Sawmills Raises $10M to Revolutionize Observability Data Management with AI

Sawmills raises $10M to cut down observability data costs with AI - SiliconANGLE

Sawmills raises $10M to cut down observability data costs with AI Telemetry data management startup Sawmills.AI Ltd. says it's hoping to help enterprises shave millions of dollars off their observability software bills after raising a fair few million itself. Today it announced it's closing on a "highly oversubscribed" $10 million seed funding round led by Team8, with participation from Mayfield and Alumni Ventures. In addition, Sawmills also announced the general availability of its flagship telemetry data explorer tool today, which is built on the open-source OpenTelemetry Collector project and enhanced with sophisticated artificial intelligence capabilities. The startup says its platform provides companies with a way to automate the management of data that flows from software applications and services to their observability tools, so it can reduce the often staggering cost of using those products. Most enterprises understand the need for observability, as it is critical for them to maintain reliable applications without unexpected downtime. But they're a lot less keen on the exponentially increasing costs associated with using such tools. According to Sawmills, the average company now spends almost $2 million annually on observability, and many report experiencing unexpected increases in those costs each month. In one famous case in 2022, the cryptocurrency exchange Coinbase Inc. was slapped with an eye-watering $65 million bill by Datadog Inc. for just one year of service. Sawmills Chief Executive Ronit Belson said the reason for these excessive costs is that observability tools process an awful lot of unnecessary data that doesn't provide useful insights. The standard business model for observability firms is to charge customers based on the data their tools consume. "In our conversations with VPs of engineering at leading companies, they consistently tell us that up to 90% of their observability data is useless, yet they're still paying to collect, process and store all of it," he said. Another challenge is that mistakes can lead to unexpected spikes in these costs. Belson related a story from one customer that explained how a single mistake from a developer resulted in additional fees of more than $250,000 in just one day. "Engineering teams need intelligent telemetry data management that not only improves data quality but also prevents costly mistakes before they happen," Belson said. "Sawmills automatically identifies optimization opportunities and implements guardrails to protect against unexpected cost spikes while ensuring you capture the data that matters." Another problem Sawmills hopes to solve pertains to the quality of observability data, which is often rendered useless because of missing data points, inconsistent formats and duplication. This low-quality data drives up costs and makes root cause analysis far more difficult, Belson pointed out. "Observability data has become the second-largest expense after cloud costs for most companies," he added. Sawmills says it solves these problems by implementing more intelligent data management. Its platform leverages OpenTelemetry Collector to gather telemetry data, and then it uses its proprietary AI algorithms to analyze it, automatically identifying opportunities to reduce spending and improve the quality of that data. Belson explains that the company's intelligent engine can process logs, metrics and traces in real time in order to detect problems such as missing data points, duplicate data and inconsistent data formats. It automatically fixes and enriches this information before sending it to the observability system. In addition, it implements "smart sampling" policies that give companies full control over their observability data streams, safeguarding against availability issues. After processing all of this data, Sawmills' AI algorithms will make various recommendations that can be used to create automated policies to prevent unexpected cost spikes. By carefully managing telemetry data in this way and sending it to more cost-effective storage resources, Sawmills can reduce the amount of information that needs to be processed by observability tools and improve its quality. The result is lower costs and more useful insights. Andy Thurai, principal analyst and vice president at Constellation Research Inc. said Sawmills' observations about the rising costs of observability are very real, due to a combination of the shift to a consumption-based model and the fact that DevOps and site reliability engineers have a tendency to carelessly infuse data into these platforms. "Most enterprises have been getting a sticker shock in terms of overage bills," Thurai said. "A lot of companies are moving towards sanitary SRE and observability practices, where they define best practices ahead of time to avoid these shocks, but cost overage remains a very real thing for many firms that don't employ these methods." The strength of Sawmills' technology derives from its foundation, the OpenTelemetry Collector project, which makes it possible to process data outside of the observability platform, Thurai said. "It allows companies to decide what to send to the observability platform and what can be sent to either cold storage or dashboards," he pointed out. According to Thurai, the main differentiator for Sawmills is it makes OpenTelemetry easier to use. "OpenTelemetry is not easy to implement, requiring a certain expertise that is hard to come by for most companies," he said. Sawmills faces a lot of competition in its efforts to combat rising observability costs, with rivals including Cribl.io Inc. and other observability data pipeline startups like SigNoz Inc., Kloudfuse Inc. and Edge Delta Inc. However, its biggest competitors are the observability companies themselves, Thurai said, with the likes of Datadog, New Relic Inc., Elastic N.V. and Cisco Systems Inc.'s Splunk all embracing OpenTelemetry Collector data. Still, Sawmills claims it has been making progress in this competitive market. It cites the example of the Indian online travel agency Via.com Ltd., which has been using Sawmill's tools for a while and said it has benefited from "streamlined costs, improve observability resource allocation and enhance data governance." Team8 Managing Partner Liran Grinberg stressed that telemetry data management is fast emerging as a critical new category for enterprise cloud infrastructure. "The Sawmills team has a deep understanding of the problem and a comprehensive vision that perfectly positions them to own this new category," he said. "This isn't just about cost reduction. We believe Samills' approach to intelligent telemetry data management will massively improve observability and become essential infrastructure for modern enterprises."

Sawmills emerges from stealth to trim enterprise observability costs and provide telemetry data sovereignty

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Data observability -- the practice of using software tools to provide a window into how an organization's entire software suite, especially the most business-critical applications, is functioning -- actually took root in the early computer era of the late 1950s, but it has renewed prominence in the generative AI era. While observability platform vendors such as Splunk and Datadog have built multibillion-dollar businesses providing tools that help their enterprise customers organize all telemetry data -- essentially, data that indicates the statuses of different processes, whether a program is functioning normally or not, and why -- the truth is that many organizations are now finding themselves drowning in too much data and associated costs in the AI era. Consider that Charity Majors, co-founder and CTO of Honeycomb, suggests that organizations should allocate 20 to 30% of their infrastructure budget to observability. And yet, a 2023 survey highlighted that 98% of companies have experienced unexpected increases in observability expenses, with 51% encountering overages on a monthly basis. Now a new San Francisco startup Sawmills AI is here to sit between observability platforms such as Datadog and Splunk and their customers, using large language models (LLMs) and other clever new proprietary machine learning (ML) models to help consolidate, summarize, trim and ultimately reduce the amount of data sent from the customer to the vendor, while empowering the customer to retain all the original data and do with it what they will. "A lot of companies have more than one observability solution," Ronit Belson, co-founder and CEO of Sawmills AI, explained in a video call interview with VentureBeat. "We strongly believe that telemetry data should be owned by the customer, not the observability vendors." Today, Sawmills AI emerged from stealth with $10 million in seed funding in an oversubscribed round led by the venture capital firm Team8 with participation from Mayfield and Alumni Ventures. Co-founded by Belson, CTO Amir Jakoby and CPO Erez Rusovsky, the company is tackling the increasing costs of observability while improving data quality and reliability. The company's smart telemetry management platform enables businesses to fully harness the potential of their telemetry data at petabyte scale, but at a fraction of the cost. "We empower engineering teams to manage, optimize, and act on their telemetry data," the company states on its website. "By addressing inefficiencies, mitigating costly data spikes and improving data quality, we enable businesses to reduce expenses and enhance the effectiveness of their observability systems." How Sawmills decided to chop observability and telemetry costs Sawmills' founders initially set out to solve a different problem, but industry conversations quickly shifted their focus. "We spoke with over 100 VPs of engineering and heads of DevOps, and their biggest problem wasn't what we expected -- it was the cost of observability solutions," Belson told VentureBeat. This wasn't just an isolated concern -- companies across industries reported paying for vast amounts of observability data that provided little actual value. "Companies are spending millions on observability, but when we asked how much of that data they actually need, the answer shocked us -- only 10 to 30%," Belson added. "That means 70 to 90% of the data sent is essentially junk." Rusovsky highlighted the challenge organizations face in managing this data explosion. "Teams are struggling because every developer is writing their own telemetry data, and there's no easy way to manage it centrally," he said. Without a system in place to filter or optimize logs, metrics and traces, data volumes continue to grow unchecked, driving up observability costs while making troubleshooting more difficult. "We are not an observability solution," said Belson. "Customers love Datadog for what it provides, but they also hate how much they pay for it." Sawmills' AI-driven solution Sawmills has developed a smart telemetry data management platform that allows companies to filter, route and optimize their observability data before it reaches their observability tools, like Datadog, Splunk or New Relic. The platform acts as a middleware layer, analyzing logs, metrics and traces in real-time using AI and ML. Key features of the flatform By using all these tools, Sawmills claims to provide significant data transmission volume and associated cost savings for its enterprise customers. "If you're sending millions of lines of logs that could be converted into a single metric, that alone can reduce the data volume by 10X or even 100X in some cases," said Rusovsky. To help summarize and consolidate customer data before it is sent to the customer's observability platform, Sawmills' platform leverages leading LLMs from top providers available as open-source models and on proprietary cloud vendor marketplaces' favored by their customers. "We're using a combination [of LLMs and ML models]," said Jakoby. "In some cases, we can use OpenAI or AWS Bedrock, but some of our customers don't allow their data to be sent externally, so we use their own cloud-based LLMs instead." Offering centralized telemetry management Rusovsky emphasized the importance of data control, noting that companies today lack centralized telemetry management: "Observability is mission-critical for all organizations, but it's also the second-largest engineering expense after cloud costs," he said. Many companies send terabytes of telemetry data daily, much of which is unnecessary. "Everybody is afraid to remove logs," Rusovsky added. "What if I need it later? That fear leads to companies sending terabytes of data a day -- much of it unnecessary." This excessive data not only drives up costs but complicates troubleshooting and monitoring. "The problem isn't just cost -- when telemetry data is out of control, it becomes harder to use," said Belson. "Too much data makes root cause analysis more difficult, and building reports takes longer." Sawmills' platform is built on OpenTelemetry Collector, an open-source industry standard for telemetry data collection. However, the company enhances it with its own proprietary layers atop it, providing AI-powered capabilities that detect anomalies, enrich data for better insights and implement smart sampling policies. Beyond cost savings, Jakoby highlighted another financial advantage. "Many customers don't realize they pay for ingesting data into Datadog, even if they later drop it," he said. "Filtering before sending can lead to massive cost savings." Early success Early adopters of Sawmills are already seeing the benefits. Edi Buslovich, VP of engineering at Via, said working with Sawmills has helped optimize telemetry data, reduce costs and improve governance. Belson emphasized that Sawmills is not competing with observability providers, but rather helping companies maximize the value of their existing tools. She noted that while engineers appreciate platforms like Datadog, they often dislike the associated costs. By allowing enterprises to control their telemetry data, Sawmills enables more efficient spending and improved system reliability. Liran Grinberg, managing partner at Team8, sees telemetry data management as an emerging category within enterprise infrastructure. He believes Sawmills' approach goes beyond cost-cutting, positioning the company as a key enabler of better observability practices. "We don't just cut costs -- we improve data quality," Rusovsky explained. "Instead of blindly sending everything to vendors, companies can finally take control of their own telemetry data." What's next for Sawmills AI? Sawmills is targeting mid-to-large enterprises with 500 to 5,000 employees, particularly those that are cloud-heavy and heavily invested in observability. The company has already secured dozens of paying customers and plans to expand its market reach following its public launch. In addition, the co-founders emphasized to VentureBeat that their customers will benefit more from ongoing usage of the platform as it better learns each customer's unique data and tool mix and needs. "When logs flow through our system, we can actually start training based on the logs and the recommendations we generate, improving our model over time," said Jakoby. "So as you use the system, the recommendations become more fine-tuned to your specific needs." Reflecting on the startup's early momentum, Belson shared that Sawmills' funding round was quickly oversubscribed, with multiple investors eager to participate. "When we started fundraising, we had multiple offers in two weeks," she said. "We planned to raise seven million and ended up raising ten. Now, it's time to take this to market." As software architectures continue to grow in complexity, the need for smarter telemetry data management is becoming increasingly critical. With its AI-powered approach, Sawmills aims to establish itself as the go-to solution for enterprises looking to optimize observability costs and data quality -- all while maintaining system reliability and vendor flexibility.