-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Thre, On 11/29/2010 07:32 AM, Mohit Chawla wrote: > Hi, > > I'd like to know how folks around here are actually scaling SOGo > deployments. Not trade secrets, but at what levels does scalability > needs to be taken care of. > > So, if I have say, two SOGo servers, then what would be the correct way > for preserving sessions across those nodes ? Are you sure you need to? I'm not sure, but if SOGo follows the REST principles then you might not have to. > > To be more specific, consider a round-robin dns deployment in place for > these nodes, with TTL ~30 seconds. Then in case of a node failure, what > all data and session stores need to be taken care of ? Does it make > more sense to have a single sql database (on a SAN maybe) or > multi-master replication across the nodes ? How big is your deployment? Have you considered separating the DB from the frontend nodes?
> What about memcached ? Does it need to be replicated as well ? At what > other levels do I need to be concerned ? Replicating memcached is a good idea if you got many hits across the servers - or if sessions needs to be replicated. Other approaches to consider: - - look into using varnish[1] to speed up repeated requests. - - if you got massive scale issues, consider sharding your users into separate boxes. - - consider using ningx for loadbalancing instead of DNS. Regards, Tarjei - -- Tarjei Huse Mobil: 920 63 413 -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkzzUxQACgkQYVRKCnSvzfLufgCeJcC35cjpklV09m8V4nLiPzd3 QyAAmQEs85s7EiwxM6jInt8NBVzWIchT =4F58 -----END PGP SIGNATURE----- -- [email protected] https://inverse.ca/sogo/lists
