Alfonso, You are correct in the fact that the *architecture* of a Lustre file system greatly influences the performance - both good and bad. That said, the MGS is very low in terms of resources used or required. So, rather than thinking of splitting the MGS and MDS in terms of performance, it is better to think of splitting them in terms of overall functionality within the Lustre file system - and in that regard (features, etc.) it is better to have the MGS on one node and the MDS on another node. As for architecting to get a high level of performance, please see: http://www.intel.com/content/www/us/en/software/architecting-a-high-perfo... [Description: cid:image001.gif@01C9FE3D.1D8A68C0] Dr. Brett Lee, Solutions Architect High Performance Data Division, Intel +1.303.625.3595 From: HPDD-discuss [mailto:hpdd-discuss-bounces@lists.01.org] On Behalf Of Alfonso Pardo Sent: Wednesday, March 19, 2014 8:36 AM To: hpdd-discuss(a)lists.01.org Subject: [HPDD-discuss] performance split mds mgs Hello, I must implement a new lustre file system and optimize it to get a high level of performance. I wonder: can I get more performance, If I split the MDS and the MGS in two different machines? Thanks!!! Alfonso Pardo Diaz System Administrator / Researcher c/ Sola nº 1; 10200 Trujillo, ESPAÑA Tel: +34 927 65 93 17 Fax: +34 927 32 32 37 [CETA-Ciemat logo]<http://www.ceta-ciemat.es/> ---------------------------- Confidencialidad: Este mensaje y sus ficheros adjuntos se dirige exclusivamente a su destinatario y puede contener información privilegiada o confidencial. Si no es vd. el destinatario indicado, queda notificado de que la utilización, divulgación y/o copia sin autorización está prohibida en virtud de la legislación vigente. Si ha recibido este mensaje por error, le rogamos que nos lo comunique inmediatamente respondiendo al mensaje y proceda a su destrucción. Disclaimer: This message and its attached files is intended exclusively for its recipients and may contain confidential information. If you received this e-mail in error you are hereby notified that any dissemination, copy or disclosure of this communication is strictly prohibited and may be unlawful. In this case, please notify us by a reply and delete this email and its contents immediately. ----------------------------

Hi Kevin, While your "short answer" is correct, to say that the MGS is *only* used at mount time is not 100% correct - though it often seems like it. :) You probably know this, but for Alfonso and possibly others, note that certain administrative actions (storage pools, quotas (as of v2.4), etc.) are configured administratively on the MGS, and Imperative Recovery (v2.2+) has further tasked the MGS. [Description: cid:image001.gif@01C9FE3D.1D8A68C0] Dr. Brett Lee, Solutions Architect High Performance Data Division, Intel +1.303.625.3595 From: HPDD-discuss [mailto:hpdd-discuss-bounces@lists.01.org] On Behalf Of Kevin Van Maren Sent: Wednesday, March 19, 2014 9:19 AM To: Alfonso Pardo Cc: <hpdd-discuss(a)lists.01.org&gt; Subject: Re: [HPDD-discuss] performance split mds mgs Short answer, no. The MGS is only used at mount time, so unless you have an extreme number of clients mounting the file system while other clients are doing large numbers of file metadata operations (open, fstat, etc), the MGS has no impact on runtime performance. Kevin On Mar 19, 2014, at 8:36 AM, Alfonso Pardo wrote: Hello, I must implement a new lustre file system and optimize it to get a high level of performance. I wonder: can I get more performance, If I split the MDS and the MGS in two different machines? Thanks!!! Alfonso Pardo Diaz System Administrator / Researcher c/ Sola nº 1; 10200 Trujillo, ESPAÑA Tel: +34 927 65 93 17 Fax: +34 927 32 32 37 This e-mail (and any attachments) is confidential and may be privileged. Any unauthorized use, copying, disclosure or dissemination of this communication is prohibited. If you are not the intended recipient, please notify the sender immediately and delete all copies of the message and its attachments. ­­