At 06:55 PM 1/26/95 CST, Keith A. Crabb wrote:
>Must be getting competitive in the network backup market, I've gotten
>two cold calls and our DEC salesrep was just around pushing DEC's
>NetWorker Save and Restore (NSR). Seems they support archiving with
>the 3.1 release available next month and they've just released clients
>for Windows NT for Intel (as opposed to Alpha), Windows TCP/IP and
>the obligatory PATHWORKS Window client all available right now.
>
>
>Sitting here waiting for a few more ADSM clients: SGI, NT, OSF/1.....
>Yes, I know they'll all on the TODO list, just a pain to see that other
>companies already have them available.
>
>---
>Keith A. Crabb Keith AT UH DOT EDU
>University of Houston Operating Systems Specialist +1-713-743-1530
>
It's important to have your key clients supported, but there's a lot more to
evaluating the worth of a product - especially important in my opinion is
the underlying architecture of the product.
The incremental architecure that ADSM uses makes it highly scalable - you'll
be able to support a far larger number of clients on a single server with
much lower network loads because of this. If your looking to scale to a
fairly large number of users (200+) over a wide area network you need an
architecture of this type. We're hoping to scale to 600 users+ over the next
two years on a single server. From our experience with a pilot project we're
currently running using ADSM I'd feel comfortable in saying we could
probably get close to 1000 users before needing to build a second server
(RS/6000). This plays a big role in overall costs - the larger we can scale
on a single server, the lower the costs. As soon as multiple servers are
required - especially if they have to be located in different areas of
campus, the cost jump significantly.
I'm aware of only one other product that uses an architecture similiar to
ADSM - Harbor. All of the others we looked at were still stuck back in the
stone age concept of Full, incr, incr, incr, incr, Full, incr, incr, incr...
Some didn't even support compression on the client side before sending data.
Do some calculations on how much data is moved across the network in a year
by the different architectures, or how much server storage you require per
user - the difference might surprise you.
Some of the products rely solely on the archive bit to determine which
datasets to back up on an incremental - this bit is not reliably set in many
environments (dos, windows for sure). This means the incrementals may not be
backing up all of the datasets they should be. The longer the periods
between full backups, the bigger this integrity exposure.
Disaster restores will also be much faster if you use the collocation
feature - keeping a clients data on as few tapes as possible. Not possible
with the full, incremental types.
I've found the ADSM clients to be much easier to use and understand than
other ones we have evaluated - the user need have no concept of save sets,
what media the data lives on, full or incremental... Just of their data.
I think ADSM has done well with the clients supported to date (even mac's -
thank you - many on campus). Sounds like their working on NT and I'm sure
Windows 95 clients (would 1 work on both?). Anyhow, for all the reasons
I've listed above, I wouldn't switch just because someone has certain
clients available sooner.
|
