LightStep Launches With [x]PM For Complex Application Performance Monitoring
Application monitoring startup LightStep has come out of stealth with its first product, [x]PM, designed to monitor complex distributed applications built with modern microservices architectures. LightStep has also announced funding to the tune of $29 million from a Series A round of $7.5 million led by Redpoint, a Series B round of $20 million, led by Sequoia, and seed funding from investors Cowboy Ventures and Harrison Metal, who participated in both rounds.
“There's been a major architectural shift in how people build software,” said co-founder and CEO Ben Sigelman. “It's hard to see what happens in any one transaction across systems using traditional approaches.”
“Our mission at LightStep is to cut through the scale and complexity of today's software to help organizations stay in control of their systems,” he said.
Sigelman built and ran global-scale monitoring technologies at Google, including Dapper, an always-on distributed tracing system that analyzes more than 2 billion transactions per second. Sigelman is also the co-creator of the OpenTracing standard, part of the Cloud Native Computing Foundation.
[x]PM has two major differences to alternative performance monitoring approaches: firstly, it captures all the telemetry data, rather than just a subset. That means absorbing a lot of data from modern applications with many micro-services processing thousands of transactions a second, which many other tools are unable to keep up with. Secondly, it's not aimed solely at troubleshooting problems, but also seeks to give advance warning of potential issues using anomaly detection.
[x]PM has collectors that live close to the application systems themselves, within a datacenter or in the VPC for cloud applications, whatever suits the application architecture. These collectors are connected to cloud-based Software-as-a-Service (SaaS) portal used to provide a central point for management and control.
Sigelman was keen to point out to me that the collectors are not merely store-and-forward satellite systems, but are in constant bi-directional communication with the SaaS. The communication is metadata, so the full firehose of application tracing data doesn't stream into the cloud, but there's some secret sauce in how the collectors and SaaS work together.
You can use
[x]PM to troubleshoot issues, yes, but you can also use it to help you predict when and where performance issues are likely to occur. This is handy for planning feature releases and capacity upgrades of systems where fully dynamic scaling of spot-priced cloud instances gets expensive. You can also get advance warning of potential failure and hopefully avoid it. It's a lot better to avoid a performance cliff than to figure out what caused an outage after the fact.
I like that [x]PM can integrate with existing systems in an enterprise. It has native OpenTracing support, and can also plug in to centralized logging systems (think Splunk and SumoLogic), as well as service mesh tools and load-balancers like nginx and haproxy. There's lots of existing investment in logging and monitoring systems, and they're not without value, they just don't cover the full range of use-cases. Tools like Splunk and SumoLogic, even security tools like ArcSight and Gigamon, all approach similar problems from their own perspective, and they all have value. Taking those existing investments and making them even more valuable is a smart approach to the enterprise market.
Something like [x]PM adds cumulative value. A tool that receives the full firehose of telemetry data modern apps spit out and can do smart things with it can happily co-exist with tools that take a sampling approach, or debugging tools you deploy when things really go wrong. It's not a case of either/or.
LightStep boasts some well known names as customers already, including Lyft, Twilio, Github, and DigitalOcean, who are all apparently using [x]PM in production.
This article first appeared in Forbes.com here.