Jupiter closes a 20% to 40% infrastructure spend visibility gap with DoiT

Infrastructure is one of the largest line items at Jupiter, and after the acquisitions nobody could say precisely how large. None of the inherited workloads had moved onto the primary cloud, so the total shifted depending on which bills you counted.

The wider gap came with the market. Solana takes a high volume of writes, which puts unusual pressure on the read side, and a class of specialist companies has grown up to serve it: Remote Procedure Call (RPC) providers, which read the chain and stream changes out of it on Jupiter's behalf. They are one of Jupiter's largest infrastructure costs, and they are young businesses whose billing works accordingly.

"They send us an invoice, we pay it, but it is a real challenge to catalog that somewhere. It is difficult to understand what is really costing us." shared Kyle.

By Kyle's estimate, 20% to 40% of Jupiter's infrastructure spend sat outside any system. Satyam Gupta, Senior Customer Success Manager at DoiT, puts it plainly: “if third-party invoices are tracked by hand, they are basically not tracking at all." A team can only forecast against the spend it can see.

DataHub is the part of Cloud Intelligence built for this gap. It takes cost data from outside the cloud providers and puts it in the same place as the cloud bill. Data arrives by API or CSV upload, carrying dimensions and metrics the customer defines, so an invoice from a vendor too small to justify a native integration becomes a source the platform treats like any other: reportable, and available to budgets, alerts and forecasts.

The capability was there but what Jupiter got was someone to build with.

Satyam Gupta, along with the DoiT Forward Deployed Engineers (FDE), started from how the invoices actually arrive, not from the API reference. He converted one real Jupiter PDF into DataHub format, ingested it and built the reporting and alerts on top, so the team could see the finished thing before committing. DoiT’s forward deployed engineers stepped in where the work went beyond configuration, while Jupiter tested the solution and defined how the data should be read, and Satyam’s team automated the flow around that feedback.

"On integrating DataHub, the DoiT team was very hands-on to the point where DoiT’s Team put together the first ingestion script and pipeline for us." shared Kyle.

What they handed back fits the finance team's habits. Invoices land in a Google Drive folder, the pipeline parses them and loads them into Cloud Intelligence beside cloud provider billing. Jupiter made minor tweaks and shipped it.

Jupiter could have tracked the missing vendors itself, but that leaves the cost data in two places, and a forecast can only run over one. 60% in DoiT and the rest elsewhere is not a position you can plan from. Kyle had run other FinOps platforms before, and that is where he draws the contrast.

"DoiT really has been able to fill the gap that I haven't seen with some of the previous FinOps platforms that I've used in the past." Kyle Espinola, Head of Platform Engineering, Jupiter

Jupiter feeds 26 outside vendors into DataHub, its RPC providers, ClickHouse and Cloudflare among them, beside its AWS and Google Cloud billing. The share Kyle once described as over 30% unreported is now in the same system as everything else.

"It's a massive unlock to get from 60% to 100% into one system, and DataHub is a great tool that is flexible and really lets us fill the gaps, but leverage all of the reporting." says Kyle.

The coverage number is not the whole return. Forecasting and historical usage analysis run across the full vendor catalog rather than the cloud slice of it, which is what Jupiter plans on. Budgets and alerts reach vendors that had no monitoring at all, so a bill that jumps gets flagged rather than discovered later. The cataloging Kyle described, remembering to record an invoice before the number is lost, is now the pipeline's job rather than a person's.

Coverage was the prerequisite. The next question is profit.

"We have very clear visibility on the revenue that a product brings in, but to understand your profit you need to know the cost. And a large piece of our cost for products is on the infrastructure spend." shared Kyle.

Allocations will map that spend to product lines across vendors, and Kubernetes Intelligence, activating this quarter, will break the shared Elastic Kubernetes Service (EKS) estate down by namespace. DoiT and AWS are working the consolidation in parallel, securing contract agreements and migration credits to bring the acquired workloads onto AWS.