April was one of the biggest engineering cycles for Akave to date, bringing major upgrades across both Akave O3 and the Akave protocol stack. Over the course of the release train, we shipped O3 v1.6.0, v1.6.1, and v1.6.2 alongside protocol releases v0.5.0 through v0.5.5.
This cycle focused heavily on scalability, operational simplicity, memory efficiency, multipart upload reliability, observability, transaction handling, and next-generation streaming upload/download paths for decentralized storage.
On the O3 side, we introduced a completely reworked multipart upload architecture, a new console and IAM deployment model, improved internal DB indexing and repository flows, dynamic IAM configuration, in-memory metadata caching, stronger AWS-compatible IAM validation, and large internal refactors around async I/O, storage handling, and buffer management.
On the protocol side, we delivered transaction loop v2 for handling zombie and undelivered blockchain transactions, multi-file upload support, experimental Upload2 and Download2 streaming pipelines, striped erasure coding, streaming encryption/decryption, enhanced IPC and indexing observability, reduced SDK allocations, Pebble-backed metadata storage improvements, and major optimizations for node-side block transfer efficiency.
Together, these updates push Akave significantly closer to becoming a production-grade decentralized S3 infrastructure layer capable of supporting enterprise-scale AI, analytics, and high-throughput storage workloads.
About Akave O3
Akave O3 is the S3-compatible object storage layer built on top of the Akave decentralized protocol stack. Every release continues to focus on three core areas:
“Security” – “Performance” – “Operability”
April heavily expanded all three through infrastructure refactors, improved AWS compatibility, reduced operational complexity, and significantly more efficient data transfer paths.
Security and Access Control
AWS-compatible IAM enforcement
One of the major themes throughout the O3 v1.6.x releases was tightening IAM compatibility and improving policy correctness.
S3 operation names were normalized to AWS IAM action semantics, ensuring bucket policies and IAM evaluations behave much closer to native AWS S3 behavior. This includes support for version-aware IAM action handling and stricter policy evaluation flows.
IAM validator logic was centralized and hardened to fail fast during configuration loading, preventing invalid policies or malformed IAM definitions from reaching runtime. ARN handling was also normalized to canonical arn:aws:iam formatting for consistency across services and integrations.
Additional IAM improvements included:
- Dynamic IAM enablement support
- IAM and Console service toggles in generated compose setups
- Migration tooling for legacy user.keys via ctrl iam migrate-classic-keys
- Stable owner/user identity handling instead of relying on access key identifiers
- Root-user restrictions and improved access-key lifecycle handling
- Access key blocking and unblocking with optional metadata and reason tracking
ACL and bucket permission handling
ACL and bucket-level authorization flows were significantly expanded across multipart uploads and object operations.
This included:
- Bucket-level ACL routing improvements
- Permission enforcement for multipart upload APIs
- CanonicalUser principal support
- Source-object permission verification during CopyObject operations
- Improved object-lock validation and legal-hold enforcement
- Stronger version-aware deletion controls
Delete marker and object lifecycle behavior was also improved to align more closely with AWS S3 semantics.
Performance and Runtime Efficiency
Multipart upload architecture overhaul
O3 v1.6.0 introduced one of the largest multipart upload refactors in the platform so far.
Legacy uploadSession implementations were removed and replaced with new PartedStore and PartedFile abstractions designed specifically for scalable multipart file handling.
The new MPU architecture includes:
- JSON-backed multipart metadata
- Layout-aware part tracking
- Lazy metadata loading
- Paginated multipart listing support
- Cleaner lifecycle management
- Reduced internal contention
- Atomic state handling for multipart tracking
Additional fixes throughout v1.6.1 and v1.6.2 improved multipart correctness even further:
- Invalid chunks are now skipped safely during multipart metadata generation
- Chunk trailer handling was simplified
- File move operations were streamlined
- Metadata publication is prevented after AIO rename failures
- Upload path caching and deletion handling were stabilized
The result is a far more predictable and scalable multipart upload pipeline that better supports enterprise workloads and large-scale object transfers.
Async I/O and storage optimization
The async I/O stack also continued evolving throughout the April cycle.
Major improvements included:
- Better io_uring fallback handling
- Dynamic ring resizing retries
- Sharded io_uring ring models
- Improved buffer pooling
- Reduced allocation overhead
- Safer syscall-based file movement
- Better TCP keepalive behavior
- Simplified storage-layer locking and state handling
Several older buffering and pipe-related implementations were completely removed to reduce internal complexity and eliminate dead code paths.
Additional AIO refinements introduced:
- Safer writer timeout handling
- Cleaner file closer logic
- More predictable ranged IO behavior
- Reduced redundant conversions in CopyBuffer flows
- Improved local and ranged file handling
DB and metadata optimization
April also introduced major changes around metadata storage and repository efficiency.
DB entity definitions and automatic index reconciliation were expanded significantly across:
- Buckets
- ACLs
- Encryption metadata
- Lifecycle policies
- Multipart uploads
- Multipart parts
- Access keys
O3 v1.6.2 introduced repository-level caching for bucket and object metadata using a simpler in-process caching layer built around MapTTL instead of the older ristretto dependency.
This reduced lookup overhead while simplifying the overall caching architecture.
Additional database and metadata improvements included:
- DB ping validation during startup
- 15-second timeout enforcement for database operations
- BSON hot-path optimizations using bson.D
- Reduced sequential object lookups
- Better handling of dropped and recreated indexes
- Correct handling of missing latest revisions in sync flows
Enterprise Deployment and Operability
Console and IAM deployment model
April significantly expanded the new console deployment architecture introduced in earlier releases.
O3 v1.6.0 added full deployment support for:
- Console services
- JWT authentication
- Refresh-token flows
- IAM integration
- User management
- Wallet balance APIs
- Health services
- Bucket listing APIs
- Sync-control APIs
- Job-state management
Compose generation and deployment tooling also became far more flexible.
New deployment configuration options include:
- o3volume
- datavolume
- mpuvolume
- --domain
- --consoledomain
- --s3gateway
- --s3host
- --iamhost
- --consolehost
- --traefikhost
Generated compose flows now support dynamic service toggles and clearer internal routing behavior through Traefik integration.
Operational simplification
Several deprecated migration and metadata repair utilities were removed entirely to simplify long-term operations and reduce maintenance burden.
Removed components included:
- Legacy migration commands
- objects-meta repair flows
- Old S3 console modules
- Unused archival/cache paths
- Redundant cache wrappers
- Legacy upload session handling
At the same time, deployment reliability improved through:
- Better non-root Docker execution support
- More robust liburing installation handling
- Improved environment variable management
- Cleaner token replacement flows
- Better GitHub raw-template fetching in setup scripts
Protocol Layer Upgrades
Akave Protocol v0.5.x
The protocol stack received major upgrades across transaction handling, SDK performance, encryption, erasure coding, observability, and transfer efficiency.
Transaction loop v2
One of the most important infrastructure upgrades in v0.5.0 was the introduction of transaction loop v2.
The new transaction system was specifically designed to better handle:
- Zombie transactions
- Undelivered blockchain transactions
- Nonce conflicts
- Network asynchrony between nodes and blockchain state
The updated loop simplified transaction management while significantly improving reliability during high-throughput upload scenarios.
Several memory leaks and monitoring span leaks inside the transaction loop were also fixed throughout the cycle.
SDK memory and transfer optimization
A major focus throughout v0.5.x was reducing SDK allocation overhead and improving transfer efficiency.
Improvements included:
- Reduced allocations in encryption paths
- Reduced allocations in erasure coding
- In-place decryption support
- Preallocated decode buffers
- Reduced connection-pool setup overhead
- Tuned gRPC dial defaults
- Faster DAG-PB payload extraction
- Direct protobuf parsing
- Streaming encryption/decryption
- Striped erasure coding utilities
Experimental Upload2 and Download2
One of the most important additions in v0.5.5 was the introduction of experimental Upload2 and Download2.
These new transfer paths introduce:
- Streaming encryption
- Streaming decryption
- Striped erasure coding
- Improved chunk handling
- Lower memory consumption
- Better throughput scaling
- More efficient reconstruction logic
This architecture represents a major step toward highly scalable decentralized data transfer pipelines optimized for large AI and analytics workloads.
Node and Network optimization
The node stack also received significant upgrades.
Major improvements included:
- Pebble-backed metadata state standardization
- Removal of legacy Badger metadata storage
- Generic metadata caching for hot read paths
- Cached validation results
- Improved block upload allocation behavior
- Streaming block downloads directly from FlatFS
- Better gRPC buffer sizing alignment
- More efficient block transfer handling
Several new observability systems were also introduced.
Monitoring upgrades included:
- Node transfer metrics
- IPC throughput metrics
- Indexer metrics
- Download observability
- Heap and live-node monitoring
- Blockchain activity dashboards
- Prometheus integration for internal spans
- Grafana dashboard expansion
These additions provide much deeper visibility into decentralized storage performance and node behavior.
Reliability and correctness
The protocol releases also included a large number of reliability improvements across IPC, indexer, PDP, and CI systems.
Fixes included:
- Pebble DB error handling corrections
- PDP processing reliability improvements
- Flaky IPC test fixes
- Indexer synchronization timeout handling
- Improved Anvil lifecycle management in tests
- More predictable transaction release behavior
- Better corruption detection during reconstruction
Together, these changes substantially improve node stability and long-running transfer reliability.
Release Recap
O3 v1.6.0
A foundational infrastructure and deployment release.
Key highlights:
- PartedStore and PartedFile multipart architecture
- Console and IAM service deployment
- DB entity/index management
- Improved deployment tooling and Traefik support
- Multipart lifecycle refactor
- Better async I/O handling
- Expanded AWS-compatible IAM behavior
O3 v1.6.1
A runtime hardening and simplification release.
Key highlights:
- Improved AIO concurrency handling
- Custom hasher pools
- Safer storage close behavior
- Simplified multipart internals
- Removal of redundant cache layers
- Better checksum handling
O3 v1.6.2
A caching, validation, and metadata optimization release.
Key highlights:
- MapTTL in-process caching
- Repository metadata caching
- Dynamic IAM enablement
- Canonical IAM ARN handling
- Improved delete/caching behavior
- Multipart metadata correctness
- Centralized validator logic
Akave Protocol v0.5.0
A major SDK and transaction architecture release.
Key highlights:
- Transaction loop v2
- Multi-file upload support
- Improved SDK buffering controls
- Reduced memory allocations
- Retry handling for PDP downloads
- Monitoring and transaction reliability improvements
Akave Protocol v0.5.1 - v0.5.3
Focused protocol optimization and observability releases.
Key highlights:
- Pebble DB support
- Reduced upload allocations
- Expanded monitoring
- Improved metadata storage efficiency
- New Grafana and Prometheus metrics
Akave Protocol v0.5.5
A major transfer and streaming release.
Key highlights:
- Experimental Upload2 and Download2
- Streaming encryption/decryption
- Striped erasure coding
- Improved DAG-PB extraction
- Better gRPC memory efficiency
- Enhanced transfer observability
- Node metadata caching
What’s Next
Streaming-first transfers
The Upload2 and Download2 architecture lays the foundation for future high-throughput transfer systems optimized for AI datasets and analytics pipelines.
Future work includes:
- Full productionization of Upload2/Download2
- Additional streaming optimizations
- Adaptive erasure coding
- Smarter transfer scheduling
- Larger-scale throughput tuning
Console and IAM expansion
The console platform will continue evolving into the primary operational control plane for Akave deployments.
Upcoming areas of focus include:
- Expanded IAM policy management
- Tenant and project scoping
- Better observability dashboards
- Credential lifecycle management
- PDP dataset visibility
- Node fleet management
Storage verification and PDP tooling
Future PDP improvements will focus on:
- Dataset lifecycle visibility
- Repair workflows
- Verification dashboards
- Better node-side repair automation
- Richer proof and retrieval tooling
Connect with Us
Akave Cloud is an enterprise-grade, distributed and scalable object storage platform designed for AI, analytics, decentralized infrastructure, and large-scale enterprise workloads.
It provides:
- S3 compatibility
- Cryptographic verifiability
- Immutable auditability
- SDKs for agentic AI systems
- Zero egress fees
- No vendor lock-in
- Up to 80% lower storage costs compared to hyperscalers
Akave works with a broad ecosystem of infrastructure providers operating across multiple regions and countries, enabling sovereign and distributed data infrastructure at scale.