- Four things really matter when choosing
binary usenet servers:
- What you need to know:
- Because so many providers now claim long
retention rates, great completion, and offer similar pricing,
Reliability is now the most important factor in choosing a
Usenet provider. There's nothing great about getting 100
days retention if you can't connect, or it goes down so often
that their completion / retention rates yo-yo all the time.
- Many factors go into our definition of
reliability: Authentication server
availability, equipment reliability
and redundancy,
server overloading,
bandwidth overloading,
billing errors,
support responsiveness, and
problem notification.
- Direct providers of usenet service
have the greatest control over all these pieces and typically
have the best reliability.
- Nearly all the major newsgroup server
companies have redundant equipment, so if a hard
drive fails, the system continues operating on the RAID mirror,
Or if a server or router fails, user sessions
automatically switch to another. All this takes money,
lots of money, and only the largest of companies can afford
to build in the kind of redundancy needed to keep a
Newsgroup Provider available 24X7.
- More than you wanted to know
- Almost all
usenet services have some sort of Authentication Server
to check users access. Resellers often have a separate
authentication server, since they must maintain their own user
lists. A failure here means you can't log on.
This is often the weakest part of a reseller's system - even if
the back-end provider has an ultra-reliable service, if the
reseller's authentication system is weak, you may not be able to
get on reliably.
- Equipment
reliability and redundancy is what most people think of when
they think of reliability. There's never a slow part
of the day with Usenet, so drives, routers, and servers are
always working at full speed. Heat, power problems,
and age all conspire to make equipment break down eventually,
especially as it gets older than 18 or 24 months. A
failure of a backend server or hard drive may mean that
the service will miss some file segments sent by their peers (reducing
completion), even if there is no impact on people accessing
the files on the front end servers. A router failure
can mean that more people have to be routed through fewer
backbones resulting in slower throughput for everyone. And
of course a hard drive failure without RAID or mirroring means
lost messages. Even RAID drives not in mirror mode take
a long time to rebuild after a failure - the best approach
is full mirroring if the provider can afford it.
- Overloaded
equipment fails more quickly and results in slower
throughput for the user. Also, as Usenet volume has increased,
and more firms offer unlimited services, the backend software
has had to be improved with new indexing methods. Many
firms had catastrophic failures shortly after introducing
unlimited access, because the loads on their equipment was
greater than anticipated. And, some companies using GPL
software, especially those without their own resources to tweak
it, must make their equipment work doubly hard to keep up with
all the volume. Other providers use commercial software
with maintenance agreements, and even these firms have to work
hard to keep releases up to date.
- Bandwidth gets
chewed up not only by user downloads (especially as more
firms introduce unlimited accounts) but also by
the peering relationships they must maintain to get the
articles in the first place. Unless a firm really
over-buys connection and bandwidth, there will likely be some
times of day when a high percentage of their users are
downloading that things might slow down. What you
have to watch out for is firms that purposely throttle their
output to you, either to reduce bandwidth bottlenecks,
or for some resellers, to prevent their purchased allotments
from getting used up too fast by their users.
- Most usenet
companies have built their own billing systems which not
only charge your credit card or paypal account, but also
communicate with the Authentication server. A few
resellers also use external services like iBill or CCBill.
Most work just fine, although we get many, MANY emails from
people who report problems with their accounts not getting
activated for days after they'd signed up, or get billed
for months after they've cancelled. Remember also that
if you pay by credit card, they will have your info on file.
If they get hacked, your credit card info could land in criminal
hands. This, combined with start and end account problems
are the number one reasons we recommend sticking with a major
firm that has been around for a while. They've usually got
their billing systems worked out and keep them up to date
against hacks.
- Support is also
a big differentiator. All the firms that actually run
the server farms themselves have staff available 7X24, although
that doesn't mean they answer support questions 7X24. Resellers
are responsible for answering all their customers' support
requests, although if there is a problem with the actual
servers they can't fix it - they still have to report it to
the back-end company they are buying from. And of
course, any of these companies can have jerks on staff -
seems to come with the territory in the computer business.
- Problem
notification can be by email to users (only very small firms
do this) or through some sort of status web page. In
the past many usenet providers had their status pages on
the same data pipes (or even servers) as their usenet equipment
- so when their usenet service was out, so was their status
page. Now all the major newsgroup server companies
have their status pages in separate locations. We
applaud the companies that report all the little problems, even
the intermittent ones - everyone has little backend problems
that create short access or peering problems - we respect the
firms that actually publish this for their users (Newshosting
especially)
Back to the newsgroup server list
|
|
