Thursday Jun 26, 2008

Open Storage: Early Customers

Now let's talk about what really matters when it comes down to Open Storage - customers.  Now, forgive me for repeating myself because I mentioned some of these customer testimonials in my first Open Storage blog.  But it just so happens that the ones I blogged about ended up in the Open Storage Adoption White Paper.    So, I'll talk about them (again), but also throw some new ones in... 

Customer Blog Comments:  Before I do, I want to point to a couple of blog comments.  The first one was submitted to Chuck Hollis' EMC blog post Do-It-Yourself Storage.  After Mr. Hollis finished criticizing Open Storage, an experienced storage and IT customer posted this comment in response to his critique:  

"I think you are missing the point. I did actually a price comparison. Building your storage solution (in my case several hundred TBs) from cheap disks using x86 servers + ZFS + Open Solaris + Solaris Cluster + ZFS, where all software is not only open sourced it is also for free, \*does\* make a huge difference. We actually started building our solution on EMC Symmetrix (great box) and EMC Celerra years ago and endup on really cheap storage + ZFS as a replacement and a way to move forward. Additionally all features like snapshots, cloning, end-to-end checksuming, remote replication, built-in compression, built-in cryptography, NFS, CIFS, iSCSI, ... are also for free. Better - they work exactly the same regardles what cheap storage or server we put underneath.

What ZFS brings to the market is the open sourced and free Google like approach to storage - how to cheaply build reliable storage from small to large scale installations.

Sure, especially for SMB market, what is needed is an easy GUI interface built on-top of Solaris + ZFS. I'm sury you will see one sooner or later."

I like the "Google like approach to storage" comment (and the fact that a disk + ZFS system replaced an EMC DMX and Celerra system for this particular customer!)  I also had a customer post these comments to my blog -  they just deployed a 2TB OpenSolaris ZFS + COMSTAR storage system for a VMWare Cluster with off the shelf components saving €2,000 in the process.  Cool stuff.   

Open Storage Customers:  Not surprisingly, a lot of these early adopters use open source to compete in their respective businesses.  You may also expect that early Open Storage adopters would come from Sun's Solaris install base - while true, many new customers do NOT come from Sun's current base.  In fact, a lot are Linux users.  They have chosen Sun because of...Storage. (Open Storage to be precise)

DigiTar provides messaging security and processing services over the Internet (antivirus, antispam, antiphishing, firewall, and archiving).  DigiTar is using Open Storage to improve the performance and efficiency of their database servers. They are using Sun's X4500 storage servers and ZFS to automate database storage administration - with ZFS they have reduced the identification and fixing of database corruption by days and/or weeks.   They are also an active member of Sun's OpenSolaris community and use the OpenSolaris community and SunSpectrum for tech support.

What I personally love about DigiTar's story is that they were (still are) a Linux shop.  So what made them a Sun OpenSolaris customer???  Storage!

Why? Read their CEO/CTO's blog:  Democratizing Storage.  He basically states that OpenSolaris was brought into the company because it made for a superior storage platform.  The clencher (and essence of Open Storage) for me is when he compares deploying Open Storage vs. a traditional storage architecture: 

"To replicate the level of redundancy we get with two X4500s, we’d have to install two completely separate storage arrays…not to mention also buy two very large beefy servers to run the databases. By using X4500s, we get the same reliability and redundancy for about 85% less cost. That kind of savings means we can deploy 6.8x more storage for the same price footprint and do all sorts of cool things like:

  • Create multiple data warehouses for data mining spam and mal-ware trends.
  • Develop and deploy new service features whenever we want without considering storage costs.
  • Be cost competitive with competitors 10x our size."

Want do do more with less?  DigiTar is with Open Storage...

Nexenta is unique in that they are an Open Storage customer - but they are using Open Storage to build and sell storage products of their own.  They bill their product as “Enterprise-class data storage for everyone!” Nexenta has built its NexentaOS and NexentaStor software appliance from OpenSolaris and ZFS; and they can deploy it over the Sun Fire X4500 as well as other HW.

Like DigiTar, Nexenta was founded by Linux gurus.  In fact, the Nexenta team developed the iSCSI stack that was adopted by the Linux community. So what platform did these experts in storage and open-source software choose to build a new storage offering?   OpenSolaris & ZFS - due to its advanced storage functionality and long history in enterprise environments.  

So now the company offers NexentaStor - a software-based NAS and iSCSI solution with unlimited incremental backups, snapshot mirroring and the inherent virtualization, performance, thin provisioning and ease of use benefits of ZFS.

Sapotek Inc.
Sapotek Inc. is a Software-as-a-Service (SaaS) provider that offers online desktop services to (currently) 200,000 users worldwide.  This SaaS provider had a classic Web 2.0 storage problem - how to massively scale, efficiently and affordably? 

Sapotek was running Red Hat Enterprise Linux on Dell servers and had been maxing out at five concurrent threads per server. So they migrated to Sun Fire X4200 servers and the Sun Fire X4500 storage server running ZFS.  They replaced four (4) Dell/EMC storage systems with one (1) Sun Fire X4500.  Sapotek also used Sun's ZFS snapshot feature, and reported that it had reduced backup and recovery times by 99%!  (From hours/days to minutes).  Here is what their cheif tech officer,  Oscar Mondragon, said about Sun Open Storage: 

"The ZFS file system feature of the Solaris 10 OS is a marvel. It creates a common storage pool where all storage performs as fast as if it were local. Our administrators can grow, add, or remove storage on the fly in a single step. Just two people administer 24 TB."

Above is just a sampling of early customers that have deployed components of Sun's hardware and software Open Storage portfolio.  Read the Open Storage Adoption White Paper for more case studies or my previous blog post.  Additionally, take a look at these customer testimonials (Some have bought commercial versions of Sun's storage offerings and some have deployed open source offerings)

  • Dow Corning Corporation: Increased storage capacity by 50% with ZFS
  • Joyent: Sun Fire X4500 gave them a hosted storage business at $1/GB as opposed to $6-$7/GB for EMC
  • Nomura Bank: Used ZFS to create a common storage pool, with no partitions to manage. Admins can now provision or grow storage, and add or remove a file system with a single command.  ZFS also safeguards data at the bank by running 64-bit checksums
  • OmniTI: During a catastrophic accident when 1.8 TBs were lost, they restored their entire database in just seconds with ZFS
  • University of Calgary: Deployed SAM-QFS software for 229TBs of storage, moves data between Sun StorageTek Disk and Tape via policy
  • University of Oxford: Their digital library project will manage 9 million library items and support an average of 9,000 library resource requests a week
  • IN2P3/CNRS: Optimized datacenter space and reduced energy consumption with X4500

Tuesday Jun 17, 2008

Open Storage: Vendor Landscape

Ok - chapter 5 in the Open Storage Adoption White Paper talks about the vendor landscape.  We'll start with Sun.  

Sun's Open Storage Differentiation
There are three areas that position Sun as the best partner for Open Storage solutions:

  1. Innovative HW systems: Sun’s hardware differentiation lies in design innovation.  Three examples of this are the Sun Fire X4500 (Thumper) which combines a four-way x64 server with 48 TB of SATA disk in a 4U rack space - one of the most efficient and dense storage servers on the planet (and there is more to come).  Sun's unique ST5800 (Honeycomb) archive platform scored perfect 10s in reliability and scalability with InfoWorld.  On the server side, Sun's Blade 6000 is the most open blade platform in the industry - delivering Solaris, Linux, Windows or VMware on single and multicore processors by Sun, AMD, and Intel in one chassis.
  2. OpenSolaris as a storage platform:  One of the most robust and reliable OSes in IT.  Sun also offers advanced open-source file systems including NFS, the upcoming Parallel NFS (pNFS) and ZFS.  ZFS can manage zettabytes of storage and offers data services including volume management, data integrity and software RAID.  
  3. Open-source storage applications: Sun has now open-sourced more high-level storage application software than any other storage vendor.  Applications like remote-mirror-copy and point-in-time-copy.  Take a look at Sun's complete open-source, end-to-end storage portfolio (a storage developer's dream):

Other "Open Storage" Efforts

In keeping with the definition of open source software + industry standard HW = Open Storage (aka a more scalable, economic storage architecture) let's look at what other vendors are doing with open source software and industry-standard hardware.  But remember, several vendors use open source software and industry standard hardware - but still limit customer choice and charge higher rates in their implementations...

IBM: In terms of industry-standard hardware, IBM sells Intel and AMD servers as well as SAS- and SATA-based disk and JBOD systems. IBM does see value in open source as it is a large Linux supporter. (However, Sun has more than 3,000 members and 30 open-source storage projects in development for OpenSolaris AND has even open-sourced its commercial applications like the Sun StorageTek Availability Suite - giving its customers full and affordable access to its own IP).  IBM’s recent investment in the storage market has been its recent acquisition of XIV - see IBM buys XIV - good move or bad?  XIV NEXTRA does use industry-standard hardware, but its software is proprietary (not open source).  XIV shows IBM has realized customers need more than what traditional disk products offer today - the design points of the XIV architecture are low cost and massive scalability.  However, the technology is new and IBM’s claims of low cost are yet to be determined.

HP: HP also sells Intel and AMD processor-based servers; as well as SAS, SATA and SCSI JBOD arrays.  HP sells ProLiant servers or industry-standard servers running Windows Storage Server (Not open source, but obviously a high-volume OS).  HP acquired PolyServe in 2007 to cluster its storage and server systems.  HP recently announced its HP StorageWorks 9100 Extreme Data Storage System (ExDS9100) - a large NAS appliance with an entry configuration of a whopping 246TB of standard disk.  HP also broke away from Windows in its implementation - using open-source Linux as the platform for this higher-capacity platform.   HP says it will ship the ExDS9100 by year's end, but without some features like CIFS support.  (And by the time the ExDS9100 ships, Sun's X4500 will have been on the market for over 2 years with over 250 petabytes installed.)

EMC: EMC primarily offers closed systems today - custom components and software that are available only through EMC.  But EMC can identify business/IT trends and adapt to them. In January 2008, EMC announced its first Storage as a Service (SaaS) or “Cloud” storage offering.  EMC is also investing in two products code-named “HULK” and “MAUI.”  MAUI is software that will provide what EMC calls a “global repository” - but not much in known at the time of this blog.   According to this blogger, HULK's official name is EMC InfiniFlex and a single system consists of a full 44U rack with up to 300 drives (10 disk trays x 30 disks).  They fit these 30 drives in 3U it looks like (and does the front-to-back cooling suggest they stick the extra disk behind the front-end disk?)  It also uses 12 dual core, 1U servers as well as 2 ethernet switches.   So, it essentially looks like industry-standard HW (storage & servers) with TBD software.  HULK and MAUI may be EMC’s first venture into the open-storage space - especially if the systems are able to work with other, third-party, industry-standard components. But the benefits to EMC’s new offerings, and just how “open” they are, are yet to be announced...

NetApp: NetApp sells their own proprietary hardware and also develops their own custom operating system called Data OnTap (while others in the industry have been moving to an open or high-volume operating system for storage - like Solaris, Linux or Windows.)  NetApp does not open-source its storage operating system software.

Dell: Dell has built its business on industry-standard, volume-based products.  Dell lets customers configure servers with industry-standard Intel and AMD processors, SATA disk drives and various Linux distributions. Dell can be credited for its online configuration and ordering services; but Open Storage customers really require enterprise-class software, services and tech support from their Open Storage vendors - like Sun offers

Next Blog...
Open Storage Case Studies

Wednesday Apr 23, 2008

HP's Upline goes Offline

HP Upline is HP's new SaaS offering for online storage, backup and data migration services - from HP's recent acquisition of Opelin.  They offer "unlimited storage" for $299 for year 1 or $599 for 2 years.

Unfortunately HP Upline crashed just a few days after it was launched.   To be fair, this stuff isn't always easy - we had several obstacles to overcome with our own SaaS compute service.

However, HP probably didn't count on an active EMC blogger as an early adopter (openness has its pros and cons).  EMC's Storagezilla posted a blog with HP's notice to customers about the crash.   EMC's own SaaS storage service, Mozy, wasted no time on capitalizing on HP's crash in true EMC fashion - launching a Google text ad titled "Shafted by Upline?" and "Is Upline jerking you around?"

Another significant point Mr. Zilla points out is that the current SaaS leader, Amazon Web Services' (AWS) total revenue for 2007 was $100M.  

Bottom line:  The market and storage industry is adopting SaaS - but the market is still new and emerging.  Like most trends, SaaS won't take over the world - but the datacenter mix will change, evolve over time.   And while simple backup technologies and strategies are not as sexy as new trends like Web 2.0 or SaaS - a simple backup strategy will still have its place in the new world. 

Monday Mar 10, 2008

Game-changing Storage Economics

In the open systems VTL space, EMC is the leader in market share.  (Sun leads in the Mainframe space btw).  How can Sun compete?  Answer - By changing the Economics at the infrastructure level.  See below: 


H/W Platform
Power Consumption
Rack Size
 24 TB  FalconStor
 1 Server, 48 SATA drives
 1,315 Watts
Sun VTL Value
 24 TB
 FalconStor  1 X4500
 1,100 Watts
(16% less)

(73% less!)

This is an example I use because both EMC and Sun (and IBM for that matter) use the same software - FalconStor

So, wouldn't you take a closer look at a product that offers similar functionality at 16% less power consumption in 73% less space? 

Thursday Feb 28, 2008

Sun's Open Archive Announcement

If you've been walking the halls of Sun StorageTek of late, you would have heard a lot of talk about the "Archive Launch" and changing IT and storage economics...

Today, Sun made a large announcement in the Archive storage space.

First a word on messaging:  Internally, Sun Systems recently went through a healthy reality check on how we message our products and solutions.  We looked at where we are in the industry and where we can, and should, differentiate.  It's no secret that Sun's core assets reside at the infrastructure level - storage, servers, processors, O/S.  These segments are the backbone of IT infrastructure - on which applications are deployed to meet business requirements and goals.  We have come to a single conclusion in which today's (and tomorrow's) messaging will focus on - the Economics of IT needs to change.  With data sprawl, longer retention periods and a paradigm shift happening in how data is generated (more and more by individuals) - traditional IT infrastructures are becoming too expensive or too inflexible...

What we announced today:  So, you will hear an overall message of changing Economics through open IT architectures and infrastructures coming from Sun.  And you will hear us announce categories of the market in which we aim to change the economics in- today's happens to be archive.  What we announced:  

Since I have personal experience with the SL3000 library and CIS - I'll paint some color on these products and their history :

Sun StorageTek SL3000 Tape Library:
10x the power savings and 50% footprint advantage vs. Quantum & IBM

First of all, there has been plenty of FUD thrown around about Sun's commitment to Tape after the StorageTek merger.  Nothing shows more vendor commitment than developing and launching a new product in that space - and we've announced two today.  In this Eco-focused world, tape becomes more relevant - not less.  Data sitting on a tape cartridge consumes 0 power and gives off 0 carbon emissions after all - tough to argue those numbers.  So Sun brings a full storage portfolio to the archive space - customers can leverage the performance of disk AND the economics of tape.

How the SL3000 came to be was a Product Manager's dream:  A) We saw a gap in our tape portfolio between entry and enterprise libraries; B) we did extensive customer research and focus groups to get customer requirements; C) we flew customers in to see and comment on the prototype D) we announce it today.

No sloppy welds:  My team was fortunate enough to conduct the research for SL3000.  When we were in Asia focus groups, customers told us something that took us by surprise.  Our customers would actually look at the inside edges of a tape library to see how it was welded together.  If the weld was "sloppy" - put together hastily - they'd notice.  In a culture of quality - the little stuff is an indicator of overall quality.  Suffice to say, we've been poking our heads inside libraries looking for sloppy welds ever since.  A good indication on how customer feedback drove this product to market (and our quality focus at Sun StorageTek).  

Some quick stats on the library itself:

  • SL3000 scales from around 200TB to 3PB (200 - 3,000 slots)
  • Supports open and mainframe environments and mixed media
  • Fully redundant robotics & non-disruptive capacity upgrades
  • Up to 10x power savings and 50% footprint advantage vs. Quantum Scalar i2000 and IBM TS3500 (Economics!)

 Sun Customer Ready Infinite Archive System (aka CIS)
Costs 46% less and consumes 1/3 the power of a 2PB EMC Centera Solution

Skunk Works?  I just learned that the origin of the term "Skunk Works" came from Lockheed Martin when they were developing one of my favorite WWII fighter planes - the P-38 Lightning.  In tech, Skunk Works can have positive and negative connotations - I personally think a lot of innovation has come from working around the process, but you need a healthy balance.   Sun's X4500 (aka Thumper) came straight from engineering and by all measures its turning out to be a huge success.   I'm supporting a Skunk Works project in fact, and I'd love to see it get off the ground one of these days (perhaps more in a later blog, but its open source Systems Managed Storage software brought out of the mainframe world into open systems, available over SourceForge). 

So while SL3000 has its origins in traditional product management, CIS (er... "Customer Ready Infinite Archive System") got its origins more on the Skunk Works side of the house - from the Field Sales and Engineering side specifically.   I don't know the full story, but I am guessing it went something like this....a Sun systems engineer is at a customer site deploying a tiered storage architecture (disk, tape, server, HSM)  for the umpteenth time and thinks, "what if we did this integration BEFORE we shipped this to customers???"  And CIS was born (or something like that...)

Call it a tiered storage platform, or ILM-in-a box, or whatever - but this is what it is (and it can be used for more than just archive btw):

  • Pre-configured and integrated
  • Server, Disk & Tape in a single rack
    • Sun X4200 or T5220 Server
    • Sun StorageTek 2540 SAS or SATA disk
    • Sun StorageTek SL500 Tape Library, supports LTO-4 tape drives
  • Comes with HSM software (SAM) - fully automated backup and data migration
  • Highly scalable - scales to petabytes by stringing multiple racks together under a single file system (QFS)
  • Ready to run - just hook it up to the network - cool stuff...

So, since we are talking archive, we compared this integrated architecture to another popular archive appliance in the market.  In a 2PB configuration, Sun's Customer Ready Infinite Archive System costs 46% less and consumes 1/3 the power of a 2PB EMC Centera solution.  Additionally, data migration cost extra for Centera customers while it comes part of Sun's solution. 

So, the industry is looking at IT economics closer than it ever has before.  Sun is innovating here at the infrastructure level - adding functionality and performance while reducing cost through open source software, integrated systems, Eco-efficient hardware and leveraging the economics of tape...

---- Update ---

Other Sun blogs discussing Open Archive:

Friday Feb 22, 2008

EMC's Big Bet

So EMC just announced an all-cash acquisition of a small startup as well as a new "Cloud Infrastructure and Services" division.  And while they do A LOT of acquisitions - EMC's Chuck Hollis doesn't think this is your everyday acquisition. 

I agree with him...

As we know, Sun and others in the industry believe the future will hold utility-like "Information Data Centers" - where customers can buy/rent/lease CPUs and GBs.  An emerging buzz term for this is Cloud Computing or Storage.  Sun is building next-gen technology (another buzz word, thank you) that can be leveraged by these emerging mega-data centers - for more on this, see Greg Papadopoulos' SAS 2008 Keynote on building network-scale computing.   

So, while Sun is providing  the technology for future "Information Data Centers" - EMC is trying to become one.

Let's look at what EMC has been up to these past couple years in this space (dates are close but not exact): 

  • 2/21/08:  EMC announces new Cloud Infrastructure and Services division to be run by ex-Microsoft exec Paul Maritz
  • 2/21/08: EMC announces all-cash acquisition of Pi Corporation, developer of Personal Information Management Technology
  • 1/22/08: EMC announces EMC Fortress SaaS infrastructureMozy online backup service 
  • 10/1/07:  EMC acquires online storage provider Berkeley Data Systems (Mozy) for $76M
  • 11/1/06:  EMC acquires online storage provider Avamar Technologies for $165M (Avamar is a de-dup technology, but it sits behind online storage service offerings like Arsenal's)
  • 6/30/06:  EMC buys RSA Security for $2.1B - they have security and authentication software, key in providing storage services

So, EMC's big bet looks like it is to become a "Utility Information Data Center" (for lack of a better term). 

More power to them if they can pull it off.  And while EMC may have already made its investments in storage infrastructure, software, server virtualization and security for this - they are going to need some pretty innovative server and networking technology to pull it off.  I think I may know a company ;-)

Wednesday Jan 30, 2008

Wow! NetApp posts EMC SPC benchmark

In a gutsy move, NetApp just posted a Storage Performance Council (SPC) Benchmark....on the EMC CLARiiON! 

Now EMC has made clear statements that they "don't participate in performance benchmarking" - EMCer Chuck Hollis blogs about this in detail.  But this is not entirely true - as EMC is an active member in SPEC NAS performance benchmarks.  So the real issue is that EMC does not participate in SPC disk array performance benchmarks.  They have been pressured to do so, but it is ultimately their choice (until now it seems).  

In a pretty bold move - NetApp looks to have acquired an EMC CLARiiON disk array and posted some benchmarks for them.  NetApp even issued a press release on it.  Now to stay above the fray (I expect a pretty heated battle over this), I won't offer any opinions or judgments.  What I will do is post commentary from EMC and non-EMC bloggers below; as well as the public SPC results...


(For the record, I do consider the SPC Council to be a good and fair 3rd-party benchmark organization that tries to replicate real customer workload behavior accurately.  They are supported by Sun, IBM, HP, NetApp, Hitachi, Fujitsu, LSI Logic and Dell...)

Relevant Blogs:  

EMC CLARiiON CX3 Model 40 pictured at right (no SnapView):

SPC-1 Submission Identifier: A00059
SPC-1 IOPS(tm): 24,997.49
SPC-1 Price-Performance(tm): $20.72/SPC-1 IOPS(tm)
Total ASU Capacity: 8,465.016 GB
Data Protection Level: Mirroring
Total Price: $517,851

EMC CLARiiON CX3 Model 40 (SnapView enabled):

SPC-1 Submission Identifier: A00060
SPC-1 IOPS(tm): 8,997.17
SPC-1 Price-Performance(tm): $59.49/SPC-1 IOPS(tm)
Total ASU Capacity: 7,054.148 GB
Data Protection Level: Mirroring
Total Price: $535,251

One stat that industry insiders are pointing out is that the EMC CLARiiON took a 2.7x hit in performance with snapshots enabled (~25,000 IOPS down to ~9,000 IOPS).  This looks to be a pretty high performance penalty and may be something EMC needs to address. 

What of Sun StorageTek SPC results?   

As stated above, we're big SPC Benchmark supporters.  It is just another good tool that gives customers more intel into choosing the best storage system for their business.  You can find a ton of Sun benchmarks on SPC, two notable ones in this midrange disk array space are:

Read more on the Sun StorageTek 6140 here and on the StorageTek 6540 here...




Tuesday Jan 29, 2008

Web 2.0 Needs Good Backup Too

To backup or not to backup, that is the question; Whether 'tis nobler in the mind to suffer the slings and arrows of outrageous downtimes, Or to protect data against a sea of troubles, and by opposing, end them...

Five years ago I was managing StorageTek's Advanced Technology Research department in RD&E (we were the "R").  One of our research probes was "Grid Storage."  At the time we used Grid Storage to describe an emerging storage architecture.  We also researched "Utility Storage" - or paying for only the storage you use as a type of service.   Even at StorageTek, where tape was king, we were talking about how "Grid Storage" could get rid of backups forever!  Imagine, multiple cheap nodes on the network, data striped across all of them - a whole section of the "Grid" goes down and you have redundancy across other sections. 

No more need for data backup and all the admin/management pains that go with it...

Today we have new innovations on the service and infrastructure side.   When talking about SaaS, I have used Amazon S3 as a prime example.  EMC just entered this space with their announcement of an online backup service available through EMC Fortress - their storage service infrastructure/platform.  On the infrastructure side, companies (and, more importantly, end users) are building grid-like enterprise storage nodes with volume components and clustered/parallel/virtualized file systems.  Sun is one vendor leading the charge here.  And Web 2.0 companies have emerged as the primary consumers and developers of these systems. 

But for some Web 2.0 companies good backup (and backup practices) is an afterthought.   Afterall, start ups can't afford enterprise protection practices like hot replication.  Oftentimes they have to restore data from backups during a crash or outage - but if their processes are not up to snuff, or their backup/restore system is faulty - they succumb to longer outages and lost data.  Bottom line - the utopia of "Grid Storage" is not here yet so having a good backup and recovery plan in place should be a necessity for Web 2.0 companies too...

NOTE:  While I am using some public examples below, I do need to note a couple of important items:

  • First, these companies should be commended for being open w/ their users.  There is something to be said for flat out honesty, and I couldn't link to these examples if these companies weren't honest and open.
  • Second, these companies most likely implemented very high quality storage systems with advanced RAID protection - even the best systems fail.  (The odds are against 3 drives failing simultaneously, but unfortunately, it happens) 
  • The point it that these were GOOD systems that failed - so even advanced systems need some sort of backup...

There have been several outages in some social networking/photo sharing sites of late.  PBase as a great example.   Read about their outage here.  Below is an excerpt from their IT dept to their end users on PBase's discussion forum: 

"On Saturday, we lost 3 disks simultaneously in our main storage system which runs on NetApp hardware. This caused an 8 Terabyte volume to have some inconsistencies which have to be analyed and repaired before we can put the volume back online. ... I wish the recovery process could have gone faster, but after a problem with the filesystem, it's important to analyze it carefully so we can be sure everything is healthy."

Digital Photography Review ( posts an update about its recent outage:

"On January 18th 2008 we had a multiple disk RAID failure...The forums are now back up and running again. There are three caveats: firstly that messages posted before 1st January 2008 are still restoring, secondly that it appears that messages posted in the last five days have been lost and that search is disabled..."

Now, let's say as a Web 2.0 company you choose to go with a storage service (SaaS) rather than an internal system.  Most notably, SmugMug uses Amazon S3 for this.  This is a viable option, and while I am a fan of Amazon S3, users need to read Amazon's Terms & Conditions first.  Here is an excerpt from Amazon S3's T&C (bold sections mine for emphasis): 

"7.1. Downtime and Service Suspensions.  In addition to our rights to terminate or suspend Services to you as described in Section 3 above, you acknowledge that: (i) your access to and use of the Services may be suspended for the duration of any unanticipated or unscheduled downtime or unavailability of any portion or all of the Services for any reason, including as a result of power outages, system failures or other interruptions; and (ii) we shall also be entitled, without any liability to you, to suspend access to any portion or all of the Services at any time, on a Service-wide basis: (a) for scheduled downtime to permit us to conduct maintenance or make modifications to any Service; (b) in the event of a denial of service attack or other attack on the Service or other event that we determine, in our sole discretion, may create a risk to the applicable Service, to you or to any of our other customers if the Service were not suspended; or (c) in the event that we determine that any Service is prohibited by law or we otherwise determine that it is necessary or prudent to do so for legal or regulatory reasons (collectively, "Service Suspensions"). Without limitation to Section 11.5, we shall have no liability whatsoever for any damage, liabilities, losses (including any loss of data or profits) or any other consequences that you may incur as a result of any Service Suspension."

So outsourcing storage may be a great option for a lot of companies - but there is also risk here...

Ay, there's the rub...

To backup is clearly the answer - it was years ago, it still is today.  But the rub is this: the time investment and cost between a poor backup process/system and a good one is probably minimal. 

Let me repeat that:  the time investment and cost between a poor backup process/system and a good one is probably minimal.      

I've been doing stuff in storage and IT for 14 years now, and I know that a basic, thought-out data protection plan will give you one of the best returns on investment in IT.  Odds are that every Web 2.0 company has some type of data protection practice in place - but it may be ill defined or largely neglected in lieu of the million other things going on there.   But a little time investment will go a long way in keeping customers confident that they can rely on you safeguarding their data. 

And, if you need to know how to pull together a good backup or data protection plan, Sun StorageTek Service Plans are a good place to start...

Monday Jan 14, 2008

IBM buys XIV - good move or bad?

Our team wrote an internal analysis for Sun Execs on IBM's XIV buy last week (thanks to Bruce Norikane for his brilliant analysis as usual).  

So, was this a good move for IBM? IMHO, yes.  I don't know if the deal will pan out for IBM (who does), and I don't know how solid the technology is (all I can do it read what is public) - but from one competitor to another, I think it makes strategic sense for IBM (and for the industry). 

Before I get to my thoughts on why, I do have to say this has been a fun analysis to do - primarily because of the Blog battle that broke out between some EMC and IBM bloggers.   There is history here too, which always make things interesting.  For those who are not storage insiders, here is the story (and feel free to use comments to correct anything I get wrong here...)

  • XIV was founded in Israel in 2002 by Moshe Yanai who is known as the father of EMC's Symmetrix disk system, one of the most successful storage products in history.
  • Moshe left EMC 8 years ago, and EMC focused its M&A money on buying software
  • Moshe continued his innovation efforts at the infrastructure level, and has been involved with several startups including Diligent 
  • Now IBM picks up XIV, and Moshe, as part of itson demand” information management strategy
  • Moshe's latest infrastructure innovation, NEXTRA, is radically different from the traditional monolithic infrastructure of EMC's Symmetrix and IBM's DS enterprise storage arrays
  • Although details are scarce, in a nutshell, Moshe's NEXTRA (pictured at right) implements an asymmetric cluster architecture with 2 types of nodes - interface modules and data modules:
    • The Interface modules perform the front-end processing and host interface. They present virtual SCSI LUNS through either FC or iSCSI host ports.
    • The Data modules store the data on 15 1TB SATA disks.
    • Interface and Data modules use dedicated 1Gb Ethernet switches to pass data and control commands. Both modules use dual 2.6 Ghz Intel processors.
    • Customers can add either Interface modules for more front end processing or Data modules to scale storage.

IBM bills its NEXTRA acquisition as a "Web 2.0" storage investment - which it should.  Web 2.0 applications demand open, flexible storage - that are both affordable and can scale massively.  Something expensive and hard to do with proprietary, monolithic architectures - but easier and cheaper to do with volume, general purpose storage "parts" strung together w/ clever software to achieve enterprise levels of capacity and performance.

So, if I may be allowed to speculate (that's what blogs are for right?) - it seems to me that IBM is positioning XIV as a Web 2.0 storage architecture to compliment its traditional DS enterprise array architecture (Sun has already taken this approach - more on this later).  EMC is positioning XIV as an attempt to help/replace IBM's "failed" DS8000 program (IBM Enterprise DS series has had a not so good showing in the enterprise disk array space compared to EMC Symmetrix and Sun's StorageTek 9900 - aka Hitachi TagmaStore, HP XP).  And I bet Moshe would love nothing better than to disrupt the market for IBM's DS series AND EMC's Symmetrix!  

With that background, here are the XIV Blog Wars that broke out last week: 

What's really happening here? (and why I think this is a good move for IBM )

What is really going on is this:  A new storage application has emerged in the data center - and it's pretty exciting.   As with any emerging application or technology, every vendor has its own terminology until the industry settles on one it likes.  Obviously I will be using some of Sun's terminology here...

What's the new data center application?  In a nutshell, Web 2.0 applications.   These are applications that store user content including media on web.  Classic application examples include Google, eBay,  Emerging examples include SmugMug, FaceBook,, and even traditional wireless companies like Verizon who send thousands of games, images and ring tones over the wireless network.

IMPORTANT POINT:  One of the most critical things I can say about this trend, it that a traditional storage application and a "Web 2.0" application can exist at the same company.  If history is our guide, there isn't one application that will overtake the other (or one architecture that will completely overtake another) - a data center will have a mix of these technologies. (The mix % is what will change over time). 

With that said, customer needs differ whether you are supporting a Web 2.0-type application or traditional storage application.  See the table below.   

 Traditional Storage Application Needs
 Web 2.0 Storage Application Needs
  • Scalability
  • Consolidation
  • TCO
  • Availability
  • Data IntegrityData Protection (BC/DR)
  • Innovation
  • Massive scalability
  • Open/flexible platforms
  • Low cost
  • Efficiency
 Customer Types:  Business & IT Management
  Customer Types:  Developers & IT Management


What's Sun Storage doing about it? 

So this is where we get to what Sun is doing about this market shift and why I think IBM's acquisition was a good idea...

First of all, Sun has invested in the traditional enterprise disk array market with the Sun StorageTek 9900 disk array - with Storage Virtualization, Thin Provisioning and the fastest performance on the planet, it's giving the market leader in this space (EMC) a run for its money

Second, Sun's development efforts are geared towards investing in open storage innovation in order to change the economics of storage, especially for Web 2.0 applications.  In this sense, Sun has developed an Open Storage Platform (See trend #1 in Top 10 Trends). 

Even more, while some companies are just announcing the acquisition of Web 2.0 infrastructures and other are leaking their development efforts for Web 2.0 infrastructures - Sun is already selling its Storage Server (Sun Fire X4500, aka Thumper) based on its Web 2.0 infrastructure offerings.

Although pricing information is scarce and unreliable around IBM's newest NEXTRA system, preliminary pricing appears to put it in the $5/GB range (about the same as traditional midrange SATA RAID)...and more than 5x Sun products like the Sun Fire X4500 already in the market.

So, the market is demanding traditional AND non-traditional storage infrastructures today for supporting application needs.  Sun can be criticized for a lot of things, but a credit to the company has always been its ability to peg future market trends and innovate.  What's new here, is that Sun is executing in the traditional storage space (with disk and tape - thanks to the StorageTek acquisition) AND the emerging Web 2.0

---- Update ---

Read about our latest Web 2.0 investment...

Tuesday Oct 02, 2007

Sun breaks World Record - but where's EMC???

Well, the Sun StorageTek 9990V disk system just posted the fastest Storage Performance Council (SPC-1) benchmark in enterprise datacenter history at 200,245.73 IOPS... (SPC-1 simulates the random I/O workloads required to support typical database, OLTP or email server applications)

I do have to say, however, that I brought this up to a colleague of mine who quickly refuted me saying, "but EMC wasn't in the benchmark, so are we really the world's fastest?"

My answer was yes - if someone doesn't show up to a title fight, then they don't get the title.  Why didn't EMC show up to the contest?  You can Google the answer to hear claims like "benchmarks don't translate to real world performance" or that "there is no good independent performance metric for storage."

But this is not a position held by the majority of disk storage players mind you - one need only look at the vendors participating on the SPC website, maybe you have heard of a few?  

But the real irony is that EMC is an active participant in SPEC for their NAS products.  So, my question is - why doesn't EMC publish how they test the performance of their systems like Symmetrix?

You see, even if EMC is not a part of the SPC (yet!) - we would like to see them publish how they test their systems performance.  You see, the value in SPC is not only in the benchmark and its results - but the fact that customers can see exactly HOW these systems were tested.  Putting the power of knowledge where it should be - in the hands of the customer.   

So the above vendors and SPC deserve credit for supporting a great philosophy - "free and open exchange of ideas and information to ensure fair and vigorous competition between vendors as a means of improving the products and services available to the general public."  (See About SPC)

In free and open idea exchange, customers win.  They need good, fair competition - and if you are a customer, would you rather make your purchase on information from a vendor spec sheet or a vendor-neutral independent auditor? 

So, congratulations to the Sun StorageTek 9990V (also sold as Hitachi Universal Storage Platform and HP StorageWorks XP24000) for being the fastest monolithic enterprise disk array on the planet!

But also keep in mind that it's not only the fastest - it also offers storage virtualization and thin provisioning so customer's get more utilization out of their products while protecting their infrastructure investments...a pretty good deal if you ask me. 



SPC Disclosure statement  
Systems Compared: Sun StorageTek 9990V, IBM DS8300 Turbo, Fujitsu ETERNUS 1100
SPC-1 Submission Identifiers: A00055, A00049,  A00053
SPC-1 IOPS(tm): 200,245.75, 123,033.40, 115,090.06
SPC-1 Price-Performance(tm): $17.31, $18.99, $16.12
Total ASU Capacities GB: 26,000.00, 9,103.36, 10,854.40
Data Protection Level: Mirroring
Total Prices: $3,466,309, $2,336,626, $1,855,100


My storage team and I focus on three of the most important aspects in any industry: customers, competitors and market trends. There is insight to gain and share in this role, so here is our take on Sun and Storage.


« July 2016