NetApp usable space calculation formula

NetApp RAW to usable space calculation formula

All of sudden today I got a question "Is there any quick formula to calculate the usable capacity for NetApp filer?" in all my career I never had to come across for any such kind of formula so curiously I started hunting on web and found a number of sites offering such tool but most of them either dead or pointing to a tool developed by Nick Bernstein. This is a small tool available for free download, where you have to select the disk, raid and some other options and you get a good numbers of usable space from your raw space. Although it gives a very close figures and sometime actual number, but I found it's not all, as there are some more constraints which one has to keep in mind while quoting usable space.

So I started hunting more on net and came across one blog from Jim of HP, where he is criticising the NetApp's not so called clear approach of giving any formula for same and after that a number of following comments by some NetApp big-shots.

After going through post and accusations made by each other on their product I found an easy to understand formula with example posted by NetApp guy. Now below is the formula given by them where I have added my points which I think is required while calculating usable size.

So first thing first, as we all know that space advertised by manufacturers are in base10 formula which is just a marketing funda but we use in base2 formula which is the actual space your system sees once you connect the disk. So I will use GB forms in base2 which is the actual space we can use.

Let's take example of 20 FC disk of 144GB which after converting in base2 number comes to (136000MB/ 1024) 132.8GB x 20 = 2656 GB
(You can check the base2 size of each disk in sysconfig -r command)

Now we will use the disks in Raid-DP so each raid group will reserve 2 disks as parity, so now we have space left-over with 20 - 2 x 132.8 = 2390 GB
(Please note that here I have taken the example of FC disk where maximum number of disks limit per RG is 28, check the NetApp Storage management guide under topic Maximum and Default Raid Group Size or see the online version on now site here)

As NetApp system stores an additional checksum of 8bytes for every 512 bytes of data, which is ~1.5% overhead or 35GB in this case. So we are leftover with 2355GB.

Now reduce 10% for WAFL overhead, so it comes to 2120GB

Now change the default aggregate snapshot reserve from 2% to 0%. Why? Because aggregate level snapshots are primarily used for metrocluster.

So to summarize that let's see the an easy step by step calculation formula
1. Check available disk capacity in base2 number (a)
2. Number of disks - hot spares - Parity disk required by raid group (depends on raid type and disk type+number) = number of disks will be used in aggregate (b)
3. a * b = Raw space available (c)
4. c - 1.5% = Space after additional checksum overhead (d)
5. d - 10% = Space after WAFL Overhead (e)

So in nutshell 88.65% is usable space after multiplying with the value of raw space in base 2 value

Link from netapp from raw to usable conversion a good read if you want to know anything further

https://now.netapp.com/Knowledgebase/solutionarea.asp?id=kb34044

Update:

On SATA disks space used for checksum in BCS type is more than 11%, but if you use ZCS the net loss from checksums is a little under 2% which is consumed from the wafl reserve

Update:

Post updated following changes in release of ONTAP 8.1.

10Gig Ethernet Card

Just got some stuffs while doing research for new 10Gb card installation, which we have purchased for our NetApp V6080 file.

"Be careful not to get lured in by marketing fluff such as TOE (TCP Offload Engine), iWARP (Internet Wide Area RDMA Protocol), CFE (Congestion Free Ethernet), and Unified Wire, are all marketing features whose value is negligible or unproven in the market. TOEs had some measurable value in dual-core, dual-socket servers, but that's since been erased by improved stateless offloads and the move to quad-core, dual socket systems. iWARP is essentially the third attempt by Intel to make their Virtual Interface Architecture (VIA), also known as Infiniband, viable in the market, only this time on Ethernet. Infiniband continues to be hard to implement, one need only look at the 80+ papers published by researchers trying to get it working to know that iWARP is heading down a hazardous path. NetEffect was the market leader in iWarp NICs and they went bankrupt in August 2008, because iWarp hadn't gained any traction in the market. Congestion Free Ethernet (CFE) is the most interesting in that it assumes most people don't know that 10GbE by design has eliminated collisions and employed measures to enable switches to apply back pressure on NICs to reduce congestion and packet loss. So CFE is a fancy marketing term for what you're already getting from everyone, but with a subtle twist. Next we have Unified Wire, the ability to move both data and storagetraffic over the same TCP/IP network with "no changes to software," another wild marketing claim. Storage traffic will eventually all travel over Ethernet, it's inevitable, but it will be via a network file system or Fiber Channel over Ethernet (FCoE) and software changes will be required. Please remember, when evaluating a 10GbE NIC remember to focus on the performance with YOUR applications, price and the physical media (CX4 or fiber), that's all that matters. Consider doing a bake-off."

I would recommend anyone who is interested in 10Gb implementation to head over to this site as it is having some very good stuffs.