Data you can trust by 9 a.m. — tested, observed, governed and kept in India.
DataOps applies software engineering discipline to data: pipelines as code, automated tests and CI/CD, orchestration, observability and data contracts — with DPDP-ready lineage and retention. Run by XcellHost engineers and our 24×7 NOC.
A finance dashboard shows yesterday's numbers. A churn model trains on a half-loaded table. A chatbot answers from a policy that was withdrawn last month. None of these are infrastructure outages — the servers are fine. The data is late, wrong, incomplete or not allowed to be there. That is data downtime.
DataOps is the discipline that prevents it. Pipelines are written as code and reviewed. Every change runs through automated tests. Orchestration runs jobs in the right order and retries the safe ones. Monitors watch freshness, volume and schema. Contracts stop producers from breaking consumers. Lineage shows where each column came from and where personal data flows — which is exactly what the DPDP Act will ask you to show.
Five kinds of data downtime. DataOps puts a test, a monitor or a rule in front of each.
Our working version of the ideas in the DataOps Manifesto (18 principles) and modern data engineering practice — written for teams that have to answer to auditors as well as analysts.
The DataOps loop has a development half and an operations half. Click a stage — or watch the run travel round.
—
From the first Git repository to the AI index — run as one service.
Repository structure, branching and review rules; isolated build environments on cloned or sampled data; automatic promotion from dev to staging to production when tests pass. Rollback is a revert, not a midnight restore.
Schema, null, uniqueness, range, referential and freshness tests on every model, with severity levels that warn or block.
DAGs with dependencies, retries, backfills, SLAs and alerts — on a managed orchestrator such as Airflow* or Dagster*.
Freshness, volume, schema, distribution and lineage monitored per table, with anomalies routed to the right owner.
Versioned schema and quality promises between producing and consuming teams, enforced in CI.
Glossary, owners, PII tags, access policies and column-level lineage in a catalogue your teams actually use.
Documents and tables flow through the same discipline before they reach a model: source versioning, chunking and embedding jobs under orchestration, index freshness monitors, retrieval evaluation in CI and removal of withdrawn or personal content from the index. Pairs with RAG as a Service, Enterprise RAG and the managed vector database.
PII discovery, masking and retention jobs; posture from DSPM and leakage control with Purview DLP.
Pick a signal to see what we watch, a sample monitor, and what happens when it fires.
—
—
The orders team owns the orders table; finance, the CRM and the AI assistant depend on it. The contract states the columns, types, owners, quality rules and personal-data tags. Try a change and see what CI says.
Indian rules turn data engineering choices into legal ones. DataOps makes them repeatable and provable.
Under the DPDP Act personal data is processed for a specified purpose, and Section 8(7) requires erasure when consent is withdrawn or the purpose is no longer served — unless another law needs it kept. We carry purpose and legal-basis tags in the catalogue and lineage, so a withdrawal reaches every downstream table and index.
Rule 8 sets three-year inactivity erasure for large e-commerce, online gaming and social media platforms, with at least 48 hours' notice before erasure, and Rule 8(3) asks for relevant personal data, traffic data and logs to be kept for at least one year for specified purposes. We implement both as orchestrated jobs with evidence.
RBI's circular of 6 April 2018 requires payment-system data to be stored only in India. Its FAQ allows processing abroad, but the data must be deleted there and brought back to India within one business day or 24 hours of processing, whichever is earlier. We pin storage and pipelines to Indian regions.
Section 16 lets the Government restrict transfers of personal data to notified countries, and stricter sector rules (RBI, SEBI, IRDAI) still apply. Lineage shows every cross-border copy — including SaaS tools and AI APIs — before a regulator or customer asks.
Summaries of the DPDP Act 2023, DPDP Rules 2025 (as notified 13 November 2025) and RBI circular DPSS.CO.OD.No 2785/06.08.005/2017-18 and its FAQ, read November 2026. Orientation only, not legal advice; your organisation remains responsible for compliance. For consent, rights and breach workflows see the DPDP Compliance Platform.
Five levels we use in assessments. The check below places you on it in two minutes.
Unanswered questions count as "Not yet".
Your 90-day roadmapData downtime = incidents × (hours to detect + hours to fix). Put in your own numbers; the improvement percentages are assumptions you control, not promises.
Downtime hours = N × (detect + fix). Cost = N × fix × engineers × engineer rate + downtime hours × business cost per hour. After = same formula with N × (1 − a), detect × (1 − b) and fix × (1 − c). All inputs are yours; no industry averages are assumed. Not a quote.
Retrieval-augmented generation is a data pipeline with a model at the end. We run it like one.
DataOps needs both: engineers who write and review pipelines, and an operations desk that notices when the 3 a.m. run fails.
We operate the tools you have, and recommend open formats and orchestrators where you are starting fresh.
* Product names are trademarks of their owners, used only to describe compatibility. No partnership or endorsement is implied unless stated on the vendor's own XcellHost page.
Tier prices match Managed Data Lake Services, where the same team runs the platform. Cloud storage and compute are billed separately at actuals.
Build & onboarding ₹1,99,999* one-time
Build & onboarding ₹4,99,999* one-time
Start with the ₹74,999* readiness assessment
* Indicative — confirm with XcellHost at proposal. INR, excl. 18% GST. All prices are reused from existing XcellHost pages: tiers, build fees, extra pipelines (₹2,999*/month), readiness assessment and governance setup from Managed Data Lake Services; DSPM, Vector DB, RAG, XcellPrivacy and Database Operations from their own pages.
Lines go to the cart at price × quantity; GST is added at 18%. Enterprise comes as a quotation.
* Indicative — confirm with XcellHost. INR. Cloud storage, compute and licences billed separately at actuals.
DataOps is a way of building and running data pipelines with software engineering discipline: code in version control, automated tests, CI/CD, orchestration, monitoring of the data itself, and clear ownership. The aim is data that arrives on time, is correct, and is used only as allowed.
DevOps ships application code. DataOps ships code and data together, and data changes every day even when the code doesn't. So DataOps adds data tests, freshness and volume monitors, contracts between teams and lineage on top of the usual CI/CD.
Monitoring the data, not only the jobs: freshness (did it update on time), volume (did the expected rows arrive), schema (did columns change), distribution (are values in normal ranges) and lineage (what is upstream and downstream). A job can succeed and still load bad data; observability catches that.
A versioned agreement, usually a YAML file, between the team that produces a dataset and the teams that use it. It lists columns, types, owners, quality rules and personal-data tags. CI checks every change against it, so a dropped or retyped column is caught before it breaks a report or model.
You need an orchestrator; which one depends on what you already run. We operate leading open-source orchestrators such as Airflow* and Dagster* and cloud-native schedulers. For a fresh start we usually recommend an open-source option to avoid lock-in.
Purpose and personal-data tags in the catalogue, lineage to every copy, masking where data isn't needed and retention as scheduled, tested erasure jobs. That makes Section 8(7) erasure and the Rule 8 retention periods provable. The DPDP Rules' core duties apply from 13 May 2027; compliance stays with you.
Under RBI's April 2018 circular, payment-system data must be stored only in India. If it is processed abroad it must be deleted there and brought back within one business day or 24 hours. We pin storage and pipelines to Indian regions and show this in lineage.
Yes. Document ingestion, masking, chunking, embedding and index refresh run as orchestrated, monitored pipelines, with retrieval evaluation in CI and withdrawn or personal content removed from the index. It pairs with our RAG, Enterprise RAG and vector database services.
Essentials ₹49,999* per month (15 pipelines) and Professional ₹1,24,999* per month (60 pipelines), with build and onboarding of ₹1,99,999* or ₹4,99,999* one-time and ₹2,999* per extra pipeline per month. Enterprise is on quote. Indicative, excl. 18% GST; cloud usage at actuals.
You do. Pipelines live in your Git repositories, data in your accounts or your XcellHost project. On exit we hand over repositories, runbooks, contracts and catalogue exports and remove our access.
A free 30-minute DataOps review: your current pipelines, the three biggest downtime risks and a package price in rupees within two business days.*
DATAOPS RUN