ONTAP Discussions

ISILON vs NETAPP for midsize enterprise customer?

EVILUTION
12,461 Views

Quick and dirty...

EMC is our primary vendor so Isilon is being pushed hard....   we are interested in what NetApp does better than Isilon for the "average Joe" customer. 

We are a midsized company that is looking for a HA solution that will be easy to grow and manage as our demands grow.

To keep things simple...  we currently have 60TB of unstructured file server data that we would like to move off of our expensive VMAX storage and onto a NAS.

Additionally we will be looking to migrate some test VMWare servers over to NAS as well as begin testing VDI solutions.

Our proposed solution from NetApp is (4) 2240s at our primary and DR sites.  We will leverage snapshots as a backup solution and replicate to the DR site.  Know I know this will do what we want and we can always had heads as they are needed.  What I want to know is what this solution will do better than what I primary vendor has to offer.  

Please keep in mind simplicy and reduced complexity is HUGE for my shop.... we have a limited number of engineers that are doing it all ... builds, deployments, capacity planning, research and dev.... bla bla bla ... We do not have time to babysit and storage frames.  I'm not really looking for ultimate storage IOP and latency showdowns.  Can anyone help with real world examples?

1 ACCEPTED SOLUTION

JWHITE_COMPUNET
9,596 Views

Things that NetApp does exceeding well compared to Isilon (or even VMAX):

Deduplication.  Dedupe is not just about saving storage.  The NetApp Cache mechanism is dedupe-aware meaning that your Virtual Workloads, especially nearly identical ones like VDI, will fit nicely into cache no matter how many of them there are.  "Cache Amplification."

With Cluster Mode, balancing workloads is manual when you want to move between storage tiers or controllers.  But that usually isn't necessary unless you're doing hardware refresh or drastically changing your workload.  Isilon has performance issues realted to the way it distributes data.  Moving data into lower tiers, locality of data that is spread all over the cluster, and having a variety of node types with different performance characteristics are some of the challenges.  With NetApp Virtual Storage Tiering, most priamry workloads can live in your lowest tier of disk while warm blocks live in large cache regions (Flash Cache, Flash Pool) and Cache Amplification from dedupe enhances this all the more.

NetApp disk IO (via WAFL) is uniquely performance accelerated.  Nobody can keep up with NetApp spindle per spindle, especially on random write (VDI), not even your VMAX.

Don't undervalue NetApp's Snapshot technology.  WAFL is unique in its ability to create large numbers of snaps that impart no performance penalty.  NetApp SnapMirror replication is second to none and is superior to Recover Point or SRDF in my opinion especially built-in compression eliminating the need for WAN acceleration.

Things that NetApp doesn't do as well:

HA Pairs.  HA Pairs impart an odd multiple personality disorder to NetApp systems.  This is MUCH BETTER in Cluster Mode than in 7-mode but getting familiar with all of the HA best practices and nuances around non-disruptive upgrades, etc.

Licensing:

Licensing on NetApp is easy.  Every NetApp differentiator is software-based.  Every model can run every feature and is compatible with every other model while features are installed via license code (Good!).  As a new NetApp owner the thing you need to be aware of is that the licensing is tiered and costs much more on the larger systems.  6200 controllers are a MONSTER jump from the 2240 and will support massive IO but the licensing costs when you add new features will be significant.

Multipating:

NetApp isn't just a NAS, it's also a SAN and performs that task admirably.  Although NetApp has multipathing, it doesn't compare with PowerPath on very large workloads.  In small and mid-range workloads however, where multiple paths aren't saturated, NetApp provides a better performance point per spindle so is worth looking at.

View solution in original post

22 REPLIES 22

aborzenkov
4,291 Views

WAFL is unique in its ability to create large numbers of snaps that impart no performance penalty.

Let's be fair. ZFS (and appliances based on ZFS) can create far more snapshots per file system with the same (lack of) performance impact. 255 snapshots per volume is not really that much, especially on a SnapVault destination. NetApp is showing its age here.

JWHITE_COMPUNET
4,291 Views

I am a big fan of ZFS for home projects.  I've done some really interesting things with it and I look forward to where it goes when it matures.  SnapVault targets could use more than 255 snapshots for sure if they were available.  NetApp promised to do away with some of these arbitrary limits, hopefully they will deliver.

Public