Share this post

October 14, 2015

USX 3.0 What’s New? – Volume-Level Snapshots

Hugo Phan - Atlantis Computing

Avatar

In this article I’ll discuss the new volume-level snapshot functionality in Atlantis USX 3.0.

Why Atlantis Snapshot?

Atlantis volume-level snapshots provide instantaneous copy images of entire volumes, which is an exact replica of the volume and its contents at the point when the snapshot was created.

Atlantis Snapshots allows for unlimited snapshots of high performance, low latency snapshots. Snapshots can both be initiated manually through the USX GUI, through the RestAPI, using the Atlantis PowerShell Tools or scheduled.

Atlantis snapshots are extremely efficient in metadata and physical space, are created instantaneously, have exactly the same data services as any other volumes, such as inline data deduplication, and have no performance impact.

Use Cases

Atlantis volume-level snapshots can be used for many applications, including:

  • Backup
  • Near continuous data protection
  • Development and Test
  • Offload processing
  • Disaster Recovery

Atlantis snapshots provide the ease of use necessary for large software defined storage environments including hyperconverged environments where the above use cases are required.

Atlantis USX Snapshots

Atlantis USX Snapshot is Atlantis’ snapshot technology implemented in all of the Atlantis products – USX and HyperScale.

Atlantis Snapshot is fast, flexible, efficient and easy, this is due to the fact that Atlantis Snapshot is deduplication aware. Yes, dedupe aware means that you can take unlimited numbers of volume snapshots without incurring additional storage capacity.

Atlantis Snapshot implements Redirect-on-Write (RoW) with a few Atlantis enhancements. Snapshot utilizing Atlantis’ implementation of RoW is efficient and results in at least 50% less write activity on the backend storage compared to the traditional Copy-on-Write (CoW) implementations of other technologies.

Lets look at this in more detail for a volume level snapshot:

When a snapshot is taken, an exact copy of the metadata is copied and points to the original data blocks. Both the original metadata and data blocks are preserved.

When a new write or a write update occurs to an existing block b1, the new write or write update goes to a separate snapshot space followed by an update to the current metadata to point to the new block location b1` in the new snapshot space. The original metadata and the original data are not touched.

Snapshots are deduplication aware in that if an incoming write is to a block that already exists in the original data or in the snapshot space then no write update is needed to the volume or the snapshot space. Only a metadata update is required to the current metadata. This results in not only fewer writes (due to dedupe) but also reduces snapshot overhead (no reading, update and writing of blocks are necessary).

The figure below shows the snapshot stages at three different times that we’ve discussed above:

  • T1. Initial volume snapshot taken
  • T2. A write update
  • T3. Two new deduplicated writes to blocks b2 and b3.

Comparison of Snapshot Technologies

  Copy-on-Write Full Clones Atlantis Snapshot
Creation time Instantaneous Duration of clone copy Instantaneous
Space Efficient Data Yes No Yes + Inline Dedupe
Space Efficient Metadata No No Yes
Metadata and snapshot location Metadata from disk and memory Metadata from disk and memory 100% working set in memory with backing to persistent media
Performance of snapshot Degraded Same as volume w/o snapshot Same as volume w/o snapshot
Performance impact High impact No impact after clone operation complete, high impact during clone operation No impact
Fast snapshot deletion No No Yes
Limitations of delete operation Limited n/a Any branch snapshot can be deleted from the tree w/o impact
Impact on up-front space reservation Yes Yes No
Impact on data services Yes No No

For more background information please read this article by my colleague Priyadarshi Prasad.

Atlantis Snapshot In Action

Lets take a look at Atlantis volume-level snapshots in action. The video demonstrates how easy and of course fast it is to take a volume-level snapshot of the entire volume, delete a virtual machine residing inside the volume and then restoring a snapshot and re-registering the deleted VM.

About the Author

Hugo Phan is a Technical Marketing Architect at Atlantis based in Mountain View, California. He is a technical evangelist for Atlantis’ 100% software defined storage solution USX and hyperconverged appliance HyperScale which provides high-performance storage and advanced data services that augments any storage using federated in-memory technology on the hypervisor layer. A vExpert, he has been active in the virtualization and cloud industries for over 10 years and has held senior technical roles at Canopy and VMware. He has spoken at User Groups, VMworld events and VMware Partner Exchange and is also a qualified VCDX Panelist. As well as holding the VCP, VCAP and VCDX certifications from VMware, Hugo is also certified in Microsoft, NetApp and ITIL certifications. He is also a Bachelor of Computing and Mathematical Sciences.

You can follow Hugo’s at community.atlantiscomputing.com/blog, www.vmwire.com and also on www.twitter.com/hugophan.

12345
Current rating: 0 (0 ratings)