[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: NetAps vs EMC



On Jan 31, "Mr. Alcourt" wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> 
> On Tue, 30 Jan 2001, Brad Knowles wrote:
> 
> > 	I don't know if this is possible with EMC, but I have to believe
> > that their management interface is more full-featured than NetApp.
>
> 
> My problem with EMC is that I don't have _any_ management interface that
> EMC is willing to tell me about.  Because of the model of local disk and
> the fact that we were forced to run at Solaris 2.6, we were forced to
> reboot after editing the /kernel/drv/sd.conf file just to allow the system
> access to access a bit more drive space that already existed inside the
> EMC array.  I'm sure there is some management interface to allow me as a
> sysadmin to control the box, but the way EMC is playing it right now, it's
> all voodoo controlled by EMC support personnel.  They seem unwilling or
> unable to provide basic information on their product even.

Note: Platform comments only apply to EMC's high-end productions and not to 
their products from their buyout of Clarion.

EMC does have a management programs. I've seen Windows based and native host 
based packages. It has a GUI based interface you can create various RAID sets, 
LUN's and associations via a drag and drop interface. They also have command 
line interfaces on the hosts to allow you to do things things like breaking a 
mirror in order to do a backup on the mirror copy.  I've also seen performance 
monitoring software.
 
> I admit, there is a good chance that my problems are due to the
> individuals at EMC that I am dealing with.  But I get nervous when I'm
> told to pretend it's just a very big disk array with no management control
> needed, until we decide to take advantage of a little more hard drive
> space and are told we need a new bin file and new microcode and only EMC
> can provide such.  (Yet the hard drives were already installed in the
> array, we just weren't using them yet.)

You experiences are not unique. They don't like customers to much with the 
insides of the frames. They want only EMC techs to manage them. That's why you 
pay through the nose for support. But I must admit, you do get good support.

While at a previous job we where evaluating server and storage solutions for a 
big Oracle database with around 1 TB of storage. We asked various local and 
corporate EMC personnel about basic performance (i.e. max I/Os per second, 
average transfer rates, peak transfer rates, etc) of their "frames".  We got 
no response from local personnel or their standard 4 color glossy sales 
package. The sales package told use everything about their products except 
performance stats.

And on the sales front, let's just say I've seen 'at any cost' behaviour in 
order to nab a sale. This varies from salesman to salesman but it stems from 
the corporation. We eventually got performance information when we had some of 
their engineering types in a presentation. But they were going all out to 
convince the company to go with EMC instead of Hitachi. (EMC won becuase 
Hitachi was a couple of weeks behind EMC in development of a couple of key 
thing we need for our environment)

And I give a few comments about the structure of their high-end frames. They 
have storage processors which handle the actual data handling on the back-end 
and IO directors which are the interfaces (Fiber Channel, Ultra-SCSI, etc) to 
your servers. Generally you will have to use only EMC approved Host Bus 
Adapters, microcode, drivers and OS patches. So beware on that front.

The storage processors use the same memory for programs/processes and  data 
caches. You can control the balance between the two, but you don't have a 
dedicated cache memory. And there are only two control/data buses in the frame 
going to all of the disks. You don't have a storage processes dedicated to 
groups of disks.

Now to the disks themselves.  All of the disks are HUGE. The last I heard, as 
of a few years ago, they were putting in 36BG drives as their standard drive. 
The drives are divided into partitions. These partitions are your basic 
building blocks for creating storages sets. You don't manage thing at the raw 
disk level. This means that you could create a RAID set out of partitions that 
are on the same physical device.

Stephen L Johnson <sjohson@monsters.org>