November 11th, 2001, 12:41 AM
High Availability for medium--small businesses
This might be a long post, but I hope its worth it to some people here:
I'm researching the best ways to bring FreeBSD into deployment in medium-small businesses. Actually, I will be doing some printable documents that help make the case for BSD in business, and I will be glad to share some of these with you guys when they are ready. I can be a somewhat convincing writer at times ;-).
I am familiar with all of the standard reasons for BSD's greatness (stability, networking performance, etc...), but I want to really get into specifics, in documenting just how businesses can improve price/performance points in their networks. For example in OnLamp's Selling BSD column, you get some good advice on how to present yourself as a Unix expert to business, but still not a lot of real content, with good arguments and rebuttals. Many businesses are getting tired of Microsoft and looking for alternatives, and of course they are turning mainly to Linux. Fine, and at times there are good reasons for using Linux, but let's get into some good reasons for recommending FreeBSD.
One area I am very concerned with now is high availability (HA): clustering, mirroring, failback, etc...
Can any users here share experiences (good and bad) in using FreeBSD for high availability? Can we start forming a list of pros and cons, and good/bad approaches to take, and what software/hardware to use?
For example, one of the common problems in starting out an HA system is the cost of a load-balancing router, which can usually starts at $10,000+. There are a few cheaper "software appliance" products out there, such as RedHill's WebMux, which starts at around $3000, but as long as I have to use a software solution, I would rather have it be FreeBSD.
Now I know that normally, for a company that seriously needs a true HA system, this kind of money is peanuts, but maybe there is a new niche in HA for medium/small businesses. For example, I have a client with a web-based business who regularly stresses out his one Linux server to the point of failure. He doesn't have the money for a standard Rackspace.com beginning cluster, at $6000/month, nor can he afford a heavy-duty Sun server, or somesuch. So we want to start a small cluster, with the ability to scale it quickly, without having to completely start from scratch if his business suddenly increases largely.
I know there are a few FreeBSD-compatible products that can help in this area, such as Polyserve. But are there any open-source tools to do the job? In short, is there a way to implement a half-decent load balancing environment with open-source tools, and better than just round-robin DNS? What scaleability restrictions would we run into?
The above questions of course apply to a web/database application server cluster. Now another question would be how to implement some sort of high-availability for a fileserver. Many businesses want to keep their windows desktops, but they are overjoyed at the possibility of a Samba server with unrestricted licences, as long as it can be made to work transparently to their users. That's no problem, but the question is, what happens when we want to scale, and provide HA in a Samba environment? How about CODA, or Vinum to deal with these problems? Can anyone share experiences?
These are just a few questions that I would like us to consider researching, and coming up with some scenarios. As I said, this is something that can benefit anyone who wants to make a living on BSD. I think the time is right. If we can get our ducks in a row, then we can make a nice niche market out of this.
I will be dealing with quite a few other questions in selling BSD solutions, and hope to post a few more threads in this vein. Stay tuned for some printable documents you can hand out to your clients and bosses.
August 14th, 2007, 02:05 AM
I have searched some time ago for making a HA server with FreeBSD.
I am very interested in this subject.
So .. I found something interesting what's wait for testing
For HA look
ggated(8), ggatec(8), ggatel(8)
For Parallel Computing look
Also you can search in google for SG Cluster.
Cose I can't post here url's
this post is realy old )