Site Map

Flanagan Consulting

Archive       5 Sept 2007             #62


By William Flanagan

Recently learned some details about a software method to achieve high availability for some server applications in small to medium data centers.  The idea is to make multiple servers (up to about 10) aware of each other so they can share the work load, replicate data on their internal disks, and automatically restart an application after a server recovers from a failure.  

Networks typically seek high availability through redundancy in hardware and transmission lines.  The very largest data centers put applications on many servers which connect to the world through (redundant, we hope) specialized load balancing devices.  Huge data bases most often are moved to network attached storage (NAS) or similar hardware dedicated to disk storage and backup.  Yes, it gets expensive, but it is the best way to achieve commercially practicable HA service.

If you aren't large enough to have the budget, but still have the desire for HA, the "SafeKit" software from Bull Evidian (www.evidian.com) is intriguing.  It runs on each server, supporting several functions:

-- Load sharing:  the servers are aware of each others' utilization, so the lightest loaded unit takes the next request.  There is no need for external load balancers.
-- Disk duplication:  in real time, transaction information is kept up to date on more than one of the servers, using their internal hard drives. 
-- Application monitoring:  If a server has a failure, SafeKit takes that server off line.
-- Restarting:  When the server hardware recovers, SafeKit restarts applications and synchronizes that local disk with the working primary server--without taking the application out of service.

For example, a traditional web hosting configuration might take the form of several servers running Apache with a pair of NAS devices to mirror a database on MySQL.  For load balancing, add one or more external boxes.  The solution Evidian offers consists of the Apache servers, but with their internal disks holding the mirrored copies of the database.  Load balancing is included.

SafeKit works with Windows, Linux, Unix (AIX, e.g.), and Solaris.  If your HA appetite can be satisfied at a scale smaller than huge, this is an intriguing solution, but not for every situation.  Evidian recommends less than 10 servers in a cluster, and less than 100 GB in the data base.  A larger file size indicates that a separate database server or network attached storage might be better.  

You say you want more for your money (if you have it)?  Next time we'll look at a database that scales as large as you need with 5-nines availability (or better).

Your discussion is always welcome.


How Can Flanagan Consulting Help You?

   We understand not only the technology of networks, but also
   the surrounding business processes:  procurement, bid
   preparation/analysis, statements of work, financial analysis,
   consensus building, and more.   www.Flanagan-Consulting.com

Find out now:  call +1.703.242.8381


"Flanagan Consulting" and "ViewsLetter" are Service Marks of W. A. Flanagan, Inc.

Flanagan Consulting
W. A. Flanagan, Inc.
45472 Holiday Drive, #3
Sterling, VA 20166
Ph:  +1.703.242.8381
Fx:  +1.703.242.8391
In Converged Networking,
We have the Experience