Re: OS X 10.8.2
Here are some rough comparative measurements of Time Machine/AFP throughput. Source machine is a modern iMac running OS X 10.8.2, connected through a Gigibit network to a modern Mac mini running 10.8.2 either as a stand-alone client or with Server 2.1.1. In Activity Monitor, I visually monitored the Network, getting a rough number for the average throughput. The Server software is installed on the mini's internal drive; the 10.8.2 client is installed on an external FW800 drive. The destination of the Time Machine backup is a freshly-erased 900 GB partition on another daisy-chained FW800 external drive. To switch targets, I merely restarted the mini using the correct startup volume, all other things being equal.
APF Target: OS X 10.8.2 (File Sharing Preference Panel):
1) Finder copy to a shared folder on the target's Startup Volume: 50 MB/sec
2) Finder copy to a folder on an external HD (connected to the target via FW800): 35 MB/sec
3) Time Machine: Create initial sparse bundle: 5 minutes.
4) Time Machine copy rate: < 5 MB/sec.
APF Target: OS X Server 10.8.2 (Server 2.1.1: Time Machine & File Sharing services enabled):
1) Finder copy to a shared folder on the target's Startup Volume: 100 MB/sec
2) Finder copy to a folder on an external HD (connected via FW800): 35 MB/sec
3) Time Machine: Create sparse bundle: 30 seconds
4) Time Machine copy rate: 35 MB/sec.
Notes:
1) Time Machine throughput to a 10.8.2 client is 5x slower than on Server 2.2.1.
2) Creating the sparse bundle takes 10x longer on 10.8.2 client than on 10.8.2 Server.
3) Time Machine throughput to a 10.8.2 client is 7x slower than a Finder copy to the same external volume on the same machine.
4) On 10.8.2 Server, AFP throughput to the Startup Volume is almost 3x faster than to an external FW 800 drive.
5) So worst-case Time Machine throughput to an external drive on a 10.8.2 client is 20x slower than the best-case Finder transfer to a startup-volume share on 10.8.2 Server.
Does anyone else see these throughput discrepencies?