Accelerate Your Data on Google Cloud Platform

Policy-based NVMe-oF storage with validated sub-millisecond latency. Choose from 5 performance tiers optimized for your workload. Deploy in minutes with Terraform.

138µs
QD1 Read Latency
2M+
Read IOPS (Ultra)
View Performance Tiers Contact Sales
MayaScale on GCP

Validated Performance Tiers

Choose the right tier for your workload. All tiers deliver sub-millisecond latency with Active-Active HA.

Why MayaScale on Google Cloud Platform?

10x Faster Than Persistent Disk

Sub-millisecond latency using local NVMe from N2 and Z3 instances. Pool local SSDs for shared, high-performance block storage.

Active-Active HA

Both nodes serve I/O simultaneously. Sub-second failover with no downtime. Synchronous replication across zones.

90% vs Traditional SAN

No expensive SAN infrastructure required. Pool local NVMe from GCE instances. $10K starting price vs $200K+ traditional SAN solutions.

Native Google Cloud Integration

Compute Engine Local SSDs

Runs on N2, N2D, and Z3 instances with local NVMe SSDs. Pool local SSD storage for persistent shared volumes.

VPC Networking

Leverages GCP's high-bandwidth networking (50-100 Gbps) for maximum NVMe-oF throughput with minimal latency overhead.

Cloud SQL Migration

Move databases from Cloud SQL to GCE for 10x better performance. Compatible with PostgreSQL, MySQL, SQL Server.

Google Kubernetes Engine

High-performance persistent volumes for stateful applications. CSI driver for dynamic NVMe-oF volume provisioning.

Vertex AI

Fast storage for AI/ML training workloads. Low-latency data loading for GPU and TPU instances.

Cloud Monitoring

Full integration with Cloud Monitoring, Logging, and alerting for performance tracking and troubleshooting.

NVMe-over-Fabrics Architecture

Pool Local SSD Storage

MayaScale transforms local SSD storage into persistent shared volumes using NVMe-over-Fabrics:

  • Local NVMe SSDs: N2, N2D, Z3 instances with local SSD storage
  • NVMe-oF Protocol: NVMe-over-TCP for cloud-native shared storage
  • Active-Active Clustering: Both nodes serve I/O concurrently
  • Cross-Zone Replication: Synchronous replication across zones
  • Standard iSCSI Support: Also supports iSCSI initiators for legacy apps

High Availability

  • Active-Active Design: Both nodes handle I/O simultaneously
  • Sub-Second Failover: Automatic failover in <1 second
  • Multi-Zone Deployment: Deploy across zones for resilience
  • No Shared Storage Dependency: Cloud-native architecture
Active-Active HA Deployment
MayaScale Architecture on Google Cloud

Dual NIC architecture with local NVMe SSDs for sub-millisecond latency. Both nodes actively serve I/O with synchronous RAID-1 replication across zones.

Perfect For These GCP Workloads

Database Migration from Cloud SQL

Move from Cloud SQL to GCE for 10x better performance and cost savings:

  • PostgreSQL, MySQL, SQL Server on GCE
  • Active-Active HA for 99.99% uptime
  • Sub-millisecond latency for OLTP workloads
  • Reduce database licensing costs with better perf/instance

NoSQL & BigQuery

Cassandra, MongoDB, ScyllaDB with ultra-low latency, plus BigQuery staging:

  • Sub-millisecond reads for real-time queries
  • High IOPS for distributed databases
  • Fast data staging for BigQuery loads
  • Dataflow and Dataproc integration

AI/ML & Vertex AI

High-performance storage for Vertex AI training and GPU workloads:

  • Fast data loading for AI training pipelines
  • Shared datasets across multiple training jobs
  • Low-latency inference with A2/A3 GPU instances
  • Cost-effective storage for large model checkpoints

GKE Stateful Workloads

Persistent volumes for Kubernetes StatefulSets with high performance:

  • CSI driver for dynamic volume provisioning
  • Sub-millisecond latency for database pods
  • Active-Active HA for production workloads
  • VolumeSnapshots for backup and recovery

Deployment Options

GCP Marketplace

One-click deployment with Deployment Manager.

  • Automatic billing through GCP account
  • Pre-configured Active-Active HA setup
  • Integrated with Cloud Console
  • Production-ready in under 15 minutes
Deploy from Marketplace

Terraform Deployment

Infrastructure-as-Code for automation.

  • Full control over configuration
  • Version-controlled infrastructure
  • Deploy to multiple regions/projects
  • GitOps-ready workflows
Get Terraform Modules

Performance Tiers - Validated Specifications

All tiers tested and validated on GCP with SNIA-compliant FIO benchmarks. October 2025.

Tier Instance vCPUs Local SSDs Capacity Read IOPS Write IOPS Latency Use Case
Ultra n2-highcpu-64 64 16x 375GB 6.0 TB 2.3M 866K 173µs Maximum performance databases, ML training, analytics
High n2-highcpu-32 32 8x 375GB 3.0 TB 900K 350K 830µs High-performance databases, enterprise applications
Medium n2-highcpu-16 16 4x 375GB 1.5 TB 700K 200K 822µs Production databases, analytics workloads
Standard n2-highcpu-8 8 2x 375GB 750 GB 380K 130K 938µs General purpose databases, applications
Basic n2-highcpu-4 4 1x 375GB 375 GB 100K 75K 630µs Development, testing, budget deployments

Note: All latency numbers are sub-millisecond (<1ms). IOPS are "Promised" marketing numbers backed by validated test results. Actual SLA performance typically exceeds promised values by 4-16%. See performance graphs above for detailed curves.

Ready to Deploy Sub-Millisecond Storage on GCP?

Choose your performance tier and deploy with Terraform in minutes. Get validated sub-millisecond latency and up to 2.3M IOPS with Active-Active HA.

View Performance Tiers Contact Sales

Frequently Asked Questions

How does MayaScale compare to Persistent Disk SSD?

MayaScale delivers 5-10x lower latency (sub-1ms vs 5-10ms) using local NVMe storage with NVMe-oF. Plus, you get Active-Active HA which Persistent Disk doesn't provide.

What happens to my data if a GCE instance is terminated?

MayaScale uses synchronous replication across nodes (typically in different zones). If one node is terminated, the other node continues serving I/O with sub-second failover. No data loss.

Can I migrate from Cloud SQL to MayaScale?

Yes. Many customers migrate from Cloud SQL to self-managed databases on GCE with MayaScale storage for 10x better performance and lower costs.

What GCP regions is MayaScale available in?

MayaScale can be deployed in any GCP region that supports instances with local SSD storage (N2, N2D, Z3 families). Contact sales for region-specific availability.

How much does MayaScale cost?

MayaScale starts at $10K vs $200K+ for traditional SAN solutions—a 90% cost reduction. You only pay for the GCE instances and local SSD storage you use.

Is there a free trial available?

Yes. Contact our sales team to arrange a proof-of-concept deployment on your GCP project. We offer technical support during evaluation.