Skip to content
Contact (852) 2763 3180
Book an Assessment

Solution Blueprint · Real-Time Data & AI

Real-Time Analytics On The Data You Already Have

Faster dashboards and ad hoc queries for many people at once, without rebuilding your data warehouse or data lake.

Layers
VisualiseAnalyseFoundation
Reading time
4 minutes
Last reviewed
October 2026
Book an Assessment
See all insights

The Problem

You’ll recognise this if:

Dashboards take more than ten seconds to load, and longer at month end.

When many people query at once, everyone slows down.

Layers of pre-aggregated tables exist only to make reports usable.

Warehouse costs rise every year with data volume.

What Good Looks Like

We agree these measures with you as the exit criteria of the PoC. We report what we measure; we do not promise a figure in advance.

  • Query times for your slowest dashboards and queries, side by side with today
  • Behaviour at your peak number of concurrent users
  • Data freshness: time from a source change to a query result
  • Infrastructure and licence cost for the same workload
Vendor figure

StarRocks 4.0 is about 60% faster year on year on the TPC-DS benchmark. Source: StarRocks 4.0 release announcement (opens in a new tab)

Vendor figure

PhoenixAI (formerly CelerData) states that StarRocks serves customer-facing analytics at tens of thousands of queries per second. Source: PhoenixAI Anywhere (formerly CelerData Enterprise) (opens in a new tab)

Reference Architecture

The Kepro data ecosystem from Foundation upwards, with the layers this blueprint uses highlighted. Product names link to their pages.

  1. Visualise

    FanRuan FineBI and FineReport query StarRocks directly. Other BI tools you already use can connect the same way.

    With FanRuan

  2. Analyse

    StarRocks (PhoenixAI Anywhere): an MPP engine with vectorised execution and a cost-based optimiser. It runs with local NVMe storage, or on shared S3-compatible object storage. Asynchronous materialised views rewrite queries automatically, and it speaks the MySQL protocol.

    With StarRocks

  3. Prepare (Not Used In This Blueprint)
  4. Connect (Not Used In This Blueprint)
  5. Foundation

    Dell PowerEdge servers with NVMe drives for the engine; Dell PowerScale or ObjectScale where data sits in a shared file or object store.

    With Dell Technologies

Data flows bottom → top

Data sourcesYour existing warehouse, data lake (including Apache Iceberg, Hive and Hudi tables) and operational databases loaded in real time.

Technology Stack

Technology stack for this blueprint
LayerProductEdition / ModelRoleKey Specification
AnalyseStarRocksPhoenixAI Anywhere (StarRocks 4.x)Analytics engineVectorised MPP engine; materialised views with automatic query rewrite; Iceberg, Hive and Hudi catalogues; MySQL protocol; row- and column-level access control
VisualiseFanRuanFineBI 7.0 and FineReport 11.5Dashboards and reportsDirect query; Metrics Hub
FoundationDell TechnologiesPowerEdge R770 or R7725; PowerScale or ObjectScaleCompute and storageIntel Xeon 6 (R770, up to 8 TB DDR5) or AMD EPYC (R7725)

How We Deliver

Five steps, each with a clear exit before the next. The timeline is agreed with you in the Design step.

  1. 01

    Data Readiness Assessment

    We profile your slowest queries, concurrency and data volumes, and estimate sizing.

    From youQuery logs, or a list of the reports people wait for.

  2. 02

    PoC

    Your real queries replayed on StarRocks side by side with today, with a cost model.

    From youA copy of, or read access to, the data behind those queries.

  3. 03

    Design

    Node sizing, Dell bill of materials, security model and migration order.

    From youAgreement on which dashboards move first.

  4. 04

    Implementation

    Parallel running, then dashboards moved over one by one.

    From youReport owners to sign off each dashboard.

  5. 05

    Support & Optimisation

    On-site support, query tuning and capacity reviews.

    From youNotice of new data sources and users.

Standards And Controls

  • Access Control

    Role-based access with row- and column-level rules, linked to your directory (LDAP).

  • Open Formats

    Queries run on Apache Iceberg tables in place, so data stays in an open format.

Before You Start

Readiness Checklist

  • A list of the dashboards and queries that matter most.
  • Read access to the source warehouse or data lake for the PoC.

Risks And How We Handle Them

  • SQL written for another engineStarRocks speaks the MySQL protocol; incompatible SQL is listed and rewritten during the PoC.
  • Judging on one averageResults are reported per query and per concurrency level, not as a single number.

Sources

  1. StarRocks 4.0 release announcement (opens in a new tab)
  2. PhoenixAI Anywhere (formerly CelerData Enterprise) (opens in a new tab)
  3. Dell PowerEdge R770 spec sheet (opens in a new tab)

Last reviewed October 2026. Vendor facts change; the linked sources are the reference.

Related

Products In This Blueprint

StarRocks
Real-time analytics and AI serving engine.
FanRuan
Formatted reports, data entry and self-service BI for many users.
Dell Technologies
Compute, primary storage, networking and protection hardware.

Related Blueprints

Real-Time Data & AI

Real-Time Customer 360: One Live Record Across Every System

A governed customer record that every channel, dashboard and AI assistant can use, built from the systems you already run.

Layers: Visualise, Analyse, Connect, Foundation, Protect
Read blueprint
Reporting & Automation

From Spreadsheets To Governed Enterprise Reporting

Fixed-format reports, online data entry and self-service analysis on one platform, with one definition for every metric.

Layers: Visualise, Analyse
Read blueprint
Infrastructure For Data & AI

A Dell Data Centre For Data And AI

Servers, storage, networking and data protection from one vendor, sized from measured workloads.

Layers: Foundation
Read blueprint

Data Readiness Assessment

Talk To Our Architects About This Blueprint

Start with a Data Readiness Assessment: we map your data sources, pain points and risks, and recommend what fits your situation, not a preset product.

Book a Data Readiness Assessment
See all insights