System Requirements
Hardware and OS requirements for each Purple8 Graph deployment tier, with practical data-size context to help you right-size your deployment.
Tiers at a glance
| Tier | Deployment | Node cap | Edge cap | Queries/min |
|---|---|---|---|---|
| Desktop Free | Local app | 1M | 10M | 1,000 |
| Desktop Pro | Local app | 10M | 100M | 5,000 |
| Pro Cloud | Managed (shared) | 25M | 250M | 5,000 |
| Cloud Plus | Managed (dedicated) | 100M | 1B | 20,000 |
| Self-Hosted Server | Your infra | Unlimited | Unlimited | Unlimited |
| Server Enterprise | On-prem / VPC | Unlimited | Unlimited | Unlimited |
Desktop Free & Desktop Pro
Desktop Free — local macOS / Windows app, 1M node cap, no key required Desktop Pro — local macOS / Windows app, 10M node cap, $29/mo
What fits?
| Use case | Nodes | Edges | Typical graph |
|---|---|---|---|
| Personal knowledge graph | 50K–200K | 200K–1M | Notes, people, books, events |
| Collection & content hierarchy | 100K–500K | 500K–2M | Libraries, taxonomies, catalogs |
| Feature flag graph | 200K–1M | 1M–5M | Features, roles, tenants, cohorts |
| Code dependency graph | up to 1M | up to 10M | Files, functions, imports, call chains |
| Process similarity / dedup | up to 10M (Pro) | up to 50M (Pro) | Processes, tasks, steps, outcomes |
Recommended hardware — Desktop tiers
| Component | Desktop Free (min) | Desktop Pro (recommended) |
|---|---|---|
| CPU | 2 cores | 4–8 cores |
| RAM | 4 GB | 16 GB |
| Disk | 10 GB SSD | 100 GB SSD |
| OS | macOS 13+, Windows 10+ | macOS 14+, Windows 11 |
| Python | 3.10+ | 3.11+ |
RAM breakdown (1M nodes, 768-dim, int8):
| Component | Memory |
|---|---|
| Graph store (RocksDB) | ~400 MB |
| HNSW index (int8 quantization) | ~900 MB |
| OS + Python runtime | ~300 MB |
| Total | ~1.6 GB |
For MacBooks
Runs on 8 GB RAM with P8G_VECTOR_QUANTIZATION=int8. Use P8G_VECTOR_BACKEND=diskann on machines with < 4 GB available.
Pro Cloud & Cloud Plus
Pro Cloud — managed shared instance, 25M nodes, $99/mo Cloud Plus — managed dedicated instance, 100M nodes, $499/mo
Purple8 manages the infrastructure. No hardware to provision.
What fits?
| Use case | Nodes | Edges | Context |
|---|---|---|---|
| Mid-size enterprise knowledge graph | 5M–25M | 50M–250M | Products, customers, orgs, transactions |
| Access management graph (AD replacement) | 5M–15M | 20M–100M | Users, roles, resources, policies, tenants |
| Multi-document RAG corpus | 10M–25M | 50M–250M | Full corpus of a large enterprise wiki |
| Healthcare patient graph | 10M–25M | 50M–200M | Patients, diagnoses, medications, encounters |
| Duplicate artifact detection | 5M–20M | 20M–100M | Knowledge artefacts, similarity edges |
Managed infra specs (for reference)
| Spec | Pro Cloud | Cloud Plus |
|---|---|---|
| Compute | Shared (2–4 vCPU) | Dedicated 16 vCPU |
| RAM | Shared pool | 64–128 GB dedicated |
| Storage | 200 GB | 1 TB NVMe |
| Queries/min | 5,000 | 20,000 |
| Concurrent connections | 10 | 100 |
| Backups | Daily, 30-day retention | Daily + PITR 7-day |
| Region | us-east-1 (default) | Choice of US / EU / APAC |
Self-Hosted Pro & Server
Self-Hosted — your infra (any cloud or bare metal), unlimited nodes, $299/mo
What fits?
| Deployment | Nodes | Edges | Use case |
|---|---|---|---|
| Single-tenant production | 25M–100M | 250M–1B | Product + customer graph |
| Multi-tenant SaaS | 50M–200M | 500M–2B | Per-tenant sub-graphs |
| Process transformation graph | 50M–200M | 250M–1B | Enterprise process mining |
| Feature enablement platform | 20M–100M | 100M–500M | Features × tenants × roles × cohorts |
Recommended hardware — Self-Hosted
Single node (up to ~50M nodes)
| Component | Minimum | Recommended |
|---|---|---|
| CPU | 8 cores | 16 cores |
| RAM | 32 GB | 128 GB |
| Disk | 500 GB NVMe SSD | 1 TB NVMe SSD |
| OS | Ubuntu 22.04+ | Ubuntu 22.04+ |
| Network | 1 GbE | 10 GbE |
With Raft HA + read replica (up to ~100M nodes)
| Component | Primary | Replica ×2 |
|---|---|---|
| CPU | 16 cores | 8 cores |
| RAM | 128 GB | 64 GB |
| Disk | 1 TB NVMe | 1 TB NVMe |
| Network | 10 GbE | 10 GbE |
Cloud instance recommendations (Self-Hosted)
| Provider | Instance | vCPU | RAM | Notes |
|---|---|---|---|---|
| AWS | r6i.2xlarge | 8 | 64 GB | Good baseline |
| AWS | r6i.4xlarge | 16 | 128 GB | Full 100M nodes with DiskANN |
| GCP | n2-highmem-8 | 8 | 64 GB | |
| GCP | n2-highmem-16 | 16 | 128 GB | |
| Azure | Standard_E8s_v5 | 8 | 64 GB | |
| Azure | Standard_E16s_v5 | 16 | 128 GB |
Server Enterprise
Enterprise — on-prem, internal cloud, or dedicated VPC (GCP / AWS / Azure), custom pricing
What fits at enterprise scale?
| Deployment | Nodes | Edges | Shards | Use case |
|---|---|---|---|---|
| Large enterprise | 50M–200M | 500M–2B | 3–5 | Global org + product + customer graph |
| Regulated industry (finance, health) | 100M–500M | 1B–5B | 3–8 | Compliance graph + RBAC + audit |
| Hyperscale / platform | 500M–1B+ | 5B+ | 8–16 | Platform-level knowledge graph |
| Multi-region federated graph | 200M–1B | 1B–10B | 8–24 | Cross-region data sovereignty |
Recommended hardware — Enterprise
3-shard cluster with Raft HA (per shard group: 3 nodes)
Each shard group = 1 Raft leader + 2 followers (quorum = 2/3):
| Component | Per shard node | Per shard group (×3) |
|---|---|---|
| CPU | 32 cores | 96 cores |
| RAM | 256 GB | 768 GB |
| Disk (NVMe) | 2 TB | 6 TB |
| Network | 25 GbE | 25 GbE |
For 3 shard groups (9 nodes total): ~2.3 TB RAM, ~18 TB NVMe
Minimum viable enterprise (3 shards, no replicas)
| Component | Per node | Cluster total (×3) |
|---|---|---|
| CPU | 16 cores | 48 cores |
| RAM | 128 GB | 384 GB |
| Disk | 500 GB NVMe | 1.5 TB NVMe |
Cloud instance recommendations (Enterprise)
| Provider | Instance | vCPU | RAM | Per-node cost (est.) |
|---|---|---|---|---|
| AWS | r6i.8xlarge | 32 | 256 GB | ~$1.61/hr |
| AWS | r6i.16xlarge | 64 | 512 GB | ~$3.22/hr |
| GCP | n2-highmem-32 | 32 | 256 GB | ~$1.54/hr |
| Azure | Standard_E32s_v5 | 32 | 256 GB | ~$1.67/hr |
Storage sizing guide
Use this table to estimate disk requirements before provisioning:
| Nodes | Avg properties | Edges (×4 nodes) | 768-dim vectors | RocksDB store | HNSW index | DiskANN index | Total (est.) |
|---|---|---|---|---|---|---|---|
| 100K | 5 props | 400K | Yes | ~500 MB | ~300 MB | ~120 MB | ~1 GB |
| 1M | 5 props | 4M | Yes | ~5 GB | ~3 GB | ~1.2 GB | ~10 GB |
| 10M | 5 props | 40M | Yes | ~50 GB | ~30 GB | ~12 GB | ~95 GB |
| 100M | 5 props | 400M | Yes | ~500 GB | ~300 GB | ~120 GB | ~950 GB |
| 1B | 5 props | 4B | Yes | ~5 TB | ~3 TB | ~1.2 TB | ~9.5 TB |
With AES-256-GCM encryption (KMS enabled): add ~10% overhead to the RocksDB store column.
Vector backend decision guide
| Backend | RAM required | Disk required | Best for |
|---|---|---|---|
hnswlib (default) | High (full index in RAM) | Low | ≤ 5M nodes, fast queries, lots of RAM |
hnswlib + int8 | Medium (4× compression) | Low | 5M–25M nodes, good balance |
hnswlib + binary | Low (32× compression) | Low | 25M+ nodes, approximate recall acceptable |
diskann | Low (2 GB RAM) | High (NVMe required) | Any scale, memory-constrained hosts |
Set with:
bash
P8G_VECTOR_BACKEND=diskann # switch to DiskANN
P8G_VECTOR_QUANTIZATION=int8 # or int8 / binary for hnswlibFeature matrix by tier
| Feature | Desktop Free | Desktop Pro | Pro Cloud | Cloud Plus | Self-Hosted Server | Enterprise |
|---|---|---|---|---|---|---|
| Cypher / Search / Traversal | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Full-text (BM25) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Export / Import | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| ACID transactions | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| KMS encryption | ❌ | ✅ (local) | ✅ | ✅ | ✅ | ✅ |
| WAL durability | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| CDC / EventBus | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Journey engine | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Human-in-the-loop | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Connectors | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Graph algorithms | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| OLAP analytics engine | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Backup / restore | ❌ | Local | Managed | Managed + PITR | Self-managed | Self-managed |
| Read replicas | ❌ | Local | ❌ | ✅ | ✅ | ✅ |
| Raft HA | ❌ | ❌ | ❌ | ❌ | ✅ | ✅ |
| Sharding + federation | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ |
| RBAC | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ |
| SSO (OIDC/SAML) | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ |
| Audit trail | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ |
| VPC / on-prem deploy | ❌ | ❌ | ❌ | Add-on | ✅ | ✅ |
| Node limit | 1M | 10M | 25M | 100M | Unlimited | Unlimited |
| Queries/min | 1,000 | 5,000 | 5,000 | 20,000 | Unlimited | Unlimited |
| Support | Community | Priority | Priority | 24/7 SLA |