DATA CENTRE SERVICE · LIVE
1 PB LIVE · ACCEPTING TENDER BIDS & PILOTSCithorum operates a 1 PB managed sovereign cloud in India running Jam end-to-end. We bring the hardware, the software data plane, the compliance posture, and the monthly benchmark + restore proof. You bring the workload — banking, healthcare, government, or enterprise — and we run it on roughly half the iron and a fraction of the energy of a conventional facility.
01 — THE OFFERING
Jam-on-Linux on AMD EPYC + NVMe. ZFS or Ceph topology. S3-compatible endpoints with isolated buckets per tenant. One MW of power gives us 300–700 PB of usable storage under Jam — the 1 PB pod is the wedge, not the ceiling.
We run it. Monitoring, backups, restore drills, security hardening, customer dashboards, billing export. Named technical contact. SLA per tier.
Monthly report: TB ingested, TB saved, restore-event audit, cost avoided. Independently verifiable. The numbers are the product.
Data plane is ours; data ownership is yours. Pull at any time; deterministic restore from any compressed envelope. No lock-in by design.
BEYOND STORAGE — A FULLY MANAGED PLATFORM
Cithorum Data Centre is not a storage box with a price tag. It is a fully managed cloud platform — observability, analytics, security tooling, AI workloads, and hosting all run on the same Jam-compressed substrate. Buyers get one vendor instead of seven, one bill instead of seven, and SOC 2 controls across the whole stack.
S3-compatible buckets with Jam end-to-end. 3–8× on production workloads, up to 100× on backup tiers. Per-tenant isolation, deterministic restore, hash-verified bytes.
Per-customer dashboards: ingest, ratio, restore audit, throughput, cost avoided. Streaming metrics and an event bus customers can attach their own tooling to. The KG substrate underneath everything else.
Spin up training, inference, and analytics jobs on the same facility your data already lives in. Egress-free reads from Jam-compressed datasets — AI training on cold archive without the cold-tier restore tax.
Audit log, RBAC, per-bucket encryption, telemetry-only architecture. SOC 2 operationally compliant; Type I attestation in flight. CERT-In logging, RBI-aligned access controls, and DPDP Act 2023 residency built in.
Self-serve dashboard for teams that want to drive it themselves. Fully managed engagement with named technical contact for tender wins and regulated estates. Same substrate, two operating modes.
Government and enterprise tenders rarely buy off-the-shelf. The platform ships with a managed-customisation lane for tender-specific schema, isolated network domains, air-gapped extensions, and procurement-officer reporting.
01b — POD HARDWARE · WHAT'S IN THE RACK
Five all-NVMe storage nodes plus one control / GPU node, all in a single 42U colo cabinet. Cost-optimised: refurbished Mellanox networking, Milan-gen EPYC, DAC copper cabling for short runs. Every architecture choice carries forward unchanged into the 1 MW build.
Supermicro AS-1115HS-TNR 1U chassis · AMD EPYC 7313P (16C / 3.0 GHz, Milan-gen, single socket) · 256 GB DDR4-3200 ECC · 8 × Samsung PM9A3 30.72 TB U.2 NVMe (PCIe Gen 4, TLC, 1 DWPD) · Mellanox ConnectX-5 dual-port 100 GbE (refurbished, vendor warranty). ~245.76 TB raw per node.
Same chassis, CPU and NIC for parts commonality. NVIDIA L4 GPU (24 GB, 72 W) for inference today; upgrade path to L40S when training revenue lands. 2 × 1.92 TB NVMe boot + local cache. Hosts the RustFS coordinator, customer dashboards, billing, audit logs, and the security-ops VM (Lucas's Kali pattern) until a regulated tenant pays for a dedicated box.
Mellanox SN2700 32-port 100 GbE ToR (refurbished, ~$5–8K used vs ~$15–20K new for SN3700C). 1 GbE managed OOB switch for IPMI / BMC access. 14 × 100 GbE DAC passive copper cables (under 3 m in-cabinet runs, 4–5× cheaper than active optical).
1.23 PB raw · ~983 TB usable after EC k=4,m=1 (exactly one chunk per node, 25% storage overhead, survives loss of any one node). ~1 Tbps east-west fabric. 4–5 kW typical draw / ~6 kW peak — Milan EPYC + Gen 4 NVMe runs cooler than Genoa + Gen 5. 7–9 U occupied in a 42 U cabinet (6 × 1U servers + 1U ToR + 1U OOB).
Full technical walkthrough — data path, replication and erasure coding math, backup strategies, power and cooling, leaf-spine topology, vocabulary glossary — is at /pod.
02 — HARDWARE DELTA · SUSTAINABILITY
The 1 MW reference model translates Jam into the thing operators can see: fewer racks, lower facility draw, a smaller battery envelope, and avoided energy that lands cleanly on the ESG line. The conservative case below uses the live 3.8× ratio. The best case uses the 100× ratio observed on backup workloads.
CONSERVATIVE · 3.8× ratio · production workloads
WITHOUT JAM
WITH JAM
rack count
facility demand
energy avoided
because the UPS envelope is smaller
BEST CASE · 100× ratio · backup-tier workloads
WITHOUT JAM
WITH JAM (BACKUP-TIER)
rack count
facility demand
energy avoided
compression vs rsync
Backup-only workloads layered with rsync-class deduplication. 100× compression observed in the April 2026 live test (123 GB → 1.18 GB). See /customers#live-test for the unedited footage and source-truth register.
02b — POWER ECONOMICS · OFF-GRID READY
Less iron means a smaller power envelope. A smaller envelope means renewables become viable. Cithorum's 1 PB pod plus Suraj's 2,000 MW captive solar network unlocks an off-grid energy economics that grid-tied data-centre operators cannot match.
Cithorum's power channel runs through Suraj's solar network in India: ~70 lakh units per year of renewable generation already in operation, available at ₹2.35/unit against the ₹8/unit commercial-tariff baseline. ~70% off the grid running cost.
A fully off-grid solar-powered 1 MW data-centre build costs ~₹2.25 Cr in India today. With Jam's -32% facility draw, the same envelope serves the workload of a 1.4 MW grid-tied conventional facility — for capex that pays back within a few tender cycles.
The smaller UPS envelope means VRLA (lead-acid) batteries are viable instead of LFP — lower capex, lower maintenance overhead, and Suraj's existing battery supply chain becomes a Cithorum advantage rather than a vendor relationship.
For backup-only workloads (the bulk of regulated archive tenders), Jam's power footprint drops to ~70 kW per MW of equivalent conventional service — a single solar cluster covers the entire facility with surplus capacity left for compute or AI workloads.
03 — LIVE TEST · APRIL 2026
VM snapshots are the workload data centres archive every day. The numbers above are what your archive looks like under the Cithorum Data Centre Service.
04 — WHO IT'S FOR
VPs of platform / infra.
Buying signal: storage cost growing faster than budget, restore SLAs slipping, hyperscaler bills no longer defensible at the board.
Public sector, defence-adjacent, healthcare.
Buying signal: regulator demands customer-controlled storage; hyperscaler exit cost is unacceptable; air-gapped or on-soil capacity is mandatory.
Operators looking to differentiate.
Buying signal: customers demand higher density and better economics without new builds. Jam adds a software layer that compounds on top of the iron they already own.
05 — COMPLIANCE & SECURITY
The architecture sidesteps the biggest objection regulated buyers raise: Jam operates on infrastructure telemetry — power, thermal, compute, throughput. It does not access or process customer application data. The compliance story is a side effect of the architecture, not a wrapper around it.
Jam reads CPU, thermal, power, and storage-throughput signal — never application payload. Customer data stays in customer buckets. Encrypted in transit (TLS), encrypted at rest, hash-verified on restore.
SOC 2 operationally compliant via continuous control monitoring; formal Type I attestation in flight (Q4 2026), Type II Q3 2027. Designed to operate inside RBI IT & Cybersecurity Framework environments and PCI-DSS-scoped estates. ISO 27001 operationally compliant with full Annex A control mapping; formal certification in flight Q4 2026.
HIPAA-aligned deployment patterns; ISO 27701 privacy controls. Genomics workloads run in customer-controlled buckets with full chain-of-custody and audit trail per restore event.
CERT-In logging and incident-response integration. MeitY empanelment route mapped through Suraj's Karnataka network. DPDP Act 2023 ready: data residency on customer-controlled storage, exit guaranteed by design.
Cithorum India Pvt Ltd is the operating entity. SOC 2 operationally compliant today; the dataroom includes the full control mapping (RBI / PCI-DSS / ISO 27001 / ISO 27701 / HIPAA / DPDP Act) for diligence and procurement review.
06 — DEPLOYMENT MODEL
Banks, hospitals, and public-sector buyers do not deploy to production on day one. The Data Centre Service walks a graduated trust ramp — every phase has a measurable exit and the customer keeps control throughout.
01
Read-only telemetry ingestion. We measure your existing estate — thermal, power, storage utilisation, throughput — and produce a baseline benchmark report. No writes, no workload changes, no risk to production.
02
Recommendations on top of the baseline: where Jam saves bytes, where cooling can be tuned, where the rack and battery envelope shrinks. Customer keeps the decision and the timing — we hand over the report and the model.
03
Customer-approved changes only, scoped per workload. Jam runs end-to-end on a defined slice; restore proofs and billing reports produced monthly. Scope expands with the customer, never ahead of them.
07 — VS WHAT YOU'RE DOING TODAY
VS HYPERSCALER COLD TIER
Cheap to put in. Slow, opaque, expensive to get out. Cithorum: same economics or better, fully restoreable in seconds, on infrastructure you control.
VS ON-PREM ALONE
You bought iron and got density. You did not get a restore audit, a benchmark report, or a software layer that compounds. Cithorum sits on existing storage and produces measured savings.
VS FLAT PER-TB CLOUD
Flat pricing wins on simplicity. Wins nothing on speed, AI loading, genomic alignment, or restore proof. Cithorum compresses AND accelerates.
07b — COMPARED TO BEST-IN-CLASS · ON THE NUMBERS
Section 07 named the comparators. Below is the like-for-like cost and capability picture against the storage stacks regulated and mid-market buyers actually benchmark us against. The two effects that compound — inline compression at 3–8× on production (and up to 100× on backup tiers) and a queryable Knowledge Graph layer shipped on the same substrate — are what move the numbers.
AWS S3 STANDARD
The hyperscaler default. World-class durability and ecosystem. But: heavy egress fees, no inline compression on customer bytes, off-soil for India, and a separate Neptune line item if you need a graph layer over the data.
Cithorum delta: approximately 50% below commercial cloud list rates effective post-Jam, zero egress, KG bundled, on-soil sovereign.
PURE · NETAPP · VAST
Industry-leading $/IOPS and deep enterprise integration. But $/TB usable runs multiples of cloud rates, the model is capex-heavy, the customer operates the iron, and no managed graph layer ships in-platform.
Cithorum delta: managed-service economics on equivalent NVMe substrate, end-to-end Jam, KG bundled, no capex commitment.
WASABI · BACKBLAZE B2
Cheap, simple, no egress fees. But the price is the price — no inline compression, no analytics or graph layer, US-headquartered with limited Indian-soil options. Restore SLAs are basic.
Cithorum delta: Jam takes the effective rate to ~$1–2 / TB-equivalent, plus restore-proof drills, KG, and tender-ready sovereign hosting.
CITHORUM DATA CENTRE
Live at M2M TechConnect's tier. End-to-end Jam, 7.85× verified compression on the April 2026 live test (135 GB → 17.2 GB at 281 MB/s encode, 1.13 GB/s decode). NVMe-bound throughput, restore in seconds, sovereign by design.
Effective rate: 3–8× below headline cloud on production workloads, >10× below on backup-tier archives.
Why the numbers look like this — the two compounding effects.
Every other vendor charges per byte stored. Jam compresses inline at 3–8× on production data and up to 100× on backup tiers (verified live in April 2026). Customers ingest 1 TB; we store ~125–325 GB. Pricing tracks compressed bytes — which is how Cithorum lands at approximately 50% below hyperscaler list rates without sacrificing margin.
Hyperscalers and enterprise vendors sell storage and graph databases as separate products with egress fees in between. Cithorum's Knowledge Graph runs on the same Jam-compressed substrate. Genomics relationships, legal citation graphs, revenue-ops joins — all queried in-platform with zero data movement and one bill. For regulated buyers this collapses an entire vendor row off the procurement diagram.
The same rack stores 3–8× more data and answers graph queries against it without lifting a byte off disk. The buyer's alternative is two vendors, two bills, an egress line, and an integration project. The Cithorum stack replaces all three — which is why the comparison above lands the way it does.
08 — THE INDIA ROUTE
Cithorum's commercial route into India runs through Suraj Chirag's network: Karnataka state programmes, Keonics partnerships, BIRAC biomedical pipelines, and central-government tender access. The market is timed; the catalysts are named.
Karnataka has announced 1,000 MW of sustainable data-centre parks across Bengaluru and two further cities. Jam's -442 kW per MW facility delta and ~3.87 GWh per year of avoided energy map directly onto the programme's sustainability mandate.
The National Informatics Centre is the largest sovereign data buyer in India; e-procurement.gov.in is the tender surface where the relevant data-centre, storage, and cloud workloads are floated. Cithorum is enrolled and actively triaging the IT / Cloud / Data Centre classifications.
Cithorum India Pvt Ltd is Udyam-registered under the Public Procurement Policy for Micro & Small Enterprises: EMD exemption, tender doc fee exemption, and a 25% procurement preference on MSE-flagged tenders. Turnover and experience eligibility relaxed against incumbent system-integrator primes.
Software Special Economic Zone registration unlocks an 18–20 year tax holiday for sovereign data-centre operators in India. Cithorum India Pvt Ltd is structuring the operating entity inside an SEZ — the running cost advantage compounds with Jam's 50% facility delta and the ₹2.35/unit captive solar economics.
India produces ~30% of the world's data and stores only ~3% of it on-soil today. The DPDP Act 2023 plus the Karnataka programme push that gap closed fast. A ten-point shift toward 13% on-soil represents a multi-Cr infrastructure opportunity — and it must run on Indian-managed substrate.
Karnataka's state electronics corporation. The operator-side channel for state-funded build-outs and regulated-workload tenders. Suraj-led network access.
INR 24.37 Cr biomedical workstream in flight via Suraj's clinical-science network. Genomics, sequencing, regulated medical archives — the workloads where Jam's FASTQ-class compression compounds.
Cithorum India Pvt Ltd is the operating entity for the India route, structuring inside a Software SEZ for the long tax-holiday horizon. UK + India presence; one balance sheet, one team, one IP register. Tender bids, project-finance lines, and pilot engagements run through this entity; the captive solar power channel and lead-acid UPS supply chain run alongside.
Read the full plan →09 — PRICING
The Data Centre Service is a managed engagement, not a self-serve product. Numbers are scoped per workload; tiers describe the shape of the deal.
PILOT
Discounted pilot tier with benchmark-rights agreement. Two weeks to first benchmark report on the customer's own workload. Convert directly into Production.
Run a pilot →PRODUCTION
Per-TB-per-month managed service with SLA, monthly billing, restore proof. Named technical contact. Customer dashboards and billing export included.
Talk to us →ENTERPRISE / GOV / TENDER
Volume terms, sovereign-data clauses, project-financed work-order option (India tender route). Air-gapped deploys, appliance option, scoped SLA.
Contact us →All tiers include the live benchmark register and 100% restore-verification on a sampled set.