[p4] Perforce roadmap?
Patrick M. Slattery
patrickmslattery at mac.com
Mon Oct 6 07:17:28 PDT 2008
You should consider a storage array that handles data duplication in
primary storage. That would alleviate the amount of storage you need,
we use NetApp ASIS (Advanced Single Instance Storage) today on the
array that stores our ESX VMs and Perforce data. I get up to 93% data
reduction on some VM volumes, I currently only get about 20% data
reduction on the Perforce volumes but that effectively allows me to
have my live and sim servers using the same disk blocks.
The are several other dedupe vendors out there, but currently most
major storage array vendors restrict dedupe to backup tasks. Give it 5
years and every halfway decent array will have it.
On Oct 5, 2008, at 5:53 PM, Frank Compagner wrote:
> Are you sure that is really what you need? The binary diff's will
> save you storage space, yes, but the cost to performance is going
> to be significant. Doing a binary diff on a 10mb (or considerably
> larger, we have lots of those) file will consume a lot of server
> resources every time a new file is submitted, or somebody requests
> a version other than the head revision from the depot.
> It depends on your actual usage patterns, but in our experience
> adding more storage is easier and cheaper than improving the server
> performance by putting in faster processors and more memory. We
> currently have about 5TB of data in the depot for a single project
> without problems, and we even have considerable amounts of the data
> set to uncompressed to improve performance. The problems we do have
> are all related to server performance and I would think long and hard
> before increasing the server load to save some storage space.
> Frank Compagner Guerrilla Games
> RC> Hi
> RC> Does anyone know if Perforce has a public roadmap of features?
> RC> I am very interested to know if it will support storing of
> binary diffs
> RC> any time soon, as I need to develop a binary data repository and
> RC> like to use a SCM system as foundation, but the amount of data
> will too
> RC> large if the SCM chosen doesn't support binary diffs. (I'll
> quickly run
> RC> out of disk space if every time someone changes 10 bytes in a
> 10mb file
> RC> the system stores another full copy...)
> RC> Cheers,
> RC> Rui
> perforce-user mailing list - perforce-user at perforce.com
More information about the perforce-user