Hi,

I need to have the following setup.

We have a bunch of webservers around the world. the webservers are hoooked to databases as well as to storage facilities.

what i want is

1) Fault tolerance. when one webserver goes down it shouldnt impact the system and end user should not experience any disruption.

2) Load balancing. Requests should be load balanced between the servers.

3) when a database entry is changed in one database the change should be reflected immediately in the rest of the dbs around the world. i would prefer multiple redundant databases then a single central one for reasons of fault tolerance.

4) when a file on any storage medium is changed, the change should be relfected on the rest of the storage facilities. the rest of the stuff which is cannot be stored on the db, is stored on the shared file system storage facilities. the storage should be available to the web application as well as the developers as a single logical storage facility.

hence a change made by the web application or a user should be relfected in the rest of the identical redundant faciities.


can any one suggest any solutions? open source as well as proprietary solutions are welcome.

Regards

I think you'll find out that in looking for "quality" high-availability solutions it's not choosing software products that will be pivotal but the costs. It may for instance seem "obvious" to do everything yourself like negotiating services between datacenters (like multiple dedicated lines and NAS storage) until you understand what the risk of doing that is. If you're opting for "quality" providers you'll find that, even though they cost more, their expertise and usage of SLA's (basically: guarantees) will mitigate risk for you and move liabilities with the other party. You might also reach that point in the project where you will be forced to look at each component again, especially legacy code and "old" HW setups, to make sure it is performant, scalable and redundant. Finding these "hidden treasures" during the project can make costs spin out of control. So what I'm saying is that, for this to work out successfully in terms of costs and risks, what you need is to work out the analysis from application level to network architecture to planning to SLA before you get the picture "right", before you start the project and before choosing software. Ofcourse you dont have to take it from me.


Everything depends on what the applications are about, what your providers can offer and what the customer is willing to pay. Next to fail-over HW there's SW like virtualisation. I don't know if Xen is "combat proven" wrt production environments, but LVS and Vmware are. There's HW routers that come with load-balancing solutions. SW wise look for HA (Heartbeat), Keepalived and taking away load by proxying content. Wrt the database stuff, yes, I'm sure that can be done but them pipes will cost you. There's companies that provide the whole (near-)realtime data shoving stuff as a package on a global scale, it's just I can't remember names and I'm not for free ads anyway.


the storage should be available to the web application as well as the developers as a single logical storage
Through experience you will find that you will never ever again allow developers on production boxen. Main responsability for sysadms is stability, while developers fsck up stability w/o taking responsability ;-p You should figure out how release management could work for you, chain developers to dev boxen and only allow supervised access to staging environment for releases.

thanks a lot for the reply.
can you please tell me what product services do your provide wrt to what i have to do?

what product services do your provide wrt to what i have to do
I'm not sure I understand what it is that do you have to do.
Do you have to set the whole thing up? What's your role?


* BTW, you can't put out job ads on LQ or try to hire people here w/o agreement from the site owner.

Heck, I'm a DEVELOPER and I don't allow myself on the production boxes... Not even if my manager is screaming at me "We don't have time to do it on dev first! Just do it in production!" (Funny, those managers don't seem to last very long. Usually they get promoted to areas they can't do much damage in pretty quickly )

@unSpawn,

My role is to propose the solution and provide it to my boss. we may also do little prototype testing before we implement the solution.

I am only seeking the names of the products. Not looking for ads or employing any one.

I would need products related to the following categories

1) Distributed clustered file system
2) Distributed clustered database
3) fault tolerant and load balancing web servers

Regards

My role is to propose the solution and provide it to my boss.
OK. Clear.


Not looking for ads or employing any one.
Now I ain't saying you would, just thought it'd be good to mention, just in case.


Distributed clustered {file system,database}
I don't usually mention non-FOSS, but Oracle does provide complete packages in a way that FOSS, separated and glue-less as they are, I think would have a hard time competing against unless you can get some good experienced Linux consultancy to help you.


fault tolerant and load balancing webservers
As far as I understand tech there are no "fault tolerant and load balancing webservers" because that is not their purpose. There's soft and hardware for that. I think I've mentioned enough search terms for you to look up.