Dell BIG-IP PowerEdge
Mar 22nd, 2004 by Alex
The Dell PowerEdge Load Balancing Server is a product that doesn’t get a lot of attention. BIG-IP is a product line of application traffic management appliances from F5 Networks that gets it’s due share of attention from the right sort of crowd. Load balancers and more; their high-end appliances can be configured with FIPS (Federal Information Processing Standard) SSL accelerator cards, providing FIPS 140-1 level 2 validation. F5 seems to to excel in providing redundant HA (high availability solutions) to those who can afford them, in the arena of tens of thousands of dollars. Load the BIG-IP software on a Dell PowerEdge 2650 and you have a rather powerful and versatile device.
BIG-IP is essentially a custom-built — designed for load balancing — kernel supported by both CLI and web-based management. The kernel is built on BSDi (BSDi Internet Super Server to be exact, but also known as BSD/OS), which is a commercial BSD UNIX distribution owned by Wind River. Since the goal of BIG-IP is to break apart a TCP/IP stream and do things that were never intended, most of the customizations are believed to be related to tracking the broken up streams that are being balanced by the device.
Because BIG-IP is built on BSDi, there are some limitations of the underlying kernel that are introduced in BIG-IP. As of this posting, the following limitations have been confirmed with F5 Networks engineers.
- SMP capabilities are limited. Dual/quad-processor servers will be seen as single processor systems.
- The largest memory size currently supported is 2GB.
- BIG-IP does not support RAID. So much for mirroring operational drives for backup purposes.
Sure enough, just as the F5 engineers said, you can only have 2GB of memory in the things. If you have more, your boot screen will get hung up before it can detect your hardware.
Copyright (c) 1996-2003
F5 Networks, Inc. All rights reserved.
BIG-IP Kernel 4.5.9 Build2
Hardware Configuration: _
The idea of having interchangeable hardware for application traffic management appliances is a cool one. Suppose a processor dies on your balancer. Simply swap the drives into a spare rack-mounted server and reboot. Having an extra server lying in the rack unused isn’t a bad idea, but how useful does it become if it can only be a single-processor system with a minimal amount of memory and no RAID support? So BIG-IP doesn’t support the extra memory. Fine, it won’t be addressed. Same with the processor. The RAID card is an issue, however. On a PowerEdge 2650 RAID-support is built into the boards by default, waiting to be unlocked by a tiny “RAID key” chip that gets inserted onto the motherboard. Not such a big deal. On other rack-mount systems, it could be a whole PCI card to remove. All of a sudden, things aren’t as interchangeable as we’d like.
The installation, configuration, and operation of BIG-IP software is pretty fluid, and well thought out. Services to be balanced are assigned to a pool by IP and port. A pool is then assigned to a virtual IP, to be listed as the public IP for the balanced services. The actual balancing can be done numerous ways: round robin, based on latency, the load on the servers, etc. A pair of BIG-IP units can be configured in either active/passive mode or active/active mode. In active/passive only one unit will field requests unless there is a failure, at which point the passive unit takes over. In active/active HA configuration, both units balance requests. All of this is configured via web-based panels or via CLI commands.
More information is coming regarding the troubleshooting of these devices.
