Restore performance using API only: what is the usual performance ?

[kristofpoppe]

Hi all,

We perform restores single and only by using API calls. We see that the restore from the datastore to the %temp% location is fast, however, offering the datastream to the API consumer is slow.
Of course we have tested underlying infrastructure: TCP/IP stack, disk and network can handle much larger amounts of data per second.

So we suspect that the internal API process is quite slow in streaming the data from the TEMP location. Is this known or can we tweak this somehow ?
Analysis shows that with such a stream there are regular CPU spikes (we suspect CPU cycles for encryption) but these are not of the nature that they would slow down the overall download that much.

Has anyone succeeded in getting a restore of several gigabytes through an API stream at a decent speed on his client ?

Thanks for sharing your insights.

[Mike Resseler]

Kristof,

Restoring the data from API's should be using the same method as through the UI. However, the speed of restore, especially with several gigabytes could take indeed while. What speeds are you seeing?

[kristofpoppe]

Hi Mike, thanks for your response.

We see very low speeds, between 0.5 and 1 Mbps although the server has enough CPU power, disk I/O and bandwidth to export data much faster to the client.
A restore of a 1.5Gb PST mailbox takes more than 2 hours.