From mboxrd@z Thu Jan 1 00:00:00 1970 MIME-Version: 1.0 In-Reply-To: References: <10af1064d3ac35a8d2f62214d5eec485@gmx.de> Date: Fri, 15 Oct 2010 12:26:00 -0500 Message-ID: From: Eric Van Hensbergen To: Fans of the OS Plan 9 from Bell Labs <9fans@9fans.net> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Subject: Re: [9fans] =?iso-8859-7?q?=F0p?= Topicbox-Message-UUID: 685fb8e6-ead6-11e9-9d60-3106f5b1d025 2010/10/15 ron minnich : > 2010/10/15 =A0: >> i wonder if making 9p work better over high latency connections is >> even the right answer to the problem. > > The reason I care is that the link from a CPU node to a file server on > blue gene is high latency. It might as well be cross-country, it's so > bad. > a mere 6 GB/s (at the bottleneck) for folks who don't understand what Ron's idea of bad is :) In comparison to the torus in both latency and bandwidth (which with aggregate bandwidth of all links could be almost an order of magnitude higher bandwidth), the path to persistent storage is quite bad. As a point of clarification, its not actually the link that is currently the main problem, its the software infrastructure we currently have deployed (since we are opting not to use the production Blue Gene file servers/service). Although with our desktop extension model which uses your laptop (or whatever) as part of the file server namespace -- latency can be quite long (in fact in many cases it is across the country, and sometimes in a different one). -eric