[tahoe-lafs-trac-stream] [Tahoe-LAFS] #235: scale up to many nodes

Tahoe-LAFS trac at tahoe-lafs.org
Fri Jan 17 23:54:04 UTC 2020


#235: scale up to many nodes
-------------------------+-------------------------------------------------
     Reporter:  zooko    |      Owner:
         Type:           |     Status:  new
  enhancement            |
     Priority:  major    |  Milestone:  undecided
    Component:  code-    |    Version:  0.7.0
  network                |   Keywords:  scalability memory performance
   Resolution:           |  openssl leastauthority
Launchpad Bug:           |
-------------------------+-------------------------------------------------
Changes (by exarkun):

 * keywords:  scalability memory performance openssl => scalability memory
     performance openssl leastauthority


Old description:

> I updated [wiki:UseCases The UseCases Page] to reflect that someone might
> want to run a managed Tahoe grid comprising one thousand nodes.  (If each
> node has a single 1 TB hard drive, that's a 1 PB grid.  Obviously there
> are lots of other options, such as each node having six 1 TB hard drives
> in a RAID-6 configuration, resulting in 4 usable TB per node or a 4 PB
> grid.)
>
> Anyway, we expect that the current Tahoe grid would have problems
> handling more simultaneously connected nodes.  One known problem is that
> pyOpenSSL uses almost 1 MB of RAM per SSL connection.  (See also #11.)
>
> This ticket can be closed when Tahoe is demonstrated to handle one
> thousand simultaneously connected nodes smoothly.

New description:

 I updated [wiki:UseCases The UseCases Page] to reflect that someone might
 want to run a managed Tahoe grid comprising one thousand nodes.  (If each
 node has a single 1 TB hard drive, that's a 1 PB grid.  Obviously there
 are lots of other options, such as each node having six 1 TB hard drives
 in a RAID-6 configuration, resulting in 4 usable TB per node or a 4 PB
 grid.)

 Anyway, we expect that the current Tahoe grid would have problems handling
 more simultaneously connected nodes.  One known problem is that pyOpenSSL
 uses almost 1 MB of RAM per SSL connection.  (See also #11.)

 This ticket can be closed when Tahoe is demonstrated to handle one
 thousand simultaneously connected nodes smoothly.

--

--
Ticket URL: <https://tahoe-lafs.org/trac/tahoe-lafs/ticket/235#comment:6>
Tahoe-LAFS <https://Tahoe-LAFS.org>
secure decentralized storage


More information about the tahoe-lafs-trac-stream mailing list