| Anthony L 2007-06-13, 1:14 am |
| Jack,
Here are some objective high level comparisons that you may want to look
into. I have extensive experience with the Celerra and I can assure you
there are many large businesses using these units exclusively with
minimal issues. Same goes for NetApp, recent versions of both products
work very well but there are some notable differences. I'm an NACE and
have implemented both solutions for a number of clients both small and
large and directly managed Celerra in a medium sized organization (100TB
managed) a couple years ago. Here's what I've seen:
Also first, let me say that I enjoy working with both products, and
they're both top notch, it's just a matter of understanding caveats and
how those caveats impact your functional requirements. Usually they
dont, and it comes down to cost/preference/politics/relationships, or
some other non-technical reason.
Availability:
I've seen and heard horror stories from both sides. If you have a
proper configuration both products should perform as advertised. I've
been in clients that ripped out NetApp and went with EMC, and vice
versa. The Reasons are numerous.
Back-end Disk management:
NetApp provides a very 'direct' way of managing disks which belong to
the system. It's very easy to ensure data is kept on separate spindles.
It's also very easy to configure any type of aggregate/raid group
configuration you'd like (only raid4 or raidDP of course..). The
celerra has its own volume manager called AVM. It's slightly more
difficult to force your data onto certain spindles, and only certain
layouts in regards to RAID group configuration per shelf are permitted.
You can still place data discreetly, but make sure to understand AVM
and it's implications, you also have the option of performing a manual
volume config. Functionally both work very well, but a couple more
caveats on the EMC side. If you're using this for CIFS only, you can
typically let AVM manage all of the disks from the start, and it will
automatically balance IO across the backend disks. Just make sure the
implementation engineer knows this and explains it to you well.
Interface and management:
The Celerra in my opinion has a more functional management interface,
both GUI and CLI. I think this is what you were alluding to, not
necessarily CIFS features. Theres one (or two for redundancy) dedicated
1U management server called the control station. This is essentailly
runs a customized linux distro which contains a set of applications
which manage the data movers (the actual NAS hardware). It also runs a
webserver which presents the GUI interface which is Java based. The GUI
has much more performance analysis features that the NetApp side, and
looks a lot better (if that matters). Also, since the control station
is a linux box, and you do have root access to it, and you actually get
a real fully functional bash shell. It'll may even be easier to manage
via a script based management infrastructure.
CIFS/NFS (file access):
Couple things worth noting off the top of my head. The Celerra does
offer more of what I'd call, configuration flexibility. basically, it
essentially allows you to perform the equivalent of what the vFiler
option would do out of the box. Just an example, you can create
multiple virtual CIFS servers and join them to separate domains. In
fact to present data via CIFS you need to create one of these. You also
have the option of presenting filesystems over 1 or many CIFS servers
simultaneously. On the NFS side, if you do need to leverage
multiprotocol access to a LARGE number of users, the usermapper in the
celerra is usually easier to configure than manually defining
usermap.cfg. There are some automated methods to achieve this with the
celerra and are EMC supported.
iSCSI (block access):
Functionally, both products perform equally well. Although Replication
manager is far more advanced than what NetApp has to offer and the
latest version (5.0) does work very well and will also control snapshots
and replication on the clariion and symmetrix platform (its equivalent
is Snapdrive and Snap manager for sql/exch/oracle). One main
differentiator is the capability of RM to automatically mount a
filesystem/database remotely and run a script. Only thing I can think
of that it wont do is the automated database moves that snapmanager can do.
Snapshots:
The mechanism for taking snapshots differs among the two vendors, I wont
get into details, but the main functional differences are 1) limits, EMC
allows 96 snaps per filesystem (NetApp Volume), NetApp allows 255.
However, with the NetApp appliance you have no option but to store your
snapshot locked blocks in your production volume. The EMC solution
allows you to store blocks required by your snapshot in a seperate disk
pool. I.E. prod FS could be on FC disk, blocks used for snaps can be
stored on ATA. Of course, EMC recommends against this, however if you
know your data well and how you use your snaps, it works well and is a
supported configuration.
Replication Options:
Celerra replicator is functionally equivelent to Snapmirror, not much to
comment, they both work well. NetApp does have snapvault however,
which allows you to keep multiple snaps in a remote appliance. To
achieve this you need replication manager from EMC
Just my 2C! Shoot me an email directly if you'd like to discuss further.
hope that helps,
Anthony
virtualjack wrote:
> On Jun 9, 10:11 am, "Rob Turk" <wipe_this_r.t...@chello.nl> wrote:
>
> Their main motivation is that they are hoping to leverage the Clarion
> infrastructure
> and our considerably larger relationship with Dell/EMC.
>
> Unfortunately EMC doesn't seem to be willing to give us any type of
> demo Celerra unit
> at this time so my analysis is strictly based on their documentation
> and answers from
> the Sales Engineers (who frankly don't seem to understand the product
> very well).
>
> So real world experience of the Celerra is what I'm most looking for.
> People who have
> migrated from one to another (either way) would be fabulous, but even
> plain real world
> experience of the Celerra would help (EMC also hasn't located a decent
> Customer reference
> yet).
>
> The 3020 is working fine although ndmpd backups are not going as fast
> as we would like -
> it is probably more of an issue with the backup system - although I'm
> beginning to think
> that my RAID group & Aggregate configuration is sub-optimal.
>
> ---Jack
>
> P.S. Sorry for the change in headers, my Palm newsreader foobar'ed
> this article.
>
> P.S. Based on what I've read so far, the Celerra has a more feature
> rich CIFS implementation.
> The spare "head/cpu board" is a nice feature as are the disjoint CIFS/
> Unix ACL's. On the other
> hand, the Celerra seems more wasteful of disk allocation and more
> difficult to manage in our
> environment which is script based.
>
|