https://seekingalpha.com/pr/17851838-datadog-achieves-aws-la…
Datadog, announced that it has achieved the AWS Lambda Ready designation, part of the Amazon Web Services (AWS) Service Ready Program. This designation validates that Datadog’s cloud monitoring platform has demonstrated deep integration with AWS Lambda.
I admit that I don’t understand what it means, except that it must ve good news for Datadog.
Saul
Hey Saul - one of your resident software engineers here - happy to try my best to explain what this means in layman terms.
Lambda is the AWS serverless solution. In 2020, pretty much no one wants to manage servers, but a lot of companies still find themselves doing this (including mine). Lambda enables you the luxury of not having to manage servers. Lambda and competing serverless solutions are basically considered part of the “future of infrastructure”. I think of lambda metaphorically, as a service that offers you “virtual servers”. You put your app on lambda, and poof, you no longer need to worry about doing java updates or operating system updates and all the other crap that comes along with managing servers. AWS Lambda handles that for you.
So basically, this AWS designation is saying that AWS lambda and datadog play nicely together. There should be few if any issues if you try to use datadog to monitor an app deployed on lambda. I am paraphrasing, this designation is sort of like Amazon coming out and saying, “Hey software engineers. Your dreams of going serverless will not be interrupted by your dream of using Datadog’s terrific monitoring solutions.”
I say this, as a software engineer who works on an app that uses Datadog, and we would like to migrate to Lambda at some point in the near future. Something I’ve said elsewhere that’s worth repeating for anecdotal purposes - we are big fans of DDOG at my company.
Long DDOG
One thing I still don’t really understand with DataDog is exactly how they fit into this model.
In a previous post (I wish I could find it so I could give a shout out to whoever wrote it), someone skillfully explained how Datadog works using the Motley Fool as a hypothetical example. The example was the fool’s web service being overwhelmed, and using DataDog to trace the bottleneck to a particularly active message board. Once the point of congestion was identified, the issue could be fixed.
I’m struggling to understand how this works if the company is not running its own infrastructure, but rather that infrastructure resides in the public cloud. I imagine that each company’s use of the public cloud resources is segmented off in a way that allows Datadog to monitor company specific resources within the larger public cloud. Is this correct?
What I’m struggling from an investment standpoint is understanding how it isn’t inevitable that public cloud providers will displace third party monitoring solutions with their own monitoring services. Will Microsoft eventually be releasing a monitoring system for Azure that makes Datadog redundant? Could Datadog be like fitbit - selling like hotcakes now, but eventually replaced by vastly superior offerings from the tech giants? Or is the multi-cloud and mixed cloud/on prem market creating the need for a platform that works across different public clouds? These are the thoughts that my obsessive “worry mind” has grasped onto lately. I started thinking about it after Muji’s Oktane post, when he mentioned one executive telling another that they just built a new data center and the other executive responding, “sorry to hear that.” I started wondering, “how does DataDog fit into a world where company’s aren’t running their own infrastructure?”
Its too bad I don’t understand the technology itself well enough to figure it out. I just have the company’s results and endorsements from experts to rely on. For me, this is enough to remain long Datadog. I just wish I understood it a little better.
BobbyBe,
That’s a good question. Cloud Titans already have products that compete with DataDog. Azure Application Insights and Amazon CloudWatch for example. The question becomes, do you want multiple tools that do the same thing, but each analyzing something different? I think the whole motto for DataDog is to break down silos and put as much information into one place as possible.
I imagine that each company’s use of the public cloud resources is segmented off in a way that allows Datadog to monitor company specific resources within the larger public cloud. Is this correct?
That is correct.
What I’m struggling from an investment standpoint is understanding how it isn’t inevitable that public cloud providers will displace third party monitoring solutions with their own monitoring services.
This is an excellent question and one that I have posed to people who have more knowledge of why my company chose Datadog as opposed to other solutions. I have not done much research on this and definitely I am interested to hear other’s opinions on this.
But what I gathered when I asked other software engineers at my company is two things. To put things simply: Firstly, datadog does it better. Have I seen/used the AWS competing monitoring product, thereby giving me some sort of special opinion on this matter? No. It’s just what I heard.
Secondly, companies don’t like the idea of relying on AWS too heavily for things. AWS already has a reputation of being very sticky. Once your contract with AWS expires, they may jack up the price, and you may have taken your company too far down the AWS rabbit hole to choose a competitor. Using a 3rd party, such as DDOG, gives them more peace of mind in that regard. It is one less thing they need to rely on AWS for.
Secondly, companies don’t like the idea of relying on AWS too heavily for things. AWS already has a reputation of being very sticky. Once your contract with AWS expires, they may jack up the price, and you may have taken your company too far down the AWS rabbit hole to choose a competitor. Using a 3rd party, such as DDOG, gives them more peace of mind in that regard. It is one less thing they need to rely on AWS for.
If I’m not mistaken, DataDog lets you monitor your whole system across multiple server providers in one monitor app or client, something that server providers would not do. I’m saying this from memory and don’t have links. Correct me if I’m wrong.
Denny Schlesinger
Hey Saul - one of your resident software engineers here - happy to try my best to explain what this means in layman terms.
Thanks so much Vinegar. Happy to have you on the board!
Saul
If I’m not mistaken, DataDog lets you monitor your whole system across multiple server providers in one monitor app or client, something that server providers would not do. I’m saying this from memory and don’t have links. Correct me if I’m wrong.
Yeah, that’s right. So, I don’t want to come off as though I am proclaiming myself as some sort of infrastructure subject matter expert - but I do dabble with these things some time. You guys are forcing this software engineer to think more like an IT manager, which is a good thing for me 
Any how, you have encouraged me to revisit the difference between the AWS competing solution and Datadog. I believe the AWS competing solution is called AWS CloudWatch. This website here talks about some of the differences: https://stackshare.io/stackups/amazon-cloudwatch-vs-datadog
In summary, from the website: “Monitor aws resources” is the primary reason why developers consider Amazon CloudWatch over the competitors, whereas “Monitoring for many apps (databases, web servers, etc)” was stated as the key factor in picking Datadog.
Vinegar101 paraphrasing: The above website quote fits in with the “Datadog enables you to view everything on one pane of glass” mantra. Alternatively, AWS CloudWatch let’s you watch only aws resources (fyi, this is just my best guess - if anyone knows better, please correct me). So if you just use aws for all your infrastructure needs, you can use CloudWatch. If you have stuff outside of aws you need to manage, I think you can’t use AWS for that. That’s where Datadog comes in.
Still though, if you are a company that has fully migrated to aws infrastructure, I think Datadog is still considered the better monitoring solution. This article is old, but it is from Datadog itself, talking about performance metric differences with AWS CloudWatch: https://www.datadoghq.com/blog/why-do-aws-cloudwatch-and-dat…
Even when that article was written (2014), it looks like the datadog solution was better lol.
Here are some quick insights…
I’ve written about DDOG in some earlier posts about how DDOG needs to keep moving fast and this is indeed very good news
Here are a couple of simple definitions ( you really don’t need to know more unless you want to do a deep technical dive).
“AWS Lambda is an event-driven, serverless computing platform provided by Amazon as a part of Amazon Web Services.”
“Serverless eliminates the need to provision and manage infrastructure components (e.g., servers, databases, queues, and even containers), allowing teams to focus on code while minimizing their operational overhead.”
Here’s how this is good news for DDOG and us ( investors in DDOG)
Datadog’s cloud monitoring platform has demonstrated deep integration with AWS Lambda.
Half of AWS users have adopted AWS Lambda.
DDOG has started with serverless with AWS. And should be a matter of time before they integrate with Google Cloud, Microsoft Azure.
At this point it appears that bigger companies which need larger infrastructure environments are adopting Lambda more rapidly.
For more insights read: https://www.datadoghq.com/state-of-serverless/
Takeaway: For now, DDOG has again proven that it’s doing it’s part to innovate and stay ahead…
Cheers!
ron
I haven’t used DDOG in production, but CloudWatch is pretty much on by default. It’s more aimed at the server-level of analysis, for example, what is that database doing, is my server still performing?.
Datadog, NewRelic and the others are much more comprehensive, and operate at an end-to-end application level. So you can trace what a user does all the way to what happens in a database, across many servers, and back again.
For some reason that I don’t quite understand, AWS has some products that are half-baked. CloudWatch is one, QuickSight (a Tableau ‘competitor’) is another.
AWS has X-ray, which I haven’t used, but thats actually a ‘competitor’ to Datadog. I believe that Datadog etc use CloudWatch and X-ray metrics to supply their data so it doesn’t look like AWS are particularly interested in competing in that space.
Datadog need to integrate with X-ray because of ‘serverless’. Perhaps a better metaphor for ‘serverless’ is ‘ephemeral servers’ or “The code runs, somewhere, and I don’t know where it’s running and I don’t care”, or “what’s a server?” 
but you still need to monitor end-to-end.
I don’t think this news is particularly ground-shaking for Datadog. However… I did search for “Elastic/New Relic/Splunk APM lambda” to check the competition, and the Elastic search returns garbage. It’s really not clear that Elastic supports Lambda at all which would be significant. I’ve contacted Elastic for their response.
cheers
Greg
Hi ronjrnb, I’ve been trying to email you off board to ask you something OT, but my emails must be going into your spam folder. Please contact me so I have a way to reach you (if my contacting you is okay with you).
Thanks,
Saul
OT
Hi Saul,
I just checked and found your email was in spam ( as well as some from other members in this board). Sorry for that…I don’t monitor this email account often. It’s tied to my Motley Fool subscriptions.
Please do send me your question/s. I’ll be happy if I can answer
Cheers!
ron
I think it’s important to understand that AWS Lamda is not a service, and not something for which users pay for on its own. It’s a platform on which anyone can build products. As noted earlier, the platform supports serverless APIs (Application Programming Interfaces), which is actually a misnomer because there definitely are servers involved. Think of Lambda as Amazon’s serverless platform which anyone can supply applications under. It’s really not hard to understand once you cut through the techno-blabber.
Imagine running a program on your phone. You start-up an app and give it some inputs and it gives you some results/outputs. You give it some more inputs and it gives you some more output. Then maybe you kill the app or restart the phone, or send the app to the background where the OS may leave it running or kill it off for you to save battery life.
It’s the same on the cloud. In typical server computing, you start-up the application and feed it data and get results. The application stays running on the server so that you can feed it more data and get results back. Often the second request you make of the app is based on the first request’s results. But, the application stays running on the server until you manually stop it, and AWS/Azure/Google Cloud charge you as long as that server is running and consuming memory, CPU, etc.
With serverless, you don’t start-up an application first you just make an API call and get back results. You don’t know it, but behind the scenes, there’s a server instance that gets started up, run, obtains results, and then terminated. The good news is that you don’t get charged for idle time since there is no application that’s continually running. The other good news, for programmers, is that you’re just calling an API. You’re not worrying about starting up the app, keeping it running, knowing when to terminate it, etc. It’s easy to use.
The problem is that there’s overhead with making these API calls since each starts up an applicatoin, and each API call is independent of each other since each starts up a new server process behind the scenes. For calls which ingest data (like from millions of IoT devices sending data into the cloud), this is great. For running “what-ifs” on a large data set, it can be a nightmare.
It’s really not clear that Elastic supports Lambda at all which would be significant.
Elastic is not listed on the AWS Lambda support pages that I saw. If you’re using Elastic as a tool to run multiple analysis queries on large data sets, you wouldn’t want to use Lambda because each query would have to startup an ElasticSearch application on your large data set, process the query, then shut-down. It would be more efficient to start the application once, then make several calls, then shut-down.
But again, it depends on what you’re doing. Small, lightweight, and non-compute-intensive functions are better/more easily done via serverless APIs. Large, compute-heavy applications are not. I would not expect the see Elastic on Lambda unless it was to support data ingestion.
Hey Smorg,
I was referring to Elastic APM, rather than ElasticSearch itself.
All the other APM vendors on my google search went straight to “How to do APM with Lambda and DataDog/Splunk/NewRelic”. The Elastic equivalent showed rubbish results.
That is, if I was looking to implement APM and I use AWS Lambda (and presumably GCP and Azure equivalents) that would be a big red flag for Elastic, and a big green flag for Datadog, NewRelic, Splunk etc.
cheers
Greg
I was referring to Elastic APM, rather than ElasticSearch itself.
Of course - sorry I missed that. The only thing I saw there was some github questions about when/if Elastic would add that. So, apparently not yet.
The other thing I missed in my rush to post on serverless was that it’s not that DataDog’s own APIs use Lambda, but that DataDog can now be used to monitor your Lambda usage without depending on parsing CloudWatch logs.
DataDog published a white paper on it: https://www.datadoghq.com/state-of-serverless/ , which references some of the cost relationships I outlined in my previous post:
The median Lambda function runs for about 800 milliseconds, averaged across all its invocations, but the tail of the latency distribution is long. One quarter of Lambda functions have an average execution time of more than 3 seconds, and 12 percent take 10 seconds or more. The long duration of some Lambda functions is notable because serverless latency impacts not only app performance but cloud costs. Lambda pricing is based on “GB-seconds” of compute time: the memory allocated to your function (detailed in the following fact), multiplied by the duration of its invocations.
So, DataDog can help you understand how your applications using Lambda may be contributing to too-high charges from AWS and help you isolate just which calls need to be changed.
Yes, when we used Lambda, it ended up costing us a ton (unexpectedly) because we were using it to shift data from one place to another which meant we invoked the lambda function a lot. Our first bill was eye-opening!
It’s hard to tell how long a lambda function will execute for, or really have any idea in advance, particularly if you don’t know exactly how many times it will be called, for example, because of unpredictable traffic. It’s also difficult to look at CloudWatch and see.
re: DDOG, I’m not sure if this was posted previously, but it seems that Datadog have released their security monitoring app (https://www.datadoghq.com/blog/announcing-security-monitorin…). That was (AFAIK) announced last quarter which seems a) very fast, and b) potentially a needle mover for DDOG.
cheers
Greg
I work on infrastructure and also have been working on Cloud. I have decent understanding of cloudwatch. As vinegar said its one of the offering from AWS and we use it a bit. However in terms of overall solution, the breath of scope, its nowhere near datadog. And mind you I dont personally have hands on experience using datadog but I just know application monitoring is something thats specific to APM vendors to give end-to-end insight. Cloudwatch can provide infrastructure monitoring and to some extend netowrk but even there you will need to have some amount of understanding how to utilize it. Still there are enough gaps that you tie in with other services provided. Its very sophisticated and you need fair degree of knowledge to use it effectively. Most software development companies will have engineers who can make use of it. However it by no means is as straight forward as DDOG judging by comments made by Bert and management. They do not have service revenue because it apparently is that so easy.
I have also done some setups on Azure and Oracle cloud. The also provide there own tools similar to AWS cloud watch but the ecosystem for Oracle needs a lot. In other words they are not there yet. Azure is comparable but they also look ease of use and frankly there ecosystem needs a lot of catchup too.
Bottomline is, most companies are not just software development companies. There do have teams of IT people working. For monitoring most companies will most likely go with some APM vendor Datadog or app dynamics etc. If one use Datadog for APM it may become easy set for Infrastructure and Networking. In my field its paramount to try and standardize as much as possible around one platform… It saves cost in terms of training and eases troubleshooting and people can communicate in one language.
Hope that helps.
Ruhaan
Long DDOG
How easy would it be to replace Datadog with something else for monitoring applications?
In a previous post (I wish I could find it so I could give a shout out to whoever wrote it), someone skillfully explained how Datadog works using the Motley Fool as a hypothetical example.
I think this is what you want:
https://discussion.fool.com/i-hate-the-business-it39s-in-it-is-h…
This is a link to Amazon’s page advertising their Lambda service.
https://aws.amazon.com/lambda/
This is outside my expertise, but it looks like a server rental by the millisecond, on-demand, rather than a “batch” rental where you upload a huge block of stuff that will keep the server occupied overnight. You pay for a few milliseconds at a time on demand, when your cellphone or desktop app needs it. Something you do changes the state of some variable in your phone app, and it tells Amazon to run something for a tiny time and returns a result. The server functionality “exists” for only the amount of time you are using it, then it is gone. Like a data hotel instead of a data apartment.
This link seems to confirm that:
https://stackshare.io/stackups/aws-batch-vs-aws-lambda
How easy would it be to replace Datadog with something else for monitoring applications?
Wrong question! How easy is it to move customers from Datadog to “something else.” People who keep asking that question don’t understand the workings of complex systems, of path dependence, and the economic law of increasing returns. A lot of investors are still thinking in terms of commodities and that frame if reference does not apply to most high tech investing.
How easy would it be to replace Google with another search engine? Why hasn’t it happened?
Denny Schlesinger
PS: I realize that my reply is rather sharp but it is in no way personal. Apologies in advance if it was taken as such. I believe this is an important lesson for tech investors and I urge you all to bone up on complexity. I highly recommend Complexity: The Emerging Science at the Edge of Order and Chaos by M. Mitchell Waldrop as a great starting point. It’s a book for the layman and very entertaining.