2 posts • joined Wednesday 13th March 2013 14:02 GMT
Inline solutions have their limitations
I think most people see through the FUD between inline and concurrent deduplication. Inline solutions are not appropriate for large enterprises and suffer their own drawbacks:
1. There is very little duplicated data within a single backup. Therefore, almost every appliance has to land virtually the whole first backup, regardless of the type of deduplication technology.
2. Inline deduplication requires very large chunks on the incoming data to be able to perform at the rates they do. This results in poor deduplication unless post-process steps are performed after all the data has landed.
3. To get competitive deduplication ratios, inline solutions need to run post-process optimization to find more duplicate data and to reclaim storage found during deduplication back to the available storage pool. This takes a lot time, causes penalizing fragmentation, and must happen post-process in order to get the results they claim and to be able to use reclaimed space for the next backup.
4. Inline systems don't scale and performance degrades over time as more data is stored on them. These systems become useless long before they reach their claimed maximum capacity specifications.
For the reasons above, anyone looking to backup a lot of data should seriously consider other solutions than inline systems.
It only makes sense to compare systems by how much performance you can have in a single dedupe domain and management pane. The equivalent four domain SEPATON system will yield 320TB/hr of backup. If you are going to gang separate systems together, then where do the comparisons end -- after 4 systems, 10 systems, 1000 systems, or maybe by the number of processing nodes, FC interfaces...