Red Hat Releases Clustering Software
Written By: R. Krause
Published On: August 15 2000
Red Hat Releases Clustering Software
July Red Hat, Inc. announced the Red Hat High Availability Server 1.0,
a specialized version of the Red Hat Linux 6.2 solution.
Hat High Availability Server is an out-of-the-box clustering solution
that delivers dynamic load balancing, improved fault tolerance and scalability
of TCP/IP based applications. It lets users combine individual servers
into a cluster, resulting in highly available access to critical network
resources such as data, applications, network services, and more. If one
server in the cluster fails, another will automatically take over its
workload. The Red Hat High Availability Server is ideally suited to web
servers, ftp servers, mail gateways, firewalls, VPN gateways and other
front-end IP-based applications where virtually uninterrupted service
product supports heterogeneous network environments, allowing individual
members of the cluster to run Red Hat Linux or virtually any other OS
including Solaris, and Windows NT. Because the Red Hat High Availability
Server is an open source product, customers are free from expensive technology
lock-in that often occurs with proprietary solutions.
Although other companies such as TurboLinux and Steeleye Technology have
released Linux clustering software, Red Hat's entry into this space will
change the scenario. With over 65% of Linux market share, Red Hat is presently
the proverbial 600-pound gorilla (we'd normally say 800-pound, but Red
Hat does not control the Linux market the way Microsoft controls the Windows
Availability Server (HAS) offers most of the traditional clustering features:
scalability, dynamic load balancing, failover. Red Hat claims that HAS
will support as many nodes as a user's hardware and network can handle.
This is in contrast to some other Linux clustering solutions, which range
from two nodes (Cobalt's StaqWare [ref. TEC NA Cobalt
Releases Linux "Clustering" Software]) to 25+ nodes (TurboLinux's
TurboCluster Server [ref TEC NA TurboLinux
Clusters - One More Step Taken]). Windows is still trying to get to
a serious cluster solution - it presently supports two-node clustering,
and hopes to support four-node clustering when the Datacenter Server version
of Windows 2000 finally ships (ref TEC NA
Microsoft Readies Win2K Datacenter for Defeaturing).
$1995, the fully-supported version of HAS is priced in the right range
(Linux clustering software is generally running in the $995 - $1995 range).
It is also available for free download by users willing to forgo automated
installation, support, documentation, and all the other nice things Red
Hat throws in when you buy it.
the scalability is truly unlimited (we have not verified this through
testing), we can see this product benefiting on the ISP/ASP market - users
who have a ton of small servers (as opposed to a few very large servers)
and who need uptime, load balancing, and scalability.
As described earlier, Linux users with lots of small servers (or server
appliances) should look at this product more closely. Users with "small-to-medium-size"
installations (less than 20 servers) should compare Red Hat's HAS, TurboLinux's
TurboCluster Server, and SteelEye Technology's LifeKeeper.
committed to Wintel will have little interest, but those keeping an open
mind vis-a-vis Windows may find that Red Hat's clustering solution may
provide enough incentive to consider switching from Windows NT to Linux
- especially if they plan to increase their server quantity significantly.
customers should also understand that the $1995 price tag, although attractive,
is only the start of the cost of ownership, as is true with all server
installations, Linux or Windows. The free-download option should be exercised
only by those users who have sufficient Linux experience and know-how.
Casual users should spend the cash for the full-price version, doing so
will save them money in the long run.
with all new products, we recommend a trial period, to ensure that the
product's performance matches the vendor's claims.