Wednesday, June 10, 2009

NetApp Releases DFM 3.8

NetApp today released DataFabric Manager 3.8, including updates to Operations Manager, Performance Advisor, Protection Manager, and Provisioning Manager. Release includes notable items such as:

   - Deduplication reports and graphs
   - IPv6 support
   - AutoSupport for Performance Advisor

See link above for complete listing of new features and enhancements to existing features.

While not required, NetApp also recommends installation of the NetApp Host Agent 2.7 for SAN and File Storage Resource Manager (FSRM) operations with DFM 3.8.

Friday, May 29, 2009

NetApp Releases SMVI 1.2, Supports vSphere 4

SnapManager 1.2 for Virtual Infrastructure has just published to the NOW site:



SMVI 1.2 is the first release to support VMware's newly-released vSphere 4.  The installation package weighs in at (hefty) 137 MB.

As always, check the interop matrix tool for specific hardware/software requirements.

Friday, May 22, 2009

Armchair Analysis of DDUP Acquisition

Lots of armchair analysis on this week's acquisition announcement can be found all over the Web. Take a look:

From Dan Warmenhoven, CEO, on acquiring DDUP:
http://www.youtube.com/watch?v=Vp_ClYPtVII&ftm=18

to end-user impact:
http://www.informationweek.com/blog/main/archives/2009/05/netapp_buys_dat.html

to long-term viability of NetApp VTL:
http://www.backupcentral.com/content/view/242/47/

...and even a legal investigation surrounding the terms of the transaction:
http://www.globenewswire.com/ca/news_releases.html?d=165921

Happy Memorial Day Weekend!

Wednesday, May 20, 2009

NetApp acquires Data Domain


Wednesday, May 13, 2009

Final Verdict on SnapLock & Dedupe

It's official! Running Deduplication on SnapLock Volumes will not invalidate SEC rule 17a-4(f):

http://media.netapp.com/documents/ar-netapp-snaplock-compliance.pdf

This has been a big debate in the legal world for some time now, with questions such as: "Does dedupe mean that a record has been actually been "altered" by users? The other question: Can one recover a given record back to its original state?

This line (from page 17) sums it up:

"...it is Cohasset’s opinion that the deduplication methodology, particularly the byte-for-byte comparisons performed on each potentially duplicate record block to determine deduplication, does not compromise the requirement of this Rule for maintaining the non-erasable, non-rewriteable recording status of each record file for the designated retention period."

Usual disclaimers apply:

This assessment represents the professional opinion of Cohasset Associates and should not be construed as an endorsement or rejection by Cohasset of the NetApp SnapLock and Deduplication capabilities or any other NetApp products. The information utilized by Cohasset to conduct this assessment consisted of a) oral discussions, b) product requirements c) functional description documents and d) other directly related materials provided by NetApp. It also included information acquired by Cohasset from publicly available sources. Additional information about Cohasset is provided in Section 3 of assessment. The content and conclusions of this assessment are not intended and should not be construed as legal advice. Relevant laws and regulations are constantly evolving and legal advice must be tailored to the specific circumstances of the laws and regulations for each organization. Therefore, nothing stated herein should be substituted for the advice of competent legal counsel.

Case closed!

Thursday, April 23, 2009

Data ONTAP 7.3.1 Certified for IPv6

For those of running IPv6 (we know you're out there somewhere), Data ONTAP is now for you:

Data ONTAP 7.3.1 Certified for IPv6

"There's no place like ::1" (0000:0000:0000:0000:0000:0000:0000:0001)

Wednesday, April 15, 2009

NetApp Updates Five Products

On vacation last week, so just ramping up on these 5 new updates published to the NOW site:


  • SnapManager 2.0R1 for Microsoft Office Sharepoint Server
  • SnapDrive 6.1 for Windows
  • SnapManager for SAP 3.0.1
  • SnapManager for Oracle 3.0.1
  • Rapid Cloning Utility 2.0


  • As always, YMMV. Give me a call if you'd like to test drive any of these updates, understand new features, learn how to demo, etc.

    We're here to help!

    Monday, March 30, 2009

    Metadata Performance in Data ONTAP

    Metadata operations power lots of day-to-day activities on your typical storage array: make/rename/remove directories, timestamps, file permissions, locks, etc.

    In this January 2009 dissertation by Christoph Biardzki, you'll find all sorts of references to NetApp, WAFL, GX, FAS3050, Consistency Points, NVRAM, RAID-DP, FlexShare, etc. as part of their HLRB 2 Supercomputer:

    Analyzing Metadata Performance in Distributed File Systems

    There's some very interesting observations in the above paper ... for example:

    "Because there are no recognizable patterns of WAFL consistency points, it can be concluded that performance is not limited by saturation of the NFS server. When using Lustre and the underlying ext3 file system it is, unfortunately, not possible to reason in the same manner: Flushing the ext3 joural seems to briefly stall all operations, independent of the load on the server. In the case of WAFL processing, a consistency point uses up resources but does not interrupt processing."

    The other important point is the need for retooling mainstream applications, introducing more parallelism in the years to come -- especially considering future deployments of Parallel NFS (pNFS) in the storage stack.

    The future of file storage is parallel. Even your metadata.

    Thursday, March 26, 2009

    New RDM Support in DRA 1.0.1

    NetApp has just released the Disaster Recovery Adapter 1.0.1 for
    VMware's Site Recovery Manager (SRM), now supporting raw device mapping via iSCSI or FCP.

    In case you're not aware, VMware SRM allows you to build, manage, and execute reliable disaster recovery plans. This makes disaster recovery rapid, reliable, affordable, and manageable.

    Don't forget -- FlexClone license is required for test failovers !!

    New NetApp Commercial

    Move over mouse & keyboard, meet the world of touchscreen storage management:

    http://www.youtube.com/watch?v=k_n572kdb2Y&fmt=18

    Seriously, the production work is pretty slick on this one. Check it out !!

    Wednesday, March 11, 2009

    Blatantly Wrong

    I’m completely flummoxed with Chris Mellor’s March 10 article in The Register stating that NetApp has had "difficulties" developing a scale-out NAS clustering product.

    Ha! Ever heard of Data ONTAP GX?

    http://www.netapp.com/us/products/storage-systems/data-ontap-gx/

    Apparently Mr. Mellor forgets that Data ONTAP GX (10.0.4) has been shipping for a number of years now, scaling upwards of 14 PB (yes, petabytes) with 24 nodes, rated at over 1 million IOPS. He also forgets that GX powers several major motion pictures, chip designs, oil/gas simulations, seismic processing, etc.

    How about a little more research next time, Chris?

    Monday, March 9, 2009

    SANscreen 5.0.1

    As part of NetApp's commitment to continuous improvement, SANscreen 5.0.1 was published to the NOW site late last Friday evening (March 6, 2009).

    Most notably, build history & scheduling views in the Data Warehouse Management Portal have been updated to provide more flexibility in creating builds. Custom reporting data models have also been redesigned to provide end-users with the most commonly used fields (capacity, date, BU, Data Center, App, etc.).

    There’s also the usual batch of resolved items as well.

    Overall, a solid point release. Feel free to schedule time in the Arrow Lab to “kick the tires” with this new release !!

    Saturday, February 28, 2009

    New Report on NetApp VTL

    Eariler this week, Enterprise Strategy Group (ESG) published their hands-on test results for the NetApp VTL 6.0 code (including deduplication):

    http://www.enterprisestrategygroup.com/ESGPublications/ReportDetail.asp?ReportID=1174

    There are some key takeaways for tape backup admins looking to move to disk-to-disk backups:


  • NetApp VTL deduplication reduced the required disk capacity for ten full Oracle DB backups from 28.87 TB to only 1.29 TB. This would enable organizations to keep backups on disk for MUCH longer, reducing the need to go back to tape for recoveries.


  • The NetApp VTL provided maximum performance straight out of the box, with no tuning and a relatively small number of disks.


  • ESG Lab confirmed that a NetApp VTL can be deployed in an existing tape environment without changes to policies or procedures and users can be performing backups in less than 10 minutes.



  • Enjoy!!

    Tuesday, February 17, 2009

    Difference Between RAID-DP & RAID-6

    I often get the question...

    "What's the difference between NetApp RAID-DP vs. other RAID-6 implimentations?"

    Answer: Most storage vendor's RAID-6 implementations incur an I/O performance penalty as a result of introducing an additional parity block. RAID-DP is optimized to reduce read I/Os, due to the way NetApp handles parity write operations.

    RAID-10 performance at the price-point of RAID-5.

    Simply beautiful !!

    Monday, February 2, 2009

    NetApp FCoE SAN Details

    I’m hearing that more and more of you are asking for technical details on NetApp’s Fibre Channel over Ethernet (FCoE) solution. Here’s a few answers to your questions:

    As you know, FCoE is a new protocol that transports Fibre Channel (FC) frames natively over 10 Gigabit Ethernet (GbE). NetApp was the first storage vendor to deliver native FCoE storage systems in 2008. More information can be found under PartnerCenter’s Product page (http://partners.netapp.com/mycommunities/PartnerCenter/1047).

    Let’s dive deeper into some of these technical items:


    TOPOLOGY
    --------------------------------------------------------------
    NetApp supports two primary FCoE topologies:

    - Mixed FCoE-to-FC topology:
    FCoE and DCB from host server connections to the FCoE-enabled switch to existing FC SAN switch infrastructures that connect to NetApp 2Gb and 4Gb FC-enabled storage systems.

    - End-to-end FCoE topology (PVR required for FCoE Target Card):
    FCoE and DCB from host server connections to the FCoE-enabled switch to PVR-enabled FCoE storage systems.

    Remember that the FCoE standard requires the use of switches (no direct host-to-storage connections).


    INTEROPERABILITY
    --------------------------------------------------------------
    As of January 2009, NetApp has qualified Microsoft Windows Server, VMware ESX, and Red Hat Enterprise Server with Emulex LP2100x and QLogic QLE8042 CNAs with the Cisco Nexus 5000 series switches. This information is subject to change, so don’t forget to use the Interop Matrix:

    http://now.netapp.com/NOW/products/interoperability/

    In addition, the NetApp Storage System must be a PCIe-capable FAS or V-Series Storage Systems: 3040, 3070, 3100 series, & 6000 series. The FAS2050 is not currently supported with the FCoE Target Card. The FCoE Target Card requires a d-patch for Data ONTAP 7.3.1 (patch not included in mainstream downloads).


    LICENSING
    --------------------------------------------------------------
    FCoE protocol support in the applicable NetApp storage systems is currently enabled via the FC protocol license. Users with a valid FCP license for the storage system may enable FCoE at no additional cost.


    SOFTWARE SUPPORT
    --------------------------------------------------------------
    SnapDrive 6.0.1 for Windows is current supported for FCoE configurations, but SnapDrive for UNIX is targeted for future support. Additional management software (such as Operations Manager) to be fully supported at a later date.

    Of course you can still provision LUNs manually, so applications (such as Microsoft Exchange, SQL Server, Oracle Database, etc.) run perfectly fine with FCoE storage.


    ADDITIONAL INFORMATION
    --------------------------------------------------------------
    Feel free to reach out to your NetApp Pre-Sales Engineering Team here at Arrow ECS. We’re here to help!!

    Thursday, January 22, 2009

    NetApp Number 1 Best Place to Work


    Way to go !!

    http://money.cnn.com/magazines/fortune/bestcompanies/2009/snapshots/1.html


    -Brian

    Monday, January 19, 2009

    NetApp Delivers Data ONTAP 7.3.1, New Best Practices, and Upcoming EOA

    Welcome to 2009!

    Busy January so far. In case you missed it last Friday (JAN 16), Data ONTAP 7.3.1 (IC.1) was posted on the NOW site, as well as the corresponding NetApp University course.

    Check out these new technical reports:


    SnapManager 1.0 for Virtual Infrastructure Best Practices
    http://media.netapp.com/documents/tr-3737.pdf

    iSCSI Multipathing
    Possibilities on Windows with Data
    ONTAP

    http://media.netapp.com/documents/tr-3441.pdf

    SnapDrive 4.1 for UNIX
    Best Practices

    http://media.netapp.com/documents/tr-3735.pdf



    Finally, don't forget that the last day to order SnapMirror for Open Systems (formally, ReplicatorX) is February 2, 2009. More info here: http://www.netapp.com/us/company/news/news-rel-20081209.html

    Monday, November 17, 2008

    NetApp Toaster

    I'm honored to be an official NetApp Toaster now:

    Technical Partner Enablement is a HUGE advantage towards Clear #1 here at Arrow.

    2008 is the first year that NetApp has offered the Distinguished Partner SE award to the Channel.

    Appreciate all the kudos. Rock on !!

    Wednesday, March 19, 2008

    Fun with Microsoft's iSCSI Software Initiator Versions

    Eariler today, doing some demo prep work. Part of the show includes Windows LUN expansions thru SnapDrive 5.0 for Windows Server 2003 SP2. Never got around to updating my Microsoft iSCSI Software Initator from 2.04 to 2.06.

    Well, well, well...

    Come to find out, Logical Disk Manager Administrative Service unexpectedly terminated. Then there's SnapDrive Vdisk "Service execution status" errors in the debug logs (HRESULT 0x1).

    Interestingly, the actual ExpandVirtualDisk operation works in MS iSCSI init 2.04 -- you see the volume size increase. The only user-level issue is a little popup error.

    Apparently, it's the last step before refreshing VDS. Users can take advantage of the extra LUN space, but dmadmin.exe (Logical Disk Mgr Admin Service) goes south (access violation) and VDS is now in an "incorrect" state -- funkytown !! FYI -- VDS error is: "Unexpected provider failure. Restarting the service may fix the problem. Error code: 80042420@0200003".

    Also of interest -- I don't recall this problem under Service Pack 1 of 2k3. My hunch is new and/or reworked code under VDS and Service Pack 2.

    Bottom-line: Your best bet is to ensure your SCSIport.sys (for iSCSI) is up-to-date (2.06) w/ SnapDrive 5.0.

    Tuesday, February 26, 2008

    Slinky 2.0 (or, Adaptive Storage Networking in a 10 Gigabit Ethernet World)

    It's Slinky, it's Slinky, for fun it's a wonderful stack. It's Slinky, it's Slinky, it's fun for an ACK and a NAK!

    Between FAST '08 and this week's launch of Windows Server 2008, I've been thinking a lot about self-tuning performance/scalability of storage networking stacks -- especially with the (eventual) adoption of 10 Gigabit Ethernet in data centers.

    For example, the (new) Next Generation TCP/IP stack -- or "tcpip.sys" in Windows Server 2008 -- renders several (older) Windows Registry settings obsolete.

    Same is true for Brocade's Fabric OS 6.0 and their whole Data Center Fabric (DCF) initiative: Being able to identify I/O patterns, QoS volumes, and perform traffic isolation are all necessary to simplify storage management.

    Also recently discovered an interesting research paper (from NetApp's R&D unit, ATG):

    http://hssl.cs.jhu.edu/~randal/papers/batsakis_lsf07.pdf


    Storage I/O stacks have lots of growth potential over the next decade -- at the host layer, within the fabric, as well as the storage array itself.

    Stay tuned for my continuing deep-dive discussion on this topic!

    Thursday, February 14, 2008

    NetApp: Free Cash Flow King

    I'm not a finance geek, but find lots of interesting parallels between IT technobabble ("geekspeak") and finance jargon.

    Anyhow, fairly detailed analysis of NTAP's valuation, as well as crunching P/E ratios, net margins, free cash flow, etc. Big takeaway here is a great story for investors, as well as IT managers looking for their next storage system.

    See it all here:

    http://seekingalpha.com/article/64605-network-appliance-cash-is-king

    Bottom-line: Rewards outweigh the risks for NTAP.

    Sunday, February 3, 2008

    New NetApp Patent: boot_diags on CompactFlash

    In case you missed it amid the Yuletide holiday festivies, today's light reading assignment is NetApp's latest United States Patent # 7,310,747 B2:

    SYSTEM AND METHOD FOR DIAGNOSTICS EXECUTION AND DATA CAPTURE IN A STORAGE SYSTEM USING NONVOLATILE MEMORY

    In other words, running boot_diags off of CompactFlash.

    As a former IT guy running NetApp arrays, I've seen some pretty interesting environmental diag codes over the years. Of course, a 100,000,000 volt lightening bolt does funny things to your power sensors.

    Saturday, February 2, 2008

    SnapManager and your Bank's ATM

    Speaking with several NetApp prospects over the past few weeks, one of the top-of-mind issues for 2008 is integrating data management with their company's applications. Typically, most vendors will attempt to "simplify" the problem by cataloging the entire infrastructure with some sort of storage resource management (SRM) software.

    Won't that work?

    From my experience as an IT guy, SRM tools tend to use the infrastructure as a proxy to manage data, thus forcing me to manage my little SQL Server database by also managing the other hundred or so components (LUNs, replication relationships, etc.). The SRM assumes I'm a storage admin that understands the underlying disk subsystem -- Ouch !!

    So NetApp's integrated data management model lets the application admins work with their data, while maintaining supreme control of the underlying storage subsystem.

    It's all about automation and application-awareness. We call it SnapManager -- comes in several flavors (pick your fave biz app).

    SnapManager is similar to your bank's ATM: Your bank can control your ability to withdraw cash at any point, not to mention the amount and any associated fees. With ATMs, we reduce the amount of cash withdrawn on any given day (thus, maintaining a healthy fractional-reserve model with quality asset control).

    Both ATMs --and-- SnapManager improve the customer experience -- so, if I need cash at 2 am on a holiday, I don't have to wait for the bank to open the next business day. Same thing applies to my business: If I'm a DBA and I need a database restore, I don't have to wake up my storage admin. I do it myself.

    Or, as I like to say, we're "virtualizing" the storage admin.

    Thursday, January 3, 2008

    NetApp MPIO DSM vs. Microsoft's Generic DSM

    Here's a great question from one of my partners:

    "What are the pros and cons of using the built in MPIO driver (DSM) from Microsoft versus the NetApp MPIO?"

    For some background on MPIO and DSMs, see James Burke's excellent post here:

    http://blogs.netapp.com/whu_who/2007/12/introduction-to.html

    Now, to answer that question...

    The Microsoft generic MPIO DSM (device-specific module) does not, by itself, have all the advantages of the NetApp Data ONTAP DSM for Windows MPIO. For example:


    -- NetApp's MPIO DSM is more stable at lower levels of the I/O stack (not all vendors enumerate & uniquely identify devices the same way),

    -- Improve performance by using NetApp's MPIO for an active/active LUN load balancing policies (Round Robin, Auto Assigned, etc.) ... thus, keeping track of IOs arriving order & sending the requests on the wire in the proper temporal order (extremely complex algorithm), &

    -- NetApp MPIO also supports more granular (configurable) failover & path recovery for NetApp LUNs.


    Microsoft's approach to licensing their MPIO kit to storage vendors (like NetApp, EMC, HP, etc.) keeps them out of the business of coding & testing all the different low-level APIs from each vendor.

    Bottom-line: The generic Microsoft MPIO offers basic path redundancy, while NetApp's DSM for MPIO gives your customers greater performance, load-balancing, & failover capabilities.

    More NetApp Blogging in 2008

    Okie-dokie. I've been super-busy throughout the Summer and Autumn of `07. Got some serious catch-up to do.

    Yes, before anyone emails me ... Onaro's SANScreen is now NetApp SANScreen. Definately big story of the day. More on that topic at a later date.

    In my next entry, going to run thru an interesting host-based multipathing question I receive from time-to-time. I'll provide "MPIO 101" links for background material.

    Saturday, June 9, 2007

    A Little Bit of Everything

    - Well, here's what a DR rookie failure looks like on a grand scale. Once again, your tax dollars at work !!

    - Will you be the first one on your block with one of these? User interface design beyond the desktop metaphor.

    - NetApp's V-Series Business Unit (VBU) has qualified 3PAR's InServ against V-Series code.

    - Sanjay from NetApp joins the blogosphere (Databases on NetApp)

    - Very savvy IT team this week quizzed me about Windows Server I/O performance between drive letters vs. volume mount points (BTW -- no difference, since VMPs are just filters that live in kernel mode); Even Microsoft and various storage vendors do performance benchmarking with volume mount points (mainly, Exchange & SQL Servers). Did you know that the IIS Team actually designed "HTTP.sys" as a kernel-mode driver in 2k3 and 2k8 Server for high-perfomance web hosting?

    - Yes, watching last evening's STS-117 launch on HDNet (1080i on my Mitsubishi 62" DLP) is still amazing (last September's STS-115 launch was just a beautiful)

    Happy Weekend !!

    Friday, May 25, 2007

    Top 10 Dying Data Storage Technologies

    After reading Computerworld's recent article, Top 10 Dead (or dying) Computer Skills, I decided the world was missing a listing of the top 10 dying data storage technologies (in no particular order):

    Without further ado...


    Top 10 Dying Data Storage Technologies in 2007


    1 - betamax
    2 - optical jukebox
    3 - LRC modules
    4 - floppy disks
    5 - NAS products based upon the Windows kernel
    6 - 1 Gbps fibre-channel
    7 - iFCP
    8 - compact cassette
    9 - laserdisc (LD)
    10 - legacy filesystems (think Tru64's AdvFS, NetWare File System, etc.)



    NOTE: I decided not to include obsolete ones (punch cards, 8-track tapes, etc.), or my predictions for the future death of currently shipping technologies.


    I also avoided specific products (Compaq HSG80) and specific array internals (legacy NetApp gear with Alpha processors).


    Have a great Memorial Day Weekend...

    Tuesday, May 22, 2007

    Do Your Baseline Transfers Use FexEx?

    Even in this age of "Web 2.0" and FiOS (see Verizon), it's amazing that Internet bandwidth just isn't suitable for terabytes of data transfer.

    See here: FedEx Still Faster than Internet.

    Of course, when using SnapMirror or SnapVault, you could easily use the LREP tool (for QSM). With VSM, you could seed it via tape or disk offline.

    Even the SnapMirror Best Practices doc states, "...but the initialization transfer (baseline) can be crippling in a WAN environment". Fortunately, easy-to-use bandwidth throttling controls exist! ;)

    Call me sometime and I'll tell you my Internet2 story -- seriously!!

    Thursday, May 10, 2007

    NetApp Gets Perpendicular

    In case you haven’t heard, earlier this week NetApp formally announced the availability of 750GB SATA disk drives for selected FAS storage systems (part number X268A-R5, in case keeping score ;-).

    What’s really neat here is the fact that instead of recording the bits horizontally – with the (traditional) longitudinal technology -- they’re being recorded perpendicular (bits are aligned vertically). This means when all the little bits lie down (as they should under longitudinal drives), they’re going to get pretty cramped and eventually do flips (otherwise known as the Superparamagnetic effect).

    Of course, perpendicular recording is just the tip of the iceberg -- research into patterned media, new read/write heads, inventing more advanced algorithms to interpret magnetic signals is just a few of the many ways to further advance the ever-evolving hard disk drive.

    Here’s my prediction: 1-terabyte (TB) SATA drives before the end of this decade!!

    Saturday, May 5, 2007

    Back from Vacation

    Just getting back into the swing of things from our Disney World vacation (which by the way, Walt Disney Animation Studios used FAS6000-series systems to produce the recently released movie, Meet the Robinsons).

    Anyway, 90-degree weather in the month of April is simply brutal. Luckily, the (fresh-squeezed) Florida orange juice is still world-class.

    Now for something else totally off-topic: I've always been fascinated in applications of cerebral ("brainwave") technology. Specifically, neuroacoustical applications. So, over the last few months, I started listening to Kelly Howell's "Deep Sleep". I've continued with the 2006 release of the "Gamma Meditation System" by Dr. Jeffrey D. Thompson. The results are simply amazing!!

    Google "resonance entrainment" or "nonlinear oscillations" for more details (another example of applied mathematics in the real world).

    More forthcoming later today...stay tuned!!

    Friday, April 20, 2007

    The Good, Bad, & Ugly of Email Archiving

    Quite often, I get to the question of "so how do you archive your email?" While responses differ, at least once a week, the answer is..."Oh, we request users archive their email using Outlook Personal Storage (or, .PST files)."

    Woah, time-out!

    The good news is that this problematic situation is very common and well-known. Bad news being that PST files on LAN/WAN is totally unsupported by Microsoft (see: http://support.microsoft.com/kb/297019).

    The next question comes from the techies in the crowd: "How come? What's the big deal?" My explanation goes like this:

    Pretend you work at a 1,000-person company. Now, let's say that half of them use PST files as their email archival with an average size of 1 gig. Now consider what happens when all 500 PST users launch Outlook when they arrive to work in the morning.

    500 x 1 = 500 GB of disk & network I/O to process simultaneously. All sorts of queuing issues -- the Windows Server Service allocates various worker threads from the kernel's Non-Paged Pool (NPP).

    Then, I/O requests are sent to the disk subsystem (let's say it's a NetApp FAS system). What if the disk subsystem does not respond in time? All incoming I/O requests are queued as "work items". Since there are finite memory resources, NPP will (eventually) run out, log a System Event ID 2019, and hang the server.

    You can actually view this steady depletion with perfmon with the "Available Work Items" counter. If you're AWI is close to 0 (zero), clients may be dangerously close to not being able to access their files.

    Sure, you could tweak to mitigate some of these negative issues, but as you scale upwards, the server itself may not be handle to workload and your house of cards falls down.

    This problem could be solved in different ways: You could...

    - use an Offline Storage file (or .OST),
    - deploy a Microsoft Terminal Services farm,
    - implement a third-party email archiving app,

    ...and back it all up with NetApp FAS systems. Remember this one key point:

    PSTs are not an enterprise storage solution !!

    Friday, April 13, 2007

    Green Spindles

    You may have not noticed one of the areas of research mentioned within the recent NetApp white paper, "Reducing Data Center Power Consumption Through Efficient Storage". While I believe most power savings come from software features, there's some pretty serious R&D work being put into intelligence control of the speed of individual drives in response to demand.

    According to EUN (Energy User News), the power requirements of today’s data centers range from 75 W/ft^2 to 150-200 W/ft^2 and will increase to 200-300 W/ft^2 in the near future.

    Look -- everyone's goal here is to conserve energy without sacrificing performance. But right now, enterprise storage systems are fairly "dumb"...basically consisting of 3 power states:

    - active
    - idle
    - off

    In active mode, the platters are spinning and the head is seeking or actively reading/writing. In idle mode, a disk is spinning at full speed but no disk activity is taking place. There's no "sleep" mode for enterprise storage system (unlike your laptop or desktop). Of course, spinning up from being powered-off incurs significant time and energy overhead at the moment.

    Come to think of it, incandescent light bulbs are a lot like today's typical enterprise disk drives -- an inefficient dinosaur.

    So imagine your storage system tends to be very underutilized between the hours of 6 pm to 12 am and 4 am to 8 am. The system would gather these metrics, analyze historic trends, and reduce the speed (or power down) specific disks during those off-peak hours in a bimodel distribution of busy & idle disks.

    There's even research being put into "power-aware" RAID designs and extending hybrid hard drives into the enterprise space.

    Wow, putting this together with using RAID-DP on ever-larger SATA drives, thin provisioning, FlexClones, Snapshots, de-duplication, automated data lifecycle policy engines, and consolidation of servers save some real dough.