Re: [WebDNA] Store module/site
This WebDNA talk-list message is from 2012
It keeps the original formatting.
numero = 109042
interpreted = N
texte = > You don't necessarily need a master if data consistency=20 > across all nodes is not critical. But let's say it is. In=20 > this case you must maintain the master on a server that's=20 > outside your dynamic node pool ... exact > This means that every db write within your node pool must be=20 > successfully replicated to the master that's outside the=20 > pool. No, in this case, it would be the single master that would need to be = replicated to the pool of slaves. With WebDNA 7.1, writing to a database = in RAM is considerably faster than in previous version. This means that = a single server with WebDNA 7.1 can handle several times the maximum = load of older WebDNA versions.=20 > This poses potential problems in and of itself, but=20 > there are more difficult problems that may exist too: >=20 > The host likely will not, or cannot, give you individually=20 > addressable access to each node in your pool via the=20 > internet -- so how can you synch those nodes when you cannot=20 > address them from your master db server? Let's say you duplicate a single WebDNA website accross several servers = with a load balancing system (f5-networks, Alteon ACEdirector Load = Balancer..). Each copy will be loaded in each server RAM. There is no = way to share the RAM accross the servers, so the common denominator will = be the disk copy. Since you have full control over how WebDNA writes to = disk, it is a matter of synchronizing text files. Once synchronized, you = need to reload the updated database into RAM. If the pool system is thought for static websites, things are getting = more complicated (and the same problem would exist with every type of = database). An idea would be to be able to detect the node WebDNA is = running on and writing to a specific individual node database = (database_node3.db) with a global merging process afterwards. = Complicated Anyway, if i were to need to handle a very high load on a webdna = website, this is what i would do: - fastest CPU speed you can find (3.8 GHz?) - fastest RAM you can afford (DDR3 2400?) with a motherboard that could = handle it - lighttpd instead of apache - Debian - working from RAM only with a [commitdatabase] when critical data is = added - an accelerator certificate :-) This could be *extremely* fast.. - chris
Associated Messages, from the most recent to the oldest:
> You don't necessarily need a master if data consistency=20 > across all nodes is not critical. But let's say it is. In=20 > this case you must maintain the master on a server that's=20 > outside your dynamic node pool ... exact > This means that every db write within your node pool must be=20 > successfully replicated to the master that's outside the=20 > pool. No, in this case, it would be the single master that would need to be = replicated to the pool of slaves. With WebDNA 7.1, writing to a database = in RAM is considerably faster than in previous version. This means that = a single server with WebDNA 7.1 can handle several times the maximum = load of older WebDNA versions.=20 > This poses potential problems in and of itself, but=20 > there are more difficult problems that may exist too: >=20 > The host likely will not, or cannot, give you individually=20 > addressable access to each node in your pool via the=20 > internet -- so how can you synch those nodes when you cannot=20 > address them from your master db server? Let's say you duplicate a single WebDNA website accross several servers = with a load balancing system (f5-networks, Alteon ACEdirector Load = Balancer..). Each copy will be loaded in each server RAM. There is no = way to share the RAM accross the servers, so the common denominator will = be the disk copy. Since you have full control over how WebDNA writes to = disk, it is a matter of synchronizing text files. Once synchronized, you = need to reload the updated database into RAM. If the pool system is thought for static websites, things are getting = more complicated (and the same problem would exist with every type of = database). An idea would be to be able to detect the node WebDNA is = running on and writing to a specific individual node database = (database_node3.db) with a global merging process afterwards. = Complicated Anyway, if i were to need to handle a very high load on a webdna = website, this is what i would do: - fastest CPU speed you can find (3.8 GHz?) - fastest RAM you can afford (DDR3 2400?) with a motherboard that could = handle it - lighttpd instead of apache - Debian - working from RAM only with a
[commitdatabase] when critical data is = added - an accelerator certificate :-) This could be *extremely* fast.. - chris
christophe.billiottet@webdna.us
DOWNLOAD WEBDNA NOW!
Top Articles:
Talk List
The WebDNA community talk-list is the best place to get some help: several hundred extremely proficient programmers with an excellent knowledge of WebDNA and an excellent spirit will deliver all the tips and tricks you can imagine...
Related Readings:
emailer error -108 (1997)
Searching Numbers (2004)
[ConvertChars] problem (1997)
syntax question, not in online refernce (1997)
Bug alert! (1997)
NetSplat and WebCat2 (1997)
No luck with taxes (1997)
[SHOWIF AND/OR] (1997)
WebCat2 - Getting to the browser's username/password data (1997)
are we ready to fix random sort? (2000)
emailer (1997)
FW: WebDNA-Talk searchable? (1997)
[WebDNA] copy some database fields into a new database (2016)
Paths, relative paths, webstar server setup andsecurity WebDNA-Talk@smithmicro.com (1997)
problems with 2 tags (1997)
Date Calulation (1997)
charging? (2001)
Errata: WCS Newbie question (1997)
I'm having trouble using [url][interpret][math] together in lookup (1997)
Preventing Safari caching (2003)