If a network is like a car, then DNS is the ignition key. To extend that analogy, DHCP might be the battery. These two services
are among the lightest available on any network, but no network can function without them. Without DHCP, address management
becomes a nightmare. Without DNS, no one can access much of anything at all.

Infoblox-1000 DNSone
Infoblox, infoblox.com
|
Very Good 8.0 |
 |
| criteria |
score |
weight |
| Management |
8 |
25% |
 |
| Performance |
8 |
25% |
 |
| Configuration |
7 |
20% |
 |
| Scalability |
9 |
20% |
 |
| Value |
8 |
10% |
 |
|
 |
Cost: $9,995; $4,995 per node for grid license
Bottom Line: The Infoblox-1000 DNS and DHCP appliance makes its mark with simple high-availability and clustering features, and its Active
Directory integration is better than Microsoft's own alternatives. We experienced some hardware issues in the lab, and the
Web interface can be sluggish across low-bandwidth connections. But if you want to silo DNS and DHCP services on your network,
this box is worth a look.
|
 |
About our Reviews and Scoring Methodology
|
|
|
|
In an enterprise network, these services usually reside on the same local server or on a central server pushing DHCP scopes
to remote sites and serving DNS requests across WAN links. Though this architecture ties the fate of remote networks to the
WAN, network managers typically choose it to avoid installing servers at remote sites where administrators are loath to tread.
Providing DNS and DHCP services in an easy-to-cluster appliance, Infoblox offers a way to bring never-die network services
to remotely (or scarcely) supported environments.
The Infoblox-1000 DNSone is simple to configure. When first powered up, the box assumes a 192.168.1.2 IP address and is immediately
accessible via the Web, or you can configure the IP parameters using the LCD panel on the front or through the serial interface.
Entering the DNS Zone
Once on the network, the device is managed from a Java interface served via SSL to a Web browser. I had no problems working
with the interface while running Firefox on Linux, Mac, and Windows, nor did IE show any problems. Two caveats: The client
is best suited to running on Java 1.4 rather than Java 1.5. And certain administrative tasks that involve changing core parameters
of the device require a full restart of the Web browser. In some instances I had to reboot my client PC to resume using the
Web interface.
The administration console is well laid out. Adding and removing zones and scopes is simple, and comfortable views of current
zone data are easily had. Because the Infoblox runs Internet System Consortium’s BIND (Berkeley Internet Name Daemon) and
DHCPD (Dynamic Host Control Protocol Daemon) packages, every option you could desire is available. Microsoft’s AD (Active
Directory) is directly supported, so the Infoblox easily serves as the DNS server in an AD environment. The solution also
fully supports dynamic DNS registration, and it’s much better at handling dynamic DNS scavenging than is Microsoft’s DNS server.
Because host information is located in a central database on the Infoblox, a DHCP lease expiration will automatically remove
address and pointer records, eliminating the problem of DNS ghosts.
Impressive HA (high-availability) and clustering capabilities allow you to configure two Infoblox-1000 devices to work as
an active/passive cluster, bringing together custom synchronization and working with VRRP (Virtual Routing Redundancy Protocol).
Intracluster communication is nicely handled by an encrypted tunnel between the devices. Beyond this, you can cluster HA nodes
into a single entity, or grid, allowing for management of the whole cluster from a master console. This includes automated
OS upgrades to cluster nodes, automated zone synchronization, and overall zone management, all of which the Infoblox makes
simple and straightforward. The $4,995 per node clustering cost is steep, but the ability to manage all the devices across
the network from a single console eases administration significantly.
Feeds and Speeds
In the lab, I built two HA clusters of Infoblox-1000 devices on separate VLANs on a Layer 3 switch. A few laptops served well
as DHCP and DNS clients, and a dual-Xeon Dell PowerEdge 2600 running Red Hat Advanced Server 4 served as a load generator.
I created a DNS zone of 100,000 records on a second Linux server and configured the same zone on the Infoblox. Handily, a
feature in the zone-creation dialog box allows you to configure a master zone and import the zone via DNS AXFR (Asynchronous
Full Transfer Zone) from another server. I then transferred the 100,000-record zone into the cluster and transferred the corresponding
reverse zone. Here I experienced a hitch. During the import of the large reverse zone, the cluster fell off the network. After
probing around the boxes with the serial console, I couldn’t get any response from them. Power cycling the units didn’t alleviate
the problem either, but when given enough time (presumably to do file system checks on the boot disk), the cluster did reboot
back to its previous state. A subsequent reimport of that zone did not trigger the same problem.
Once I had the large zone in place, I ran query performance benchmarks against the cluster. At the top end, the Infoblox appliances
handle about 22,700 queries per second, which should suffice for almost any application. I did note that the queries per second
rate dropped significantly when importing larger zones using the GUI. The Infoblox-1000 performs well when compared to a Dell
PowerEdge 2800 with dual 3.4GHz Xeon EM64T CPUs running Red Hat Enterprise Linux 4 and a stock BIND name server, which handles
about 36,000 queries per second. I conducted my query tests against a 100,000-record zone, but the seed list was limited to
100 records to better simulate real-world experience. Using a seed list of 100,000 records, the Infoblox’s performance fell
off significantly due to opportunistic cache aging. The PowerEdge nearly maintained its previous performance level.
Next I tested the clustering fail-over speed. I ran a script to continuously query the VIP (Virtual IP Address) of the cluster
every second. I then forced a fail-over. Initially this did not go as planned, with the passive node failing to assume the
primary role. I replaced the questionable node with another Infoblox unit and brought it into the HA cluster. This configuration
worked flawlessly. I witnessed successful fail-over events completing within five seconds.
Clustering Infoblox-1000 devices is unbelievably simple. From box to cluster node, it took five minutes to configure the node,
and only a few more to synchronize the zone and scope data. Infoblox has done a remarkable job. If you need to guarantee DNS
and DHCP availability for a substantial network, these appliances deserve a close look. Although DNS and DHCP aren’t especially
challenging to administer, the out-of-the-box high-availability clustering combined with the ease of configuration provided
by the Infoblox-1000 is attractive.