant by performance is that when a request for work is placed on the stack provided by the cloud vendor, the cloud vendor can measure and commit to the latency (response time) with which the cloud vendors’ stack responds to that request.
If the above is the requirement for infrastructure performance assurance, then it is clear that the Rackspace acquisition of Cloudkick is irrelevant as Cloudkick provided nothing but commodity monitoring functionality as a service and does not provide anything close to the infrastructure performance assurance that is described above as being necessary for cloud vendors to grow their businesses.
The next step for the cloud vendors is to address the question of applications performance management or APM. In all cases except for SaaS based clouds, the cloud vendor is not responsible for the application itself, however, it is in the interest of the cloud vendor to make the performance management of applications hosted on their clouds as easy for the customer as possible.
How Will These Needs be Met?
It is clear that cloud vendors have no business trying to be monitoring vendors. For the reasons listed above, being a monitoring vendor is outside of the core competence of the cloud vendors. Therefore cloud vendors will have to partner with vendors who offer monitoring as a service for the various layers of IaaS and PaaS stacks.
The Cloud Infrastructure Performance Management problem is the most challenging problem in this mix. There is only one MaaS vendor today who offers an effective Infrastructure Performance Management solution on a multi-tenant MaaS basis and that is AppFirst. There are other vendors who offer very good IPM solutions (Akorri, Xangati, Virtual Instruments, and CA (Virtual Assurance)), but these solutions are currently designed for on premise use at enterprises and are not offered on a multi-tenant MaaS basis.
On the Applications Performance Management side, there is a much easier answer. New Relic has already proven the viability of APM as a Service for Java, .NET, Ruby, and PHP environments by demonstrating its market leadership in this space with thousands of paying customers and numerous partnerships with cloud vendors.
What Should Cloud Customers Do?
If you are going to put a performance critical application into a public cloud you need to address two performance management issues:
- You should demand of your cloud provider that they be able to prove the responsiveness of their “stack” to you on a continuous basis and alert you whenever that responsiveness falls below an agreed upon threshold. This is an essential part of your being able to know if you application is running well, and if it is not where you should start the troubleshooting process.
- You should procure either on your own or through the cloud vendor an APM solution that is offered on a cloud friendly basis. This does not necessarily mean that you have to go with the most convenient route of using a MaaS APM service like New Relic. You could choose instead to go with an APM solution that allows for its agents to live behind firewalls and “phone home” to the APM management system installed in your data center. Such offerings are available from AppDynamics, dynaTrace, Opnet, and Optier. You will likely find that your legacy APM solution that you use to manage your current Java application in production will not meet these needs.
The acquisition of Cloudlick by Rackspace points out the need for IaaS cloud vendors to get serious about offering an Infrastructure Performance Management solution to their customers – but fails to deliver such a solution to the customers of Rackspace. Cloud customers should focus upon finding true cloud ready Infrastructure Performance Management and Applications Performance Management solutions as a part of putting performance critical applications in public clouds.