The Local-First Telecom Platform: Building a Dual-OSS Tracker on IndexedDB with Zero Dependencies

"When network engineers in regional control rooms have to download 150MB spreadsheets over an unstable 3G hotspot just to verify cell alarms, your cloud portal isn't modern—it's an operational liability."

1. The Problem with Cloud-Centric OSS Portals in Frontier Operations

In national telecommunications networks spanning Tanzania, engineering operations are split across disparate Operations Support Systems (OSS). In our operational perimeter, radio access networks (RAN) and core transport telemetry lived predominantly inside two major platforms: Huawei iMaster MAE (formerly U2000) and Huawei NetEco (site power and green energy management).

Every morning, tier-2 and tier-3 NOC engineers faced a brutal bottleneck. Central management had mandated a centralized web portal built on a standard cloud stack: a heavy client-side framework pulling hundreds of megabytes of daily site status, MML command logs, and power alarms via REST endpoints.

In Dar es Salaam headquarters on high-speed fiber, the system took 4 to 6 minutes to load. But for field supervisors and regional operations centers in Dodoma, Mbeya, and Kigoma—connecting through cellular modems subject to weather attenuation and regional fiber cuts—the portal repeatedly timed out. Engineers spent an average of 45 minutes every morning waiting for tables to render before they could dispatch field maintenance teams.

2. The Architectural Pivot: Local-First with Browser-Native IndexedDB

Rather than upgrading cloud database instances or adding caching proxies that still required round-trip network hops, we made an uncompromising architectural bet: the browser must become the database.

Instead of requesting pre-computed aggregations from a central backend, the browser downloads raw, compressed delta feeds directly into the client's local IndexedDB. Once persisted, all filtering, cross-system joins (matching MAE radio cell alarms with NetEco solar battery voltages), and MML batch command script generation execute in-memory with sub-millisecond latency.

3. Chunked Streaming & Web Worker Ingestion

Parsing 80,000 site rows and matching alarm hierarchies on the browser's main thread would trigger severe frame drops and "Page Unresponsive" warnings. To prevent UI freezing, we offloaded ingestion to a dedicated Web Worker that processes streamed text chunks.

// Web Worker: Streaming parser with deterministic batch commits
self.onmessage = async function(e) {
  const { blob, schemaVersion } = e.data;
  const stream = blob.stream();
  const reader = stream.getReader();
  const decoder = new TextDecoder('utf-8');
  
  let buffer = '';
  let batch = [];
  const BATCH_SIZE = 1000;

  while (true) {
    const { done, value } = await reader.read();
    if (done) break;
    
    buffer += decoder.decode(value, { stream: true });
    const lines = buffer.split('\n');
    buffer = lines.pop(); // Keep partial trailing line

    for (let i = 0; i < lines.length; i++) {
      const record = parseMmlRecord(lines[i]);
      if (record) {
        batch.push(record);
        if (batch.length >= BATCH_SIZE) {
          await writeToIndexedDB(batch);
          self.postMessage({ type: 'PROGRESS', count: batch.length });
          batch = [];
        }
      }
    }
  }
  if (batch.length > 0) {
    await writeToIndexedDB(batch);
  }
  self.postMessage({ type: 'COMPLETE' });
};

4. IndexedDB Index Design for Multi-Dimensional Querying

The schema was structured with composite indexes designed specifically around the daily diagnostic questions NOC managers ask:

• "Which 4G LTE cells in Mwanza currently report VSWR high alarms AND have battery autonomy below 4 hours?"
• "Show all sites pending modernization with scheduled fiber cutovers today."

// Setting up indexed object stores for instant compound lookups
const request = indexedDB.open('TelecomNocDB', 3);

request.onupgradeneeded = function(e) {
  const db = e.target.result;
  
  if (!db.objectStoreNames.contains('sites')) {
    const siteStore = db.createObjectStore('sites', { keyPath: 'site_id' });
    siteStore.createIndex('by_region', 'region', { unique: false });
    siteStore.createIndex('by_status', 'operational_status', { unique: false });
    siteStore.createIndex('region_status', ['region', 'operational_status'], { unique: false });
  }

  if (!db.objectStoreNames.contains('alarms')) {
    const alarmStore = db.createObjectStore('alarms', { keyPath: 'alarm_id', autoIncrement: true });
    alarmStore.createIndex('by_site', 'site_id', { unique: false });
    alarmStore.createIndex('by_severity', 'severity', { unique: false });
    alarmStore.createIndex('site_severity', ['site_id', 'severity'], { unique: false });
  }
};

5. The Empirical Results in Production

Deploying this local-first architecture yielded immediate, transformative operational results:

• Sync Time Reduced from 45 min to 3 min: Field engineers only fetch binary deltas, reducing morning bandwidth consumption by 84%.
• 100% Offline Capability: During a major undersea fiber severance event in 2024, regional operations teams continued querying site registries, running batch MML generators, and tracking work orders entirely offline without interruption.
• Zero Cloud Infrastructure Maintenance: By eliminating server-side rendering farms and Redis query caches, infrastructure compute expenses dropped to zero for that operational subsystem.