Reason we/I quote raw is that you can carve up the volumes in a myriad of ways, and depending on how you do so, it will impact useable space. Probably a more accurate measurement (but not necessarily meaningful) is "What is LARGEST useable storage configuration for the storage system?" i.e. in this situation, you COULD use 36GB drives, set aside 1 disk for parity, factor in WAFL/OnTap overhead, set snap reserve to zero, and come up with a single data drive value x 17 disks (of course, in reality, you would LIKELY have a SR of 10-20% and set aside 1 disk as hot spare, but you get the idea). Would you increase Max RG Size to 18? Hmm, that's another topic for discussion...one might go with small RG's and hence, increase overhead, decrease likelihood of double disk failure, and decrease SOME performance...so, I ask again (hypothetically, that is), realistically, how SHOULD a vendor quote storage capacity?!?
I, personally, explain this to ALL of my customers so that they can understand the correlation of max storage to useable...NetApp employees are HIGHLY encouraged to do so when they are hired, particularly SE's.
Lance
-----Original Message----- From: Luke Mewburn [mailto:lukem@wasabisystems.com] Sent: Tuesday, February 20, 2001 9:29 AM To: White, Lance Cc: 'toasters@mathworks.com' Subject: Re: F85 series
On Tue, Feb 20, 2001 at 08:16:55AM -0800, White, Lance wrote:
No, it uses FW Ultra SCSI-3 disks, not FC. Yes, NDMP backup is supported. Limitations? Well, as it's positioned to be the NEW entry level Netapp
(at
3300 NFS ops/sec, it actually about 30% faster than the old low end F720, and about 30% less expensive), it IS limited to 648GB MAX storage (6 internal plus 12 external in 18GB & 36GB flavors), as opposed to the
F720's
limit of ~500GB. Still doin' hot swap, RAID4, snapshots, OnTap, WAFL, everything that's cool about a Filer and more! :-D
The box sounds cool. It's about time a box of this price point was available again :)
I have a (long-time) gripe in the way it's advertised. It's a gripe I have with every vendor. I wish NetApp would advertise the *usable* diskspace of the machine (as seen by a client OS or df, for example), not the raw diskspace calculated by taking the disk vendor size and multiplying by the number of disks.
I know this makes the boxes appear more expensive per usable megabyte, but it's being more honest to the client, and once you explain to the client to take into account all the overheads, and how to do the same math on competing options (e.g, h/w raid + vxfs on Solaris, EMC, etc), then you (NetApp) appear to be delivering accurate technical information.
For example, there's no WAY an F720 has 500 GB of usable space. Out of the box, with default snap reserve, it barely has 300 GB, where GB == 2^30, which is what client based tools (such as df and the `Properties' box in Windows) use, not 10^9, which is what (overzealous) disk vendors use.
The usuable capacity of an F720 is trivial to calculate: 14 x 36 GB disks, utilised as: 1 hot spare, 1 parity, 12 data disks
Each data disk has 24840 MB usable, calculated as: 34500 megabytes (rightsized by OnTAP) - 10% wafl overhead (is this 8% in OnTAP 6.x?) - 20% snap reserve (default; but often reduced to 10%) = 24840 megabytes
So, 12 x 24840 = 298080 MB = ~ 291 GB
Tweaking the numbers to 8% wafl overhead (from OnTAP 6) and 10% snap reserve gives an F720 approximately 334 GB of space (from 500 GB `raw').
Based on the info Lance White gives above, an F85 with 36 GB drives, consisting of 15 data disks, 1 hot spare, 2 parity, 8% wafl overhead, 10% snap reserve would have a usuable capacity at the client of: 15 x (34500 - 8% - 10%) = 15 x 28556 = 428490 MB = 418 GB
418 GB is a bit less than the ``648 GB MAX storage'' that Lance quotes above.
In general, take away approximately one third of the raw capacity that NetApp quote and that will give you an guestimate on the usuable capacity that will suit most client's needs.
Luke.
PS: I'm sure some people have seen or heard this rant before :)
Disclaimer: I like NetApp. I worked at NetApp for a while as an SE. I just believe in being honest to the technical people you're selling to, especially those who have to justify to their management why the (not cheap) 500 GB box they bought ran out of space at 300 GB.
-- Luke Mewburn lukem@wasabisystems.com http://www.wasabisystems.com Luke Mewburn lukem@netbsd.org http://www.netbsd.org Wasabi Systems - providing NetBSD sales, support and services.
On Tue, 20 Feb 2001, White, Lance wrote:
performance...so, I ask again (hypothetically, that is), realistically, how SHOULD a vendor quote storage capacity?!?
...with the most common configuration [0], accepting the default numbers and vendor's recommended operating configurations. In NetApp's case, filers with shelves should be quoted retaining a parity and hot spare drive with the default WAFL overheads and 20% snapshot reserve.
Think of it this way: when a typical customer orders a filer he opens it up, plugs it in, gives it a name and a few network numbers, and <poof> how much space does he see? *That* should be the number in the quoted literature.
Specifically, if we were considering the case of the self-contained F85, this would be (6 drives minus 1 parity minus 1 hot spare) 4 x 36 GB drives, minus ~10% filesystem minus ~20% snapshot reserve, for a total usable space of roughly 100 GB. In this way, there is no suprise to the customer; management isn't upset that the "216 GB" file server (36x6) they paid for is only giving them half of that to use; etc.
Vendors can say also there is a "maximum usable" configuration available; "however, some important features may not be available in this configuration. Your SE can help you decide what is right for your installation." For instance, you could quote a "maximum usable" capacity of 162 GB, which is 5 drives (sacrificing the hot spare) and no snapshot reserve (5x36x0.9).
All IMHO of course.
I've got a small spreadsheet which takes the disk capacity and number of drives and calculates this for me, 'cause I know the marketing capacities are useless.
[0] I know there are a lot of "configurations" (1 shelf, 2 shelves, etc.) but what I mean is if you buy one shelf, most people will, and NetApp recommends, keeping 1 disk as a parity drive and 1 as a hot spare. Sure, one could create a RAID group with 7 data drives and no parity and hot spare, but I doubt that is "common."
Until next time...
The Mathworks, Inc. 508-647-7000 x7792 3 Apple Hill Drive, Natick, MA 01760-2098 508-647-7001 FAX tmerrill@mathworks.com http://www.mathworks.com ---
Well, I'm sure you see why your idea wouldn't work. Someone like EMC will come along and quote raw capacity, without any overhead for Snapshots, or the filesystem, without hot spares, etc. And they will come out look better than NTAP. Heck, they ALREADY try to compare non-parity protected systems with NTAP's parity protected systems. It doesn't matter if the customer gets it, flips the switch, and says, "Hey, where's all that space?" It's too late; the sale has been made, and the company is already invested in EMC. The switching cost to another vendor is non-zero.
Another thing to remember is that since EMC and other vendors are more "established" names, they are often purchased as part of a higher-level strategic decision, by upper-level managers. The mid-level, detail-oriented systems administator doesn't have much choice, even if they understand the distinctiong between the methods for measuring available disk space. And if that admin *is* able to influence the purchase decision, then he is probably smart enough, like you, to know what's what when it comes to capacity and can see past the marketing layer.
Bruce
-
Bruce Sterling Woodcock -
Todd C. Merrill -
White, Lance