High-performance sync with Æoncase One order of magnitude faster than Dropbox in mixed workloads
The Æoncase Sync Appliance can be deployed behind your corporate firewall, allowing you to retain control over your data. Its performance and scalability thus depends ultimately on the hardware and environment where it is hosted. Æoncase Sync Appliance was designed to yield good sync speed atop even inexpensive, commodity hardware, yet what performance levels can be expected? In order to answer this question, we timed it against the popular Dropbox cloud-based sync service, which itself touts a 5x speed advantage against other solutions.
The Æoncase Sync Appliance will typically be used in your private network, with the ensuing speed and reliability advantage compared to cloud-based solutions. In order to negate this inherent edge, the Sync Appliance was installed on a server located in the Netherlands, and the client on a desktop computer in Germany.
Server (The Netherlands)$15/month DigitalOcean 2-core, 2GB RAM VM
40GB SSD Disk
Linux Centos 7.1
Client (Germany)4-core Intel i7-3770 3.40GHzM
2x3TB HD RAID-1
Furthermore, so as to calibrate the performance of the Sync Appliance on inexpensive hardware, the server consisted of a DigitalOcean VM with 2 GB RAM and only 2 virtual cores, priced at $15/month as of May 2019. Most computers acquired in the last 4 years, including the desktop computer used in the test, are more powerful, so in an actual deployment performance will be considerably better.
Sync performance is critical when working with large number of files — it makes the difference between getting the data you need to proceed with your work in the time it takes to go grab a coffee, or leaving something for the day after. The benchmark thus consisted in transferring 10000 files totaling 503 MB, with sizes following a Pareto distribution mimicking real-life patterns.
Æoncase Sync Appliance: 10x faster sync even on modest hardware
Even on very inexpensive, low-power hardware, the head-to-head comparison against Dropbox shows Æoncase Sync Appliance uploads data over 10x faster on average. The results for download speed are also telling: whereas Æoncase downloaded the files in less than 6 minutes, Dropbox needed over 54 minutes.
The difference is apparent in the resources needed by the client: whereas Dropbox needs over 8 minutes of CPU time when uploading and 13 minutes when downloading, the Æoncase client requires but 1’01” and 2’28” respectively, while using one fourth of the memory. Higher CPU usage not only slows down your other work on the computer, but moreover maps directly to increased power needs and battery drain.
Making sync fast
The Æoncase sync engine uses a number of techniques to speed up sync and let you make the most of your time.
Files are prioritized based on their nature and size, so that you will not have to wait for a bulky video to be transferred before a spreadsheet is synced.
Multiple-lane parallel transfer
The Sync Appliance is configured to transfer up to 20 files at once, and places transfers in “multiple lanes” to both optimize bandwidth usage and the median of the file transfer latency.
Concurrent upload and download with update reordering
Many sync engines perform separate download and upload cycles: they first compare what has changed since the last sync on each end, perform sync in one direction, then in the other. If something changes in between, it is not detected until the next cycle. The Æoncase sync engine syncs in both directions in parallel, and needs not wait for a cycle to end before syncing more changes. For example, if a file being uploaded is edited, the first upload will be canceled and the new version transferred right away. Likewise, if a synced file is edited in another computer while another is being uploaded, the engine needs not wait for the latter transfer to complete before downloading the former file.
Smart rename-aware reconciliation
The sync engine detects moves and renames, instead of registering them as a pair of file deletion and creation.
Duplicate content detection
The engine detects when a whole file (or file chunks when differential transfer/storage is activated) is already available in the receiving end, bypassing the transfer of the corresponding data. This both speeds up syncing and decreases bandwidth usage.