Globally distributed databases, locally fast

Azure Cosmos DB development services for applications that need sub-10ms reads across continents, multi-model flexibility, and the elastic scale that relational databases can't match. Data models designed for your query patterns, not forced into a relational shape. We build on Cosmos DB with the cost discipline that turns a powerful database into a sustainable one.

  • NoSQL API
  • Sub-10ms reads
  • Global distribution
  • RU-optimised

Why Entalogics for Cosmos DB

Four things every
Cosmos DB deployment
actually needs.

The Cosmos DB accounts we audit always have the same problems — a partition key chosen by guessing, cross-partition queries on every hot path, default indexing policy burning RUs on fields nobody queries, and a monthly bill that climbs 30% without traffic growing at all. Cosmos DB is powerful. Most implementations pay for that power without using it.

Cost01

Your partition key is your cost structure.

A bad partition key creates hot partitions that consume 80% of your RUs while the rest sit idle. We model your access patterns before provisioning and choose partition keys that distribute reads and writes evenly. The difference between a $500/mo and $5,000/mo Cosmos bill is usually the partition key.

Architecture02

Denormalise for your queries, not for your diagram.

Cosmos DB is not a relational database. Joins are expensive. Cross-partition queries are expensive. We model documents around how your application reads data — embedding related data, using change feed for materialised views, and keeping every hot-path query partition-aligned.

Performance03

Custom indexing policy, not the default.

The default policy indexes every property — burning RUs on every write. We define indexing policies that include only the fields your queries actually filter on. Write-heavy workloads see 40–60% RU reduction from this change alone.

Scale04

Autoscale or provisioned — matched to your traffic pattern.

Autoscale for variable traffic that spikes unpredictably. Provisioned throughput for stable workloads with known patterns. Serverless for development and low-traffic apps. Each tier chosen by measurement — not by default.

When Cosmos DB, when not

Cosmos DB is a tool.
Not the right database for everything.

Cosmos DB excels at global distribution, elastic scale, and sub-10ms latency. It also charges per request unit and punishes bad data models harder than any relational database. We'll tell you on the first call if Cosmos DB fits.

PICK COSMOS DB WHEN

  • Global distribution with single-digit latency across regions — Cosmos DB is built for this
  • Elastic scale with unpredictable traffic — autoscale handles millions of requests without capacity planning
  • Multi-model needs — NoSQL, MongoDB API, Cassandra API, Gremlin, or Table from one account
  • IoT, gaming, e-commerce, or real-time apps where low latency at any scale is non-negotiable

CONSIDER ALTERNATIVES WHEN

  • Relational data with complex joins and transactions — Azure SQL or PostgreSQL is simpler and cheaper
  • Single-region application with predictable traffic — Cosmos DB's global distribution value disappears
  • Budget-constrained with small data — Cosmos DB's minimum cost exceeds what a managed PostgreSQL handles for less

WE SAY NO WHEN

  • "Cosmos DB because it's NoSQL." That's a category, not a reason.
  • "Migrate our SQL Server to Cosmos." Relational workloads rarely benefit from a document database.
  • "We'll figure out the partition key later." Later means a re-architecture and data migration.

What we build on Cosmos DB

Six product surfaces.
One quality bar.

The shapes of Cosmos DB development we deliver most. Each modelled for query patterns and cost-optimised from day one.

  • S01

    Real-time application backends

    E-commerce, gaming, social — high-throughput read/write with sub-10ms latency. Documents modelled around access patterns, not entity relationships.

    COSMOS NOSQLCHANGE FEEDAUTOSCALESDK V3
  • S02

    IoT data platforms

    Device telemetry ingestion at millions of events per second. Hierarchical partition keys combining device ID and date. TTL for automatic data expiration. Change feed for real-time analytics.

    COSMOS NOSQLHIERARCHICAL KEYSTTLCHANGE FEED
  • S03

    Multi-region global apps

    Active-active multi-region writes with conflict resolution. Users in Tokyo, London, and New York all get sub-10ms reads from their local replica.

    MULTI-REGIONMULTI-WRITECONFLICT RESOLUTIONSLA
  • S04

    MongoDB API on Cosmos DB

    Existing MongoDB workloads running on Cosmos DB's managed infrastructure. Zero code changes. Global distribution and elastic scale without managing replica sets.

    MONGODB APICOSMOS DBGLOBAL DISTRIBUTIONMANAGED
  • S05

    Cosmos DB cost optimisation

    Existing accounts with runaway costs. Partition key analysis, indexing policy tuning, throughput right-sizing, and change feed for materialised views that eliminate cross-partition queries.

    RU ANALYSISINDEXING POLICYAUTOSCALECOST TUNING
  • S06

    Relational to Cosmos migration

    SQL Server or PostgreSQL data modelled for Cosmos DB. Documents denormalised around query patterns. Change feed replacing triggers. Migration with performance validation.

    DATA MODELLINGMIGRATIONCHANGE FEEDSDK V3

The playbook

Patterns we
ship on repeat.

Cosmos DB patterns from real production deployments — not quickstart tutorials.

  • P01

    Partition key modelling before provisioning

    Access patterns documented. Read/write ratios measured. Partition key chosen for even distribution and query alignment. Never chosen by field name alone.

  • P02

    Custom indexing policy

    Only the fields your queries filter on are indexed. Composite indexes for ORDER BY queries. Spatial indexes only where geo-queries exist. Default policy replaced on day one.

  • P03

    Change feed for materialised views

    Write-optimised container plus change-feed-projected read containers. Every query partition-aligned. Cross-partition fan-out eliminated from hot paths.

  • P04

    Hierarchical partition keys for multi-tenant

    Tenant ID as the first level, entity type as the second. Data isolated per tenant. Queries scoped to a single logical partition. No cross-partition reads for tenant-specific data.

  • P05

    Autoscale with budget caps

    Autoscale with max RU/s set to a known ceiling. Budget alerts in Azure Cost Management. No bill surprises from a traffic spike that provisioned 100,000 RU/s permanently.

  • P06

    SDK V3 with bulk and batch

    Bulk operations for high-throughput ingestion. Transactional batch for multi-document atomicity within a partition. SDK configured with direct mode and TCP for lowest latency.

Signature case

An e-commerce backend,
remodelled from hot partitions to even distribution.

A B2C e-commerce platform on Cosmos DB — monthly bill at $8,200, single partition absorbing 78% of traffic, cross-partition queries on the product listing page, and default indexing policy burning RUs on 40 unused fields. Remodelled the partition key, custom indexing policy, change feed for materialised product views, and autoscale with budget caps in 6 weeks. Monthly bill dropped to $3,100. Latency unchanged at 7ms.

Before

$8,200/mo · hot partition at 78% · cross-partition product queries · default indexing · no autoscale cap

After

$3,100/mo · even distribution · partition-aligned queries · custom indexing · autoscale with cap

  • Monthly cost−62%
  • Hot partition load78% → even
  • To fully optimised6wk
  • Latency unchanged7ms

Engagement shape

Eight to ten weeks
to a measurable ship.

A typical Cosmos DB engagement. We model and deploy container by container — the current data stays live while we work.

  • W01

    Audit + RFC

    Two senior Cosmos DB engineers. Partition key analysis, RU consumption profiling, indexing policy review, cost breakdown. A ranked, dollarized RFC.

  • W02–03

    Data model + first container

    Access patterns modelled, partition keys chosen, indexing policy defined, first container deployed with SDK integration. Real RU consumption measured.

  • W04–08

    Container by container

    Each container modelled, migrated, and validated. Change feed wired for materialised views. Autoscale configured. Your application keeps running.

  • W09+

    Handoff

    Cost dashboard live. Indexing policies tuned. Runbook handed to your team — or we stay on retainer.

Stack

Tools we
reach for first.

Our default Cosmos DB development stack — picked for production.

  • DatabaseCosmos DB NoSQL API · MongoDB API · Cassandra API · Table API
  • SDK.NET SDK V3 · Node.js SDK · Java SDK · Python SDK
  • ModellingPartition key analysis · Indexing policy design · Change feed
  • CostAzure Cost Management · RU consumption analysis · Autoscale
  • MigrationAzure Data Factory · Custom migration tools · Change feed
  • MonitoringAzure Monitor · Cosmos DB Metrics · Datadog · Application Insights

Engagement

Three ways
to work with us.

No hourly retainer that bills for "thinking time." Pick a lane that matches your stage; everything is fixed-quote or transparently rated.

FIXED SCOPEone-off build

Ship a Cosmos DB deployment, end-to-end.

A defined scope, a fixed price, a senior-only team. From data modelling to production in 6–12 weeks.

$15k–$30k

FIXED SCOPE

  • Senior engineers only
  • Fixed quote in week 1
  • Code, infra, runbook — yours
Plan a fixed build
DEDICATED TEAMmonthly

Hire dedicated Cosmos DB engineers.

Embedded engineers in your team. Senior data architects specialising in Cosmos DB. Pause, resize, end with 30 days' notice.

$5k / eng / mo

PER ENGINEER

  • Same senior bar as fixed-scope
  • Embedded in your team
  • Founder-direct escalation
Hire dedicated Cosmos DB devs
ENGAGEMENTcustom

Strategic Cosmos DB partnership.

A long-term partner for data-intensive applications — cost optimisation, global distribution architecture, migration strategy, hiring help.

custom

PROCUREMENT-FRIENDLY

  • Multi-quarter roadmap
  • Architecture & hiring partner
  • Procurement-friendly paper
Speak to the founder
FAQ

Sharp questions,
straight answers.

Partition keys, runaway bills, Cosmos DB vs Azure SQL — the questions we get on every Cosmos DB discovery call.
Model your access patterns first. The partition key should align with your most frequent query filter and distribute data evenly. A bad key creates hot partitions that waste RUs. We model before provisioning — changing a partition key after deployment means a data migration.
Usually three things: hot partitions wasting provisioned RUs, cross-partition queries on hot paths, and default indexing policy indexing fields nobody queries. We've cut bills 40–60% by fixing these three issues alone.
Cosmos DB for global distribution, elastic scale, and sub-10ms latency on document workloads. Azure SQL for relational data with complex joins and transactions. Most applications don't need Cosmos DB — we'll tell you if yours does.
Yes. The engineers who write the RFC ship the data model. No handoff mid-engagement. Direct access throughout.
Yes. We audit partition keys, indexing policies, throughput settings, and query patterns. Most existing deployments have 30–60% cost reduction available without changing application code.

Founder-direct

Tell us whatyou're building.

Thirty minutes with the founder. We'll bring a senior Cosmos DB engineer, the relevant playbook, and a candid read on whether Cosmos DB is the right database — or whether Azure SQL, PostgreSQL, or DynamoDB fits your workload better.