Tracking and
cracking your network to ensure that there are no performance problems is not
easy. More than just identifying mystifying bottlenecks, ensuring network
efficiency requires an almost high understanding of how your network is
functioning down to the nitty gritty stuff. It is also the ability to have a
somewhat “thick skin” to withstand the heat when problems inevitably arise.
Also, it is remembering that just because something
is wrong with the network doesn’t mean we get to blame the network itself
at all times.
To keep that
network humming, there are certain areas you really need to tweak that can lead
to significant performance gains. This also ensures that your system is blazing
that is essential to your needs.
Disk
striping
Often the
network itself is not the bottleneck. If you have several hard drives, you can
combine them into one logical drive where the data is “striped” across them.
There are some downsides but performance is not one of them. This is called “RAID
0” by computer experts.
Balance
your system load
In the same vein
as disk striping, you don’t want all your I/O for your NIC, hard drives, and
tape drives, etc. on the same load. Most servers have several, so put some
thought into how you can optimize this. Specifically, realize that data does not
go directly from the hard drive to the NIC if they are on the same load. All
the components still have to chat with your CPU, so if there is contention, it
is faster if they are on separate loads.
Clean
up network protocols
If you are one
of those people that still have NetWare IPX, Appletalk, NetBEUI, and TCP/IP
protocols bound to every interface on your server, get rid of the ones you are
not using. Refusing to let go of the past often results in increased costs,
downtime, and fragility of core systems. Instead of holding meeting after
meeting to figure out how to get a 10-year-old package transferred to new
infrastructure, throw it out and get something new. The upfront costs may be
more, but they will pale against the long-term costs you will incur by not
severing these ties.
Adjust
your TCP/IP settings, particularly the window size
If you cannot seem
to find any bottlenecks, and you are not getting the throughput on your WAN
that you think you should, remember the bandwidth delay product and check your
window size.
Implement
WAN bandwidth saving models
There are lots of technologies out there that
fundamentally change the model of networking and result in dramatically lower
WAN utilization or latency. Some examples are terminal servers, content
networking, WAFS, and Web services. You may also want to investigate your web of
leased lines and costly WAN charges. Linking multiple sites with T1 lines,
MPLS, and even Frame Relay used to be the only way to guarantee connectivity,
but everything has changed since with improvements in technology. Rather than
curse at your monthly WAN bill, it is high time to investigate your
alternatives.
No comments:
Post a Comment