Datadog Reviews

Datadog provides us with a solution for data ingesting for all of our application metrics, resource metrics, APM/tracing data etc. 

We use it for use in dashboards, monitoring/alerting, SLO targets, incident response etc. 

We have a lot of applications across multiple languages/frameworks etc., and have deployed in Kubernetes across multiple regions in AWS, along with underlying managed resources such as SQS, Aurora, etc. 

Datadog makes understanding the state of these seamless. We are a company with millions of daily active users, and this level of detail is excellent.

Datadog has allowed us to rapidly spin up alerting and monitoring that helps our incident responders get alerted quickly when our SLOs are in danger and helps to quickly resolve issues. 

It is the single most important tool we have from an SRE perspective. 

It also provides us with an easy way to get information at a glance for all of our services through APM and create unified dashboards that track our underlying resources, such as databases, queues, etc., alongside application data. 

It has been invaluable to our organization.

The management of SLOs and their related burn-rate monitors have allowed us to onboard teams to on-call fast. 

Management of resources using infrastructure-as-code has been a recent game-changer for us. Combining the two has allowed us to provide product teams with a total solution for getting their applications attached to user-focused alerting and monitoring within a matter of days rather than months - and has clearly impacted our ability to discover and respond to significant production incidents.

Managing dashboards as IaC is a bit hard to work out at times. I use custom tools to convert JSON dashboards to Terraform resources. Ideally, I'd like for some sort of building tool for this to be built into the app. For example, a templating system that can easily be exported to IaC would be transformative for us. 

There are also some aspects of the API that can be a bit verbose - especially in the area of new features like SLOs - and take some time to understand. That said, overall, they're well-documented enough to be a minor concern for us.

I've been using the solution for over five years.

I have never seen a major outage that prevented us from using Datadog, although I can't speak for other teams/time zones

This product is massively scalable - I haven't seen any issues as we continue to onboard new technologies and teams

Datadog provides us with a number of direct lines to support, although I haven't personally required their assistance.

We previously used LightStep for APM and switched to Datadog to unify all of our application data.

Most elements are quite simple to set up. However, some types of data collection require organization-wide engineering buy-in.

We handled the initial setup in-house.

I have been using Datadog products and capabilities increasingly over the last 4 years, from POC to widespread adoption. 

The capabilities we use are unique for each use case and can be combined in various ways to provide the full observability coverage needed to maintain stable operations and shift from becoming more reactive to proactive. 

Our organization uses both site/service reliability for the range of backend and frontend services, custom monitoring, and dashboards that can be dynamic and reused for multiple teams.

The capabilities we use are unique for each use case. They can be combined in various ways to provide the full observability coverage needed to maintain stable operations in order to become more proactive. 

Our organization uses both site/service reliability for backend and frontend services. Custom monitoring and dashboards that can be dynamic and reused for multiple teams. 

We continue to increase the size of our footprint as we get more and more positive experiences.

The APM, RUM, synthetics, and infrastructure host maps have been some of the most popular and commonly used features. 

Overall, the Data UI and the usability of customer features continue to improve. 

The RUM session data and replays are much more convenient and applicable than other tools I have worked with in the past, and by combining multiple capabilities or features together, there is full visibility across the technology stacks and can identify specific bottlenecks or areas for risk and vulnerabilities to be likely to exist. 

Watchdog insights take the work out of the hardest part, helping us identify the issues before our customers.

It is very difficult to make the solutions fit perfectly for large organizations, especially in terms of high cardinality objects and multi-tenancy, where the data needs to be rolled up to a summarized level while maintaining its individual data granularity and identifiers. Tagging is imperative. However, the solutions could be improved for these needs in the future.

I've used the solution for over four years now.

The stability is excellent.

You can work with engineering to make it work for your needs. They are excellent at supporting their customers.

Technical support is excellent.

Neutral

I previously used New Relic, App Dynamics, Heap, Clicktale, and more. Datadog has incorporated many of the features we were looking for into a one-stop shop.

The initial setup is simple and straightforward.

We had an in-house team working directly with Datadog engineering support and technical enablement.

We looked into New Relic, App Dynamics, Heap, Clicktale, and more. Datadog has many of the features we were looking for in one place.

We use all versions of the solution.

We deploy various services for our main platform on AWS across multiple regions. We have a development environment, a staging environment, a QA environment, and a production environment. We deploy our many services across hundreds of instances. 

We have many server farms, all responsible for various services on our market intelligence platform. The deployment of each server farm or even individual instances varies depending on what stood up. We have instances built in three different ways, with two different pipelines and some even on user data scripts.

My team has a 24/7 on-call schedule where we need to be ready to handle and mitigate incidents with the platform at any moment. 

We have countless monitors set up on Datadog that alert directly to our queue using an email that generates a ticket. 

The actionable steps for each type of monitor and its associated incident are easily included in the alerts whenever something is triggered. We generate links to the Datadog monitors and can instantly drill down into what went wrong and for how long.

The features I have found most helpful are synthetic monitoring, APM, and alert features. The monitoring functionality, in general, and tagging infrastructure are great.

Synthetics have become bread and butter for us as we have migrated many tests over to Datadog. We have simplified and consolidated our synthetic tests while also making them more robust with the help of your tagging. 

A large portion of our monitoring is based on synthetics results, and alerts integrate seamlessly without an incident queue system. We use dashboards heavily. 

The metrics capabilities are extremely helpful, and we use virtually all of the widgets.

My main place of improvement for Datadog would be the documentation. While the tool is robust with many different capabilities, users would greatly benefit from more examples in the documentation. 

The number of current code snippets available in the docs is not enough, and some need to be updated even today. 

One function I would add would be a button to generate a report of the performance of a synthetic test and the performance of each of the steps in the test over time.

This timeline varies in terms of how long we've used the solution. We have one platform completely in the cloud and one still on-premises. We've had the solution for many years on AWS.

We primarily use Datadog for:

• Native memory
• Logging
• APM
• Context switching
• RUM
• Synthetic
• Databases
• Java
• JVM settings
• File i/o
• Socket i/o
• Linux
• Kubernetes
• Kafka
• Pods
• Sizing

We are testing Datadog as a way to reduce our operational time to fix things (mean time to repair). This is step one. We hope to use Datadog as a way to be proactive instead of reactive (mean time to failure).

So far, Datadog has shown very good options to work on all of our operational and development issues. We are also trying to use Datadog to shift left, and fix things before they break (MTTF increase).

We are currently in a POC and do not own Datadog at the moment. 

So far, there have been a few issues due to security. There are two main security issues. 

The first is moving data off-prem. This has been resolved to a point (filtering logs, etc). However, there is still an issue with moving a JFR as a JFR potentially contains data that is not allowed off-prem.

The second security issue is more internal, however, the main installation requires root access or using an ACL. Our company does not use ACLs on our Linux platform. This is problematic since the install sets a no-login on the Datadog user.

The most valuable aspects of the product include the APM and profiler.

These two have given us insights into things that are very difficult to track down given the standard OS (Linux) tools. 

The native memory tracking is super difficult to see exactly where it comes from. I attended a course (continuous profiling), and it showed me the potentially very important capabilities.

If you add these details to a standard dashboard, or a sub-dashboard for techy people, or even just a notebook, it would be easy to identify issues before they occur.

Combining these details with the basic tools (infra, logging, APM, and good rules), Datadog can easily show the details that a true engineer would need. It isn't just for monitoring, however, I see the value in it for engineers.

I have done every training offered (and in a short period of time: two days for 20 courses).

I find the training great. That said, it is set for the LCD (lowest common denominator). Of course, this is very helpful to sell the product, yet, to really utilize the product, you need to get more detailed.

If I did the training as it is written and I cut/paste a bunch of stuff and see the cut/paste work, I didn't really learn anything. Later sessions (I quit using the editor and switched to VI) stopped cutting and pasting, and learned much more.

I've used the solution for one month.

I' give stability a thumbs up.

We are not sure yet in terms of scalability. The off-prem solution seems to scale well (although had issues with the training slowing down).

Technical support is great.

Positive

I previously used Dynatrace and Elastic. We didn't switch. We are in a POC.

The initial setup is simple yet complex. There are too many teams are needed.

We did the initial setup in-house.

In terms of ROI, the labor saving is probably the biggest. The NPR is probably second - although management would probably reverse these.

Pricing and licensing is fairly complicated. A GB for .1 sounds great, however, once you put all 16 or so prices together, it adds up fast. A cost model sheet on the main site would be very helpful.

We are currently in a POC.

We work with all product versions.

We use Datadog for three main use cases, including:

• Infrastructure and application monitoring. It is ensuring that our services are available and performant at all times. This allows us to proactively address incidents and outages without customers contacting us. This includes monitoring of cloud resources (databases, load balancers, CPU usage, etc.), high-level application monitoring (response times, failure rates, etc.), and low-level application monitoring (business-oriented metrics and functional exceptions to customer experience.
• Analyzing application behavior, especially around performance. We often use Datadog's application performance monitoring on non-production environments to evaluate the impact of newly introduced features and gain confidence in changes.
• End-to-end regression testing for APIs and browser-based experiences. Using Datadog's synthetic testing checks periodically that the system behaves in the exact correct way. This is often used as a canary to detect issues even before users reach them organically.

Since we integrated Datadog, we have had increased confidence in the quality of our service, and we had an easier time increasing our delivery velocity. 

We have seen time after time that the monitors we have carefully created based on all ingested data are detecting issues quickly and accurately. 

This means we allow ourselves to manually test things less frequently. We have also had an easier time investigating application errors and slowness using Datadog's APM and log explorer products which allow us to introspect any part of the system, in its execution context.

The most valuable features include:

• Integrated observability data ingestions: All data that Datadog collects is connected. This allows easily connected logs with failed requests, and slow database questions with services and requests.
• Broad integrations allow us to monitor our entire production environment in a single place, not just cloud resources. Since all parts stream metrics, logs, and events to Datadog, we can have unified dashboards and manage monitors and incidents all from the same page.
• A high level of configuration. We can configure and modify many parts, from how data is collected from our applications to how Datadog parses and visualizes it. This means that we always get the best experience, and we don't need to find ten different products that do small things well or settle on one product that does everything badly.

Since the Datadog platform has so many separate features, solving so many use cases, there are often inconsistencies in feature availability and interoperability between products. 

Older, more mature products tend to be complete (many features, customization, broad integrations, etc.), while newer products will often be at a "just above minimum viable product" phase for a long time, doing what's intended yet missing valuable customizations and integrations.

We've used the solution for 12 months.

The solution scales very well on technical aspects, being able to ingest large quantities of data from many services. However, the pricing often doesn't scale naturally, and effort has to be put in to keep ongoing costs at a reasonable amount.

Customer service and support are generally very high-quality. In most cases, they reply very quickly and offer well-researched and relevant responses. This is contrasted with many vendors who take a long time to reply and send links to documentation instead of understanding the problem.

However, we had cases where support took several weeks to reply to a complicated request and sometimes eventually responded that the issue cannot be resolved. These are rare edge-case occurrences.

Positive

A large part of the initial setup was straightforward. We were able to collect about 80% of the relevant and 90% of the meaningful insights from just a couple of hours of connecting the AWS integration and the Datadog APM agent. 

Getting it to 100% and configuring and customizing things to our unique situation, took about two weeks. Datadog's documentation and support team were extremely helpful during both phases.

We handled the setup in-house.

From the number of outages stopped or shortened (which lead to lost revenue from non-renewals) and the number of hours saved on investigations (which correlates to engineering salaries), I estimate that the ROI of the implementation time and monthly charges to be between 10x and 20x.

We use the solution as a SaaS deployment.

We use the solution primarily for distributed tracing, service insight and observability, metrics, and monitoring. We create custom metrics from outbound service calls to trace the availability of back-office systems. 

We use the flame graph to get insights into our GraphQL implementation. It helps highlight how resolvers work. 

However, it's lacking in tracing which GraphQL queries are run, and we use custom spans for that.

Prior, the team only had Instana, and few people used it. The main barriers to entry were the access (since it was not integrated into our SSO) and the user experience, which made it hard to follow. We had an on-prem version, and it wasn't the snappiest. The APM has made observability and tracing more accessible to developers.

Flame graphs are pretty useful for understanding how GraphQL resolves our federated queries when it comes to identifying slow points in our requests. In our microservice environment with 170 services. There are complex transactions over the course of a single user request since we essentially operate as a middle layer with 90 back office systems we integrate to.

I often have issues with the UI in my browser. I tend to have a lot of tabs open, yet have issues with it not responding or not showing data. A couple of times, pasting the URL into an incognito window shows the data that's there.

I've used the solution for two years. 

The initial setup was complex and required a bit of tweaking to get everything configured correctly and into our pipelines.

We are using a mixture of on-prem and cloud solutions to bridge the gap with healthcare entities in the service of providing patients with the medication they need to live healthy lives.

Since we're a heavily regulated company, a lot of our solutions grew from on-premises monoliths. However, as we scaled out, it became harder and harder to move forward with that architecture. Today, we're investing heavily in transforming our systems from monoliths into distributed systems.

With this change in mind, the ability for us to connect the dots using Datadog has been invaluable.

We have an API that serves as a critical aspect of our system for generating new requests for us to process in service of a patient. This service has many tentacles, and it was always hard to track down how issues from this API are affecting things downstream. Since we've added more instrumentation in this API, Datadog has changed our status from a reactive posture to a proactive one.

It has also served as a prime example to other applications on what the benefit of a well-instrumented system is for that application and other applications around it. Due to this, more and more people are using Datadog.

We like the distributed tracing and flame graphs for debugging. This has been invaluable for us during periods of high traffic or red alert conditions. It has also informed our developers on how our various systems are interconnected and the downstream effects of the problems we might encounter for certain services.

We're still working on getting widespread adoption of these products. Still, we're already seeing a shift in the developer's perspective from application-specific and starting to look at things from a more holistic systems perspective.

While this is not part of the question, this is relevant: Now that I've learned more about RUM, this will be something that we will heavily leverage moving forward to give us a whole complete view of our system from the front and back end perspective.

Once Datadog has gained wide adoption, it can often be overwhelming to both know and understand where to go to find answers to questions. Currently, we use a combination of documentation and COPs to ensure that folks know how to leverage what we have in Datadog properly.

While the guides for Datadog go a long way, a way to customize the user experience from "advanced" to "novice" mode would go a long way.

I've been using the solution for two years.

It has never failed us and therefore I consider it to be very stable.

It's magic. For the most part, we just installed the product and a lot of it just worked out of the box.

Technical support is excellent.

Positive

We have used Splunk, Sentry, and a suite of hand-made solutions. We switched since the Datadog solution was both comprehensive and cohesive. It was also easier to onboard people since the solution was well-documented and standardized.

For the most part, it was really painless to set up.

We implemented the solution in-house.

We're still early on in our transformation process. That said, we are gaining a lot of steam in terms of adoption. Both the engineering team and the product team are seeing tremendous value from this solution.

Adding more tooltips and links to documentation or how-tos within the application would really go a long way for those trying to get their feet wet with Datadog.

Datadog is a SaaS solution we tried for URL and synthetic monitoring. You record a transaction going into a website and replay that transaction from various locations. Datadog is mainly used by the admin, but three or four other guys had access to the reports and notifications, so it's five altogether.  

We probably tried no more than 8 percent of what Datadog can do. There are so many other bits and modules. I've only gone into about half of what APM can do in the Datadog stack.

We could detect outages on particular websites or problems in specific locations. If I had paid for the full solution, I'm sure I could get a lot of value out of Datadog.

I like the amount of tooling and the number of solutions they sold with their monitoring. Datadog was highly intuitive to use. 

Datadog needs more local Asia-Pacific support, and if they don't have a SaaS solution in Asia-Pacific, they should offer an on-prem version. I'm told that's not possible. 

I have used Datadog for about two or three years.

I was only using Datadog to monitor on a small scale. 

I'd rate Datadog support four out of 10. It was primarily an issue with support in the Asia-Pacific region. I sent them several emails, and they responded around three weeks later. 

They said it went around the houses. Nobody knew who to respond to. That's not good enough. They should have at least told me they'd received the email. I used to work in support.

Neutral

We were just trying Datadog, and we've switched temporarily to Site24x7. We're looking for one of the bigger ones. They've all given us proposals, whereas Datadog hasn't come forward with a proposal for what they could do.

I used Datadog because I already had a relationship with them at a previous company. However, that guy's moved on now, and I wanted to see how good they were. 

Setting up Datadog is pretty straightforward. I have a lot of experience doing that sort of thing. It took maybe a day and a half to deploy because I was picking externally facing websites.

I deployed it by myself. One person is enough for the small system we had. However, if we were moving forward, I'd recommend at least two or three people to manage it. 

Datadog would've cost around $850 a month based on the loads we were doing, and you could estimate roughly what you would be paying monthly. I liked their pricing model. It was flexible, so you only paid for what you used. I rate Datadog pricing eight out of 10. 

We looked at several URL and APM monitoring solutions like Site24x7 and Pingdom. They weren't big players like Dynatrace or any of the those that had already provided us a request for information. 

Even with our negative experiences, I'd still give Datadog an eight out of 10. Datadog is a complete solution with easy-to-use templates and excellent scalability. People should know exactly what they're going to configure before they try it out. The trial is brief. Don't start a trial until you know exactly what you're going to do. 

You must be certain that you can meet any internal security requirements. If you're in the Asia-Pacific region, you might not be able to run something that's running abroad.