DR 101: Synchronous, A-Synchronous and Near-Synchronous Replication Technologies
By Joshua Stenhouse, Zerto Technical Evangelist
When evaluating the most suitable disaster recovery technology to protect your virtual infrastructure, one major choice is the type of replication used to move the data. In this post I will explain each of the different types of data replication, and reveal some characteristics of each that you may not be aware of to help you choose the best replication technology for protecting and recovering your data.
Synchronous Replication
Synchronous replication ensures all data written in the source storage is simultaneously written in the target storage, and waits for acknowledgement from both storage arrays before completing the operation. This relies on matching storage between source and target with fibre channel latencies to minimize the performance overhead of the link between the storage arrays. Because of the potential for performance impact, synchronous replication should only ever be performed in the storage layer and not performed by a virtual appliance technology. For this reason I will only compare storage based synchronous replication technologies in this post.
A-Synchronous Replication
A-Synchronous replication does not write data to both the source and target storage simultaneously, it uses snapshots to take a point in time copy of the data that has changed and sends it to the recovery site on a schedule. The frequency is typically set on a schedule of hours, depending on the number and frequency of snapshots that the storage and application can withstand. A-synchronous replication can be performed by the storage array or by using a VM-level technology, but with storage based replication being the most predominant I will focus this type of replication for comparison.
Near-Synchronous Replication
Near-Synchronous replication is always-on and constantly replicating only the changed data to the recovery site within seconds. Because it is always-on it does not need to be scheduled, doesn’t use snapshots and writes to the source storage don’t have to wait for acknowledgement from the target storage. I will use the near-synchronous hypervisor-based replication of Zerto Virtual Replication for comparison.
To help you choose the most appropriate replication technology I created the below table to explain the features and limitations of each method:
|
Networking |
Synchronous Replication | A-Synchronous Replication | Near-Synchronous Replication |
| Replicate over any distance, avoiding regional disasters | – | ✓ | ✓ |
| Compressible replication traffic | – | ✓ | ✓ |
| Utilize cheaper IP links including VPNs | – | ✓ | ✓ |
|
Data Loss & Recovery |
Synchronous Replication | A-Synchronous Replication | Near-Synchronous Replication |
| *Data loss (disk & in-memory data) | – | – | – |
| *Data loss on disk writes | ✓ | – | – |
| Seconds of data loss on disk writes | – | – | ✓ |
| Seconds of data loss of all data (disk & in-memory)* | – | – | ✓ |
| Data corruptions immediately written to target | ✓ | – | – |
| Point in time recovery to increments in seconds | – | – | ✓ |
|
Performance & Snapshots |
Synchronous Replication | A-Synchronous Replication | Near-Synchronous Replication |
| Always-on protection with no scheduling overheads | ✓ | – | ✓ |
| No site link performance overhead on writes | – | ✓ | ✓ |
| Snapshots for point in time recovery | ✓ | ✓ | – |
| No performance impact of snapshots | – | – | ✓ |
| No storage overhead for snapshots | – | – | ✓ |
| Application consistency without snapshots | – | – | ✓ |
| Point in time recovery to increments in seconds | – | – | ✓ |
*Subject to the frequency on which the application can quiesce writes to disk without the overhead of utilizing snapshots.
As you can see, each data replication technology has different attributes and depending on the requirements of your workloads and service level agreements you can use the replication method that best meets your requirements.
Become even more knowledgeable with our Disaster Recovery Essential Guide!
