Saturday, June 19 2004 @ 08:49 AM CEST Contributed by: bart Views: 2652
CIO.COM is suggesting in this article that IT certificates are mostly useless.
Slashdot has a bit of discussion about it as well.
I believe there are a bunch of problems with how
IT certificates and qualifications are being used,
not that they are useless in all cases.
THe first issue that comes to mind is the large
group of MCSE certified people who may have
learned how to answer a bunch of questions, but
are utterly clueless about the actual technology
they are dealing with. (yes, there are also
MCSEs out there who do know their stuff, I know)
A second issue is that the certificates that are around, mostly deal with specific products and not with the technology behind them. I seriously wonder what companies are trying to achieve by hiring people who have not in any possible way proven to udnerstand the technology and only have shown to be able to reproduce litteral knowledge from a book. Such a person will have knowledge that gets outdated as quick as the product (s)he is 'specialized' in.
It makes many people waste a lot of time and
money. IT professionals waste their time on the
certificates while not learning anything usefull
and either those professionals or the companies
they work for are wasting their money on not
getting anything.
There have been certification programs in the
past that did work, Novell's certified network
engineer did mean somethign for a long time.
What sets such certificates apart?
For a NCE certification, you have to know things
at two different levels:
You MUST know basic networking
You must know the product
The thing is that without a demonstrated strong
knowledge of networkign in general, you will not
even get the oppertunity to qualify for the
product.
Then, to obtain the certificate, you have to
demonstrate understanding what you are doing,
and you have to do so in practise. Yes, you also
have to known all the details.
When you are looking for someone to operate the
restarting of crashed windows machines, a MCSE
will be what you are looking for. For a network
expert that will last a few years longer then
todays bit of 'hot' technology, I would go look
somewhere else really, at least not have a MCSE
certificate play any role of significance.
MCSE is not unique, and is likely not even the
worst example of a certificate that has been
devaulated completely. It is a good example
however because there are a lot of people who
have it, and it is a popular 'filter' when
companies are recruting.
By using this filter, companies are filtering
out the people who they need most, those who
didn't go for the bullshit and instead went for
things that actually work.
As long as the average MCSE can't explain things
like for example the
scaling limitations of a NAT implementation,
they should definitely not be doing network
infrastructure of any sort.
Authored by:
bart on
Saturday, December 11 2004 @ 05:17 PM CET
First of all, there is the number of external ports on the NAT gateway that limits the total number of connections that can be handled simultaneously. THere are a few smart tricks to allow more connections then ports, but that just moves the problem to the next resource, the number of sockets.
Then, to do NAT, the gateway needs to keep a state table.
This state table requires memory, and performance will be unacceptable if it would end up being swapped, so it is limited by physical memory.
Then, the bigger the state table gets, the more time the gateway will need to locate the proper entry for a packet in transit.
Last but not least, there is no easy way to do load balancing dynamically with multiple NAT gateways. If you'd keep state tables in sync so a packet can go through any of a number of gateways then you just cause all gateways to run out of the same resource at the same time, and more improtantly, that is at the exact same number of simultanious connections as a single gateway. If you use a seperate load balancing device, then you just move the state table problem to that device, which imposes a similar limit.