Skip to main content

APM for Enterprise: How Does It Scale?

Larry Haig

It is easy to feel that so called "second generation" Application Performance Management (APM) tooling rules the world.

And for good reason, many would argue – certainly the positive disruptive effects of support for highly distributed / Service Orientated architectures, and the requirements of many fast moving businesses to support a plethora of different technologies are a powerful dynamic. That leaves aside the undoubted advantages of comprehensive traffic screening (as opposed to "hard" sampling), ease of installation and commissioning (relative in some cases), user accessibility, flexible reporting and tighter productive association between IT and business – in short, empowering the DevOps and PerfOps revolution.

So, modern APM is certainly well attuned to the requirements of current business. What's not to like?

Could these technologies have an Achilles heel? Certainly, they are generally strong on lists of customer logos, but tight lipped when it comes to detailed high volume case studies.

Hundreds or thousands of JVMs and moderately high transaction volumes are all very well (and well attested), but how do these technologies stack up for the high end enterprise? What other options might exist?

It could be argued that an organization with tens of thousands of JVMs and millions of metrics has a fundamentally different issue than those closer to the base of the pyramid. Certainly these organizations are fewer in number, but that is scant comfort for those with the responsibility of managing their application delivery. Whether in banking/financial trading, FMCG or elsewhere, the issue of effectively analyzing daily transaction flows at high scale is real. The situation is exacerbated at peak – one large UK gaming company generates 20-30,000 events per second during a normal daily peak. During the popular Grand National race meeting, traffic increases 5-10 times – creating the need to transfer several terabytes a day into an APM data store.

The question is: which if any of the APM tools can even come close to these sorts of volumes?

It is certainly possible to instrument these organizations with second generation APM – but what snares lie in wait for the unwary, and what compromises will have to be made?

To some extent, the answer depends upon the particular technology deployed. All will have their own weaknesses, but those architected around collector/analysis servers are likely to be particularly vulnerable to the effects of extreme data volume unless high scale technology/architectural interventions have been made "under the covers". Cloud based solutions may duck this bullet (although they are not guaranteed to do so), but come with their own security concerns, at least in theory.

So, you are a high volume Enterprise, and have plumped for second generation APM. What issues may arise? Essentially, software agent based APM is likely to evidence stress in one or more of three principal areas:

■ Length of data storage/"live" access

■ Data granularity

■ Production system performance overhead

Compromises essentially hinge around reducing the data flows processed by the APM to reduce the amount of data written to disk, or improving the inherent efficiency of such data handling. Traditionally, this involves sampling rather than screening all transactions; and this is an option for some. However, sampling has no value for businesses needing to identify and analyze a particular single customer session.

Other approaches are to increase the hardware capacity of collector/server components, or reducing the application server to collector ratio. Either way, these compromises run the risk of eroding the underlying value proposition supporting much of second generation tool philosophy. In addition they will push the architecture of these solutions to their limit and potentially expose fundamental issues in how they scale.

Open Source approaches to extreme scale have evolved using NoSQL – creating products such as Hadoop and ElasticSearch. The pedigree of these is generally good, in that they have been developed as strategies within companies such as Google and Facebook to deal with the problems of ultra-high volume environments.

Certainly, integration of these technologies into their tooling by APM vendors can be a potential solution, providing that they have been architected/implemented appropriately – and tested with extreme scale in mind.

Given that most if not all major volume Enterprises have de facto constraints on their flexibility and speed of adoption of extension technologies (not to mention change generally), perhaps there is a case for revisiting "traditional" APM tooling models. These certainly had (and have) a track record of delivering value in large enterprise deployments, albeit without some of the bells and whistles offered by later entrants. Any high scale developments made by these vendors would certainly have the advantage of leveraging the often considerable sunk investment made in them.

Provided that any constraints are well understood, and appropriate investment is made in initial commissioning and ongoing support, then this option would in our view be worth adding to the mix – for consideration, at least.

Alternatively, perhaps a "dual tool" approach may have validity – second generation APM pre-production, and traditional high volume solutions in the live environment.

For Enterprises with extremely strong nerves, and appropriate skills, "building your own" using Open Source technologies is a possibility, although it is likely to be both extremely high risk and costly. Such an approach comes with its own ongoing maintenance challenges as well.

We would like to see more open sourcing of the key components of APM, for example the agents that instrument Java and .Net applications. These, conforming to open standards, enable a flexible approach to open-APM. Choose your agents, your transport method (Apache Flume, FluentD etc.), and your data storage and analysis methods (Elastic Kibana) that are appropriate for your scale and company skillset.

Either way, we would strongly suggest that major enterprises face these issues squarely, and certainly not make significant investments in APM without appropriate high volume (production scale) Proof of Concept preliminary trialling.

Above all, put little trust in marketing. Prove it in your environment – ideally in production.

Larry Haig is Senior Consultant at Intechnica.

This blog was written with contributions by James Billingham, Performance Architect at Intechnica.

Hot Topics

The Latest

In MEAN TIME TO INSIGHT Episode 14, Shamus McGillicuddy, VP of Research, Network Infrastructure and Operations, at EMA discusses hybrid multi-cloud network observability... 

While companies adopt AI at a record pace, they also face the challenge of finding a smart and scalable way to manage its rapidly growing costs. This requires balancing the massive possibilities inherent in AI with the need to control cloud costs, aim for long-term profitability and optimize spending ...

Telecommunications is expanding at an unprecedented pace ... But progress brings complexity. As WanAware's 2025 Telecom Observability Benchmark Report reveals, many operators are discovering that modernization requires more than physical build outs and CapEx — it also demands the tools and insights to manage, secure, and optimize this fast-growing infrastructure in real time ...

As businesses increasingly rely on high-performance applications to deliver seamless user experiences, the demand for fast, reliable, and scalable data storage systems has never been greater. Redis — an open-source, in-memory data structure store — has emerged as a popular choice for use cases ranging from caching to real-time analytics. But with great performance comes the need for vigilant monitoring ...

Kubernetes was not initially designed with AI's vast resource variability in mind, and the rapid rise of AI has exposed Kubernetes limitations, particularly when it comes to cost and resource efficiency. Indeed, AI workloads differ from traditional applications in that they require a staggering amount and variety of compute resources, and their consumption is far less consistent than traditional workloads ... Considering the speed of AI innovation, teams cannot afford to be bogged down by these constant infrastructure concerns. A solution is needed ...

AI is the catalyst for significant investment in data teams as enterprises require higher-quality data to power their AI applications, according to the State of Analytics Engineering Report from dbt Labs ...

Misaligned architecture can lead to business consequences, with 93% of respondents reporting negative outcomes such as service disruptions, high operational costs and security challenges ...

A Gartner analyst recently suggested that GenAI tools could create 25% time savings for network operational teams. Where might these time savings come from? How are GenAI tools helping NetOps teams today, and what other tasks might they take on in the future as models continue improving? In general, these savings come from automating or streamlining manual NetOps tasks ...

IT and line-of-business teams are increasingly aligned in their efforts to close the data gap and drive greater collaboration to alleviate IT bottlenecks and offload growing demands on IT teams, according to The 2025 Automation Benchmark Report: Insights from IT Leaders on Enterprise Automation & the Future of AI-Driven Businesses from Jitterbit ...

A large majority (86%) of data management and AI decision makers cite protecting data privacy as a top concern, with 76% of respondents citing ROI on data privacy and AI initiatives across their organization, according to a new Harris Poll from Collibra ...

APM for Enterprise: How Does It Scale?

Larry Haig

It is easy to feel that so called "second generation" Application Performance Management (APM) tooling rules the world.

And for good reason, many would argue – certainly the positive disruptive effects of support for highly distributed / Service Orientated architectures, and the requirements of many fast moving businesses to support a plethora of different technologies are a powerful dynamic. That leaves aside the undoubted advantages of comprehensive traffic screening (as opposed to "hard" sampling), ease of installation and commissioning (relative in some cases), user accessibility, flexible reporting and tighter productive association between IT and business – in short, empowering the DevOps and PerfOps revolution.

So, modern APM is certainly well attuned to the requirements of current business. What's not to like?

Could these technologies have an Achilles heel? Certainly, they are generally strong on lists of customer logos, but tight lipped when it comes to detailed high volume case studies.

Hundreds or thousands of JVMs and moderately high transaction volumes are all very well (and well attested), but how do these technologies stack up for the high end enterprise? What other options might exist?

It could be argued that an organization with tens of thousands of JVMs and millions of metrics has a fundamentally different issue than those closer to the base of the pyramid. Certainly these organizations are fewer in number, but that is scant comfort for those with the responsibility of managing their application delivery. Whether in banking/financial trading, FMCG or elsewhere, the issue of effectively analyzing daily transaction flows at high scale is real. The situation is exacerbated at peak – one large UK gaming company generates 20-30,000 events per second during a normal daily peak. During the popular Grand National race meeting, traffic increases 5-10 times – creating the need to transfer several terabytes a day into an APM data store.

The question is: which if any of the APM tools can even come close to these sorts of volumes?

It is certainly possible to instrument these organizations with second generation APM – but what snares lie in wait for the unwary, and what compromises will have to be made?

To some extent, the answer depends upon the particular technology deployed. All will have their own weaknesses, but those architected around collector/analysis servers are likely to be particularly vulnerable to the effects of extreme data volume unless high scale technology/architectural interventions have been made "under the covers". Cloud based solutions may duck this bullet (although they are not guaranteed to do so), but come with their own security concerns, at least in theory.

So, you are a high volume Enterprise, and have plumped for second generation APM. What issues may arise? Essentially, software agent based APM is likely to evidence stress in one or more of three principal areas:

■ Length of data storage/"live" access

■ Data granularity

■ Production system performance overhead

Compromises essentially hinge around reducing the data flows processed by the APM to reduce the amount of data written to disk, or improving the inherent efficiency of such data handling. Traditionally, this involves sampling rather than screening all transactions; and this is an option for some. However, sampling has no value for businesses needing to identify and analyze a particular single customer session.

Other approaches are to increase the hardware capacity of collector/server components, or reducing the application server to collector ratio. Either way, these compromises run the risk of eroding the underlying value proposition supporting much of second generation tool philosophy. In addition they will push the architecture of these solutions to their limit and potentially expose fundamental issues in how they scale.

Open Source approaches to extreme scale have evolved using NoSQL – creating products such as Hadoop and ElasticSearch. The pedigree of these is generally good, in that they have been developed as strategies within companies such as Google and Facebook to deal with the problems of ultra-high volume environments.

Certainly, integration of these technologies into their tooling by APM vendors can be a potential solution, providing that they have been architected/implemented appropriately – and tested with extreme scale in mind.

Given that most if not all major volume Enterprises have de facto constraints on their flexibility and speed of adoption of extension technologies (not to mention change generally), perhaps there is a case for revisiting "traditional" APM tooling models. These certainly had (and have) a track record of delivering value in large enterprise deployments, albeit without some of the bells and whistles offered by later entrants. Any high scale developments made by these vendors would certainly have the advantage of leveraging the often considerable sunk investment made in them.

Provided that any constraints are well understood, and appropriate investment is made in initial commissioning and ongoing support, then this option would in our view be worth adding to the mix – for consideration, at least.

Alternatively, perhaps a "dual tool" approach may have validity – second generation APM pre-production, and traditional high volume solutions in the live environment.

For Enterprises with extremely strong nerves, and appropriate skills, "building your own" using Open Source technologies is a possibility, although it is likely to be both extremely high risk and costly. Such an approach comes with its own ongoing maintenance challenges as well.

We would like to see more open sourcing of the key components of APM, for example the agents that instrument Java and .Net applications. These, conforming to open standards, enable a flexible approach to open-APM. Choose your agents, your transport method (Apache Flume, FluentD etc.), and your data storage and analysis methods (Elastic Kibana) that are appropriate for your scale and company skillset.

Either way, we would strongly suggest that major enterprises face these issues squarely, and certainly not make significant investments in APM without appropriate high volume (production scale) Proof of Concept preliminary trialling.

Above all, put little trust in marketing. Prove it in your environment – ideally in production.

Larry Haig is Senior Consultant at Intechnica.

This blog was written with contributions by James Billingham, Performance Architect at Intechnica.

Hot Topics

The Latest

In MEAN TIME TO INSIGHT Episode 14, Shamus McGillicuddy, VP of Research, Network Infrastructure and Operations, at EMA discusses hybrid multi-cloud network observability... 

While companies adopt AI at a record pace, they also face the challenge of finding a smart and scalable way to manage its rapidly growing costs. This requires balancing the massive possibilities inherent in AI with the need to control cloud costs, aim for long-term profitability and optimize spending ...

Telecommunications is expanding at an unprecedented pace ... But progress brings complexity. As WanAware's 2025 Telecom Observability Benchmark Report reveals, many operators are discovering that modernization requires more than physical build outs and CapEx — it also demands the tools and insights to manage, secure, and optimize this fast-growing infrastructure in real time ...

As businesses increasingly rely on high-performance applications to deliver seamless user experiences, the demand for fast, reliable, and scalable data storage systems has never been greater. Redis — an open-source, in-memory data structure store — has emerged as a popular choice for use cases ranging from caching to real-time analytics. But with great performance comes the need for vigilant monitoring ...

Kubernetes was not initially designed with AI's vast resource variability in mind, and the rapid rise of AI has exposed Kubernetes limitations, particularly when it comes to cost and resource efficiency. Indeed, AI workloads differ from traditional applications in that they require a staggering amount and variety of compute resources, and their consumption is far less consistent than traditional workloads ... Considering the speed of AI innovation, teams cannot afford to be bogged down by these constant infrastructure concerns. A solution is needed ...

AI is the catalyst for significant investment in data teams as enterprises require higher-quality data to power their AI applications, according to the State of Analytics Engineering Report from dbt Labs ...

Misaligned architecture can lead to business consequences, with 93% of respondents reporting negative outcomes such as service disruptions, high operational costs and security challenges ...

A Gartner analyst recently suggested that GenAI tools could create 25% time savings for network operational teams. Where might these time savings come from? How are GenAI tools helping NetOps teams today, and what other tasks might they take on in the future as models continue improving? In general, these savings come from automating or streamlining manual NetOps tasks ...

IT and line-of-business teams are increasingly aligned in their efforts to close the data gap and drive greater collaboration to alleviate IT bottlenecks and offload growing demands on IT teams, according to The 2025 Automation Benchmark Report: Insights from IT Leaders on Enterprise Automation & the Future of AI-Driven Businesses from Jitterbit ...

A large majority (86%) of data management and AI decision makers cite protecting data privacy as a top concern, with 76% of respondents citing ROI on data privacy and AI initiatives across their organization, according to a new Harris Poll from Collibra ...