February 22, 2012

NetApp Introduces New Controllers

One of the more popular controllers (at least, in my experience) that NetApp offered was the FAS250/FAS270.  From the front it looked like a standard DS14 disk shelf, but in the back it contained controller module(s).  You could deploy it as a single controller, or in a HA pair and it made a great option for SMBs – it didn’t consume a lot of space since the controller was built into the shelf and the price was entry level.  These models went away in favor of the 2000 series controllers which still offered internal drives (12 in the FAS2020/2040 and 20 in the FAS2050) but they used their own form factor and were a dedicated storage appliance.  A problem with these controllers was that when you upgraded the system, the internal drives could not be taken out and put into a disk shelf (such as the DS4243).  There were ways around this limitation, most notably just not ordering internal drives in the controllers and only using external storage to ease any concerns over future upgrades, but it still was a limitation in my opinion.

Today NetApp is announcing two new controller models that are reminiscent of the FAS200 line.  They are the FAS2240-2 and the FAS2240-4, they are 2u and 4u in size respectively.  Early performance numbers indicate between a 2-3x performance improvement over the FAS2040 depending on workload type. The FAS2040 will stick around to complete the FAS2000 lineup.  This means all current controllers will be able to run the latest ONTAP software from NetApp (the FAS2020 and FAS2050 did not support ONTAP 8.x).  As I alluded to previously, the FAS2240 is a storage shelf with the controllers inserted into the back.  The 2240-2 is a 2u system and based on the current FAS2246 SAS shelf, while the 2240-4 is a 4u system and based on the current FAS4243 shelf.  The FAS2240-2 utilizes 2.5″ SAS drives and supports either 450 or 600GB drives as of today.  The FAS2240-4 utilizes 3.5″ SATA drives and supports 1, 2 or 3TB SATA drives as of today.  Both systems can be ordered with either 12 or 24 drives.

Some quick notes on the new models:

  • Will require ONTAP 8.1+
  • Supports a mezzanine card, which can be either a 2 port FC card or 2 port 10 GbE card
    • If you put a FC card in the mezzanine slot, they can be either target or initiator ports much like onboard FC ports on other controllers today
  • Will support cluster mode, but you have to use the 10 GbE mezzanine card for cluster communication so only iSCSI/CIFS/NFS will work and must be served out of the GbE ports
  • Will come with ONTAP Essentials, which means all storage protocols are included (as well as things like Operations Manager, Protection & Provisioning Manager, DSM/MPIO)
  • Ability to convert from a controller into a disk shelf (much like the FAS200 line)
  • Will not have support for the FlashCache card or FCoE

Front view of the FAS2240-2 controller

Rear view of the FAS2240-2 controller

Front view of the FAS2240-4 controller

Rear view of the FAS2240-4 controller

Also, and in my opinion this is a big one, the maximum volume size is 54TB on the FAS2240 and the maximum volume size with dedupe and/or compression enabled is ALSO 54TB!  This is one of the best features of 8.1, 64 bit aggregates in 8.x allowed us to grow beyond a 16TB aggregate limit but we were still limited (at least in some environments) on the volume size to a maximum of 16 TB (depending on the controller model) when using compression and/or dedupe.  To reiterate, as of ONTAP 8.1 the maximum volume size for dedupe/compression is now equal to the maximum volume size for the controller – which means it could be anywhere from 30 TB on the (now) entry level FAS2040 to 100 TB on the highest end FAS6280.  To determine what the maximum volume size is for your controller, check the System Configuration Guide.

Anyone out there looking at these new controllers from NetApp?

Popularity: 14% [?]

Some Notes on the New Controllers and Shelves from NetApp

Here are my notes on some of the new product announcements from NetApp, notably the 3200 and 6200 series controllers, and the DS2246 disk shelf.

ONTAP 8.0.1

  • Will have a GA status in March 2010
  • Will include data compression for volumes
    • Data is compressed as it is written to disk
    • Enabled on a per-volume basis
      • Compression scanner can be used to compress existing non-compressed volumes
    • Can be used with deduplication to reduce footprint even further
    • Works by utilizing compression groups, these are 32KB in size and made up of 8 4KB blocks
    • Use cases are data that is not performance sensitive, archival data, backups etc
  • Unified Connect
    • Use a CNA to connect to a 10GbE switch to provide FCoE, iSCSI, NFS and CIFS
  • Will support all of the new systems
  • Data Motion for Volumes
    • Non-disruptive volume movement for SAN
    • Works between different drive types: SSD/FC/SAS/SATA
      • Note: SSD and SATA are not recommended on the same controller
    • Movement within the controller currently
    • Movement between the same aggregate type, e.g. 64 bit to 64 bit

6200 Series Controller

  • Up to 3PB capacity and 2x more PCIe connectivity
  • Built-in 10GbE, 8Gb FC and 6Gb SAS
  • Double the current performance of the 6000 series
    • Improved performance comes from hardware upgrades as well as ONTAP 8 code
  • PCIe slots are now v2.0 x8
  • 2x slots compared to the 6080

6280

    • 1440 drives, 2880TB capacity, 8TB flash cache supported
    • 24 processor cores
    • 192GB RAM
      • As of 8.0.1 only 96GB will be used, future software update will make use of the entire 192GB RAM
    • 8GB NVRAM
    • 24 maximum PCIe slots
    • 8 or 16 onboard 8Gb FC ports
    • 8 onboard 10GbE ports
    • 0 or 8 onboard 6Gb SAS
    • 100TB maximum aggregate size
      • 16TB maximum if using deduplication or compression

6240

    • 1440 drives, 2880TB capacity, 6TB flash cache
    • 16 processor cores
    • 96GB RAM
    • 8GB NVRAM
    • 24 maximum PCIe slots
    • 8 or 16 onboard 8Gb FC ports
    • 8 onboard 10GbE
    • 0 or 8 onboard 6Gb SAS
    • 100TB maximum aggregate size
      • 16TB maximum if using deduplication or compression

6210

    • 1200 drives, 2400TB capacity, 3TB flash cache
    • 16 processor cores
    • 48GB RAM
    • 8GB NVRAM
    • 8 maximum PCIe slots
    • 8 or 16 onboard 8Gb FC ports
    • 8 onboard 10GbE
    • 0 or 8 onboard 6Gb SAS
    • 70TB maximum aggregate size
      • 16TB maximum if using deduplication or compression

3200 Series Controller

  • 50% more PCIe connectivity
  • Up to 2PB storage capacity

Expansion Module (IOXM)

  • 3u expansion module

Shelves

DS2246

    • 24 x 2.5″ Drives in a 2u enclosure
    • 10k RPM SAS in 450 or 600Gb (No SATA)
    • Uses a 6Gb SAS interconnect
      • 24 Gbps per SAS port (6 Gbps x 4 lanes)
      • Existing PCIe 4 port SAS HBA already supports 6Gb/sec
    • DS2246 and DS4243 shelves must be used in separate stacks

SSD

  • Uses the existing DS4243 shelf
  • 24 x 100Gb SSD
  • Supported in 3160/3170, 3210/3240/3270, 6040/6080, 6210/6240/6280
  • Requires ONTAP 8.0.1
  • Requires multi path HA cabling (MPHA)
  • Should be SSD and HDD stacks on separate SAS adapters to ensure full bandwidth available to SSD IO
  • MUST be a full shelf, cannot mix and match drive types
  • Flash Cache vs SSD
    • Flash Cache is good for:
      • Random read intensive workloads
      • Hot data is dynamic or unknown
      • Server/desktop virtualization, file shares/home directories, databases
    • SSD is good for:
      • Workload is random IO intensive
      • Every read must be fast
      • Databases for mission critical applications

My Take

  • the IOXM module is a welcome addition, seemingly every 3100 system that went into the field was slot limited if the customer was considering Flash Cache or 10GbE
  • Big difference in the lower models in each series, 3210 and 6210
  • Will be interesting to see how well compression works, a number of customers have expressed interest already but is the performance penalty worth the saved space?
  • SSD shelf is a good addition especially with non-disruptive volume movement, if the database gets pounded at the end of the month the storage administrator could non-distruptively move it to SSD and then move it back to a lower tier once that work is complete
  • Speaking of data motion, it’s great to see it but it’s still somewhat limited – currently it must be in the same controller and a similar aggregate type.  It would be nice to be able to move a volume from one controller head to the other if one was under heavy load
  • Also along those lines, will any future update to Operations Manager allow you to automate data motion similar to VMware’s distributed resource scheduler(DRS)?

Popularity: 20% [?]