The new kid on the Software-Defined storage block and the question you ask yourself is of course: What differentiates it from other (Software-Defined) storage solutions. First of all, it’s built on a decade of experience at Facebook and Amazon. The key architect, Avinash Lakshman, has built the well-known open source database management solution Cassandra at Facebook and co-authored Dynamo at Amazon. It’s 100% software based and embraces web-scale principles which include high performance and security through distributed compute, storage and network. Data can be replicated easily thanks to its in-the-box multi-datacenter support what provides cloud-like availability. The power of the web-scale approach is using distribution as foundation for scalability and availability.
- Hyper-Converged: Compute and storage together in a single server node with interlocked scalability. Provides huge performance and cost benefits.
- Hyper-Scale: Compute and storage at different server nodes, scale separately in capacity and performance.
IT is free to scale at the performance and cost level needed. The business can profit from the performance and cost benefits of Hyper-converged, or scale independently and save on licensing cost (hypervisor licencies) when there is only need for storage. All the magic runs in a docker container and this is the reason why Hedvig runs on virtually every hypervisor (VMware, Hyper-V, Linux KVM/Xen) as well as baremetal server deployment scenario’s.
Web-scale trusts on distribution of everything to ensure availability instead of using special, mostly costly, hardware. The cluster balances itself automatically when a disk or server fail occurs, in order to keep data secure and protected. This is not limited to servers or disks, it also applies to rack and even datacenter failures. We call this build-to-fail and it is the same technology as Amazon, Facebook and Google use to guarantee their availability.
Hedvig supports most popular storage protocols. Protocols can be set specific to each vDisk or LUN. No other storage platform provides this level of freedom in one solution.
- File: NFS v2, v3 en v4. SMB3 support is in development (expected later this year)
- Block: iSCSI
- Object: Openstack Swift and Amazon S3
Virtual disk features
Hedvig enables you to set features and protocol support for each separate virtual disk.
Features as read cache, inline deduplication and compression can be set per vDisk. Also, virtual disks can be created in batch and managed through policies.
You can balance protection and performance by adjusting the replication policy. This enables you to control the way data is protected versus the performance penalty you will get. Set the number of replicas you want per vDisk (up to 6 replicas) and the number of commits you want the cluster to wait for minimizing impact on performance. This ensures that data is distributed to different servers, racks and data centers based on your protection needs.
Performance is boosted by making use of a number of smart innovations. RAM is being used to handle intensive IO tasks such as deduplication and compression. Random IO is made sequential in RAM before it’s written to optimize disk performance. Flash or solid state are being used to enable storage-side or server-side caching. All writes are distributed over the cluster as a whole, wherein the application is waiting for a majority of write confirmations instead of waiting for all. These optimizations ensure great performance, despite a comprehensive feature set.
Wat is Server-side caching?
Server-side storage proxies handle file and block storage towards the vDisks using local server cache (Tier 1: Ram, Tier 2: Flash or SSD) to boost performance. Features like deduplication are also handled at the server-side further improving storage performance.