APM and Utility Computing

Utility Computing is currently more of a concept than a reality – primarily because the industry has not figured out how to really solve all of the problems that must be solved in order for an application system to be delivered as a utility in the same manner that electricity is delivered to the outlet in your home or office. Let’s have a look at the problems and how they are being addressed:

1. Storage must become virtualized. The good news is that substantial progress towards this has been made by the SAN vendors, who have largely succeeded in delivering offerings that separate an application or application system from dependence upon any single physical entity of storage.
2. The Network must be become virtualized. If your network has a decent level of physical redundancy (multi-homed NIC’s in servers, redundant switches and routers, etc.), this combined with the inherent ability of TCP/IP to reroute itself is also very close to reality. The one fly in the ointment is that the newest network application (VOIP) places demands upon the continuous reliability of the TCP/IP network that no other application has placed to date.
3. The Server and Operating System level of resources must become virtualized. Thanks to continued great work by VMWare (now part of EMC, a leading vendor of SAN’s and storage virtualization), a server can now live as an instance of software (a VM with its associated OS and applications). There are still issues with how well a virtual server performs and scales relative to a physical server for some applications, but these are being worked on and progress is being made.
4. The Applications that deliver the real business value to the users of the Utility must become consistently reliable in their delivery of functionality and performance to their users.

It is in this last area of consistent applications performance that Utility Computing meets its greatest and mostly unresolved challenge. Supporting an applications system is still very much a “hand crafted by elves” exercise wherein each new application and each new release brings with it new challenges that must be resolved by expensive and dedicated domain experts (in many cases actual applications developers who “know the code”). In order to resolve this last layer of virtualization (virtual applications that just work), one of two things must happen:

1. The Application Development community must agree upon and then implement a standard set of application instrumentation in the source code of all newly developed applications. While this is the ideal solution, such a standard does not exist, and even once it exists, it would take a decade or two for all of the existing applications to get rewritten to incorporate such instrumentation.
2. Application Performance Management tools (products) must evolve that can notice when performance and reliability issues have arisen and then through “under the covers” code instrumentation that is applied outside of source code perform meaningful and actionable root cause analysis.

The Application Performance Management (APM) industry has a long way to go before it can deliver on #2 above in any kind of a meaningful and comprehensive way. Vendors like Wily Technology can deliver upon effective instrumentation and root cause analysis for a particular environment (in the case of Wily, J2EE based applications). The incumbent set of infrastructure monitoring vendors (CA, HP, BMC, NetIQ, IBM/Tivoli) are largely useless in pursuit of this objective. Point solutions exist for non-J2EE web applications from a variety of very small startups. Perhaps the most encouraging source of progress will come from vendors like Identify Software with their AppSight product, who come from the world of helping developers and application support teams do root cause analysis of applications problems, but who are not currently built or sold to be used as production applications monitoring solutions. The combination of products like AppSight, with products that can do fine grained analysis of production applications performance (no scripts), may in fact yield the best chance for solving this layer of the problem.

In summary, real Utility Computing is not going to happened until the APM industry makes applications as virtual, reliable and manageable as the network is today. A substantial opportunity exists for the right set of vendors to partner or merge with each other to solve this problem. We’ll just have to see what happens.

Bernd Harzog
CEO
APM Experts
www.apmexperts.com
bernd.harzog@apmexperts.com

ProactiveNet and Coradiant Partner: 33% + 33% = 66%

ProactiveNet and Coradiant Partner to Provide Real-Time Web Application Performance Analytics

ProactiveNet and Coradiant have announced a technology partnership and the seamless integration of their two products. The strength of the ProactiveNet solution has always been its ability to correlate degradation in back end infrastructure metrics with performance data. Unfortunately, ProactiveNet has always relied upon scripted synthetic agents to generate performance information, which has several drawbacks (see "Why Is Measuring applications Performance With Scripts a Bad Thing)." The strength of the Coradiant solution is that it can gather real application performance metrics on every transaction issued by every user. Unfortunately in order to accomplish this task, you have to insert the Coradiant appliance into your network via either a network tap or a mirror port on your switch (and of course pay the extra money for the appliance).

While this is good progress for both companies, at the end of the day it still leaves them 33% short of a true APM solution. A true APM solution must incorporate true response time measurement, with back end infrastructure root cause analysis, and also provide insight into the applications layer of the system should the problem reside there. This solution might be appropriate for enterprises that run packaged applications and who therefore have no access to the underlying source code, but it is still 33% short of the mark for any enterprise that has custom developed applications as a part of the mix.

Bernd Harzog
CEO
APM Experts
bernd.harzog@apmexperts.com
www.apmexperts.com