Storagezilla

Isilon One Million.

-140 Isilon S200 Nodes.

-One Volume.

-One File System.

-One Million+ IOPS for the SPECsfs 2008 CIFS and SPECsfs 2008 NFS Benchmarks.

But first a proof point. Unlike many of it’s competitors Isilon aggregates system resources in every node to provide linear scalability. To show this the team tested 7, 14, 28, 56 and 140 nodes by powering them on as required.

Over the course of the testing for NFS each node averaged about 8,100 IOPS/Node and 11,500 IOPS/Node for CIFS. Note, the NFS and CIFS benchmarks aren’t the same (While the platform configuration was.) so the numbers aren’t comparable to one another.

They are however comparable to everyone else who has run the same benchmark. But not everyone who’s submitted for one has submitted for the other and those who do can submit results from two different configurations.

In Isilon’s case it’s the same configuration with the same number of nodes for each benchmark. Lets net out the published CIFS numbers to see how they stack up.

 

126% higher than it’s nearest competitor and significantly higher than all Isilon’s published competitors.

Plotting the NFS results we have Isilon 75% higher than their nearest competitor and significantly higher than all their published competitors.

But when we look at the numbers on a per file system basis we end with with something really interesting. Isilon is using a single file system in both benchmarks while many of it’s competitors are using multiple file systems due to ghastly volume size limitations or architectures which cannot aggregate system resources and as a result require the use of multiple file systems to prevent a single controller from becoming a massive bottleneck.

With some quick mathematics run on the published numbers Isilon is shown to be more than 1700% faster on a per file system basis than it’s closest competitor and close to 5000% faster than a volume size challenged competitor on the CIFS benchmark.

On the NFS benchmark it’s from close 200% faster on a per file system basis to it’s nearest competitor to over 1000% faster than a volume size challenged competitor.

So, through this repeatable benchmark what have we learned?

1: Isilon requires only One file system regardless of the workload type.

PROVEN.

2: Isilon OneFS aggregates systems resources and provides linear scalability.

PROVEN.

3: This linear scalability allows Isilon to deliver IOPS to CIFS & NFS workloads at a massive rate.

PROVEN.

Isilon, with a file system spanning all nodes in a cluster, the ability to move tasks between all system resources in the cluster and the ability to control redundancy down to the level of an individual file, is far and away the leader in Scale Out NAS.

And with the growth of File storage projected to far outstrip the growth of Block storage over the next ten years this is The Decade Of Isilon.

Oops

Ouch!

Distribute.IT has revealed that production data and backups for four of its shared servers were erased in a debilitating hack on its systems over a week ago.

…

“In assessing the situation, our greatest fears have been confirmed that not only was the production data erased during the attack, but also key backups, snapshots and other information that would allow us to reconstruct these Servers from the remaining data,” the company reported.

You may think that I’m saying the hack is wrong – and anyone conducting such a malicious attack is certainly being particularly unpleasant. But the simple truth is that such an attack should not be capable of rendering a company unable to recover its data.

It suggests multiple design failures on behalf of Distribute.IT:

• Backups were not physically isolated; regardless of whether you can erase the current backup, or all the backups on nearline storage, there should be backup copies that are sent off-site and removed from such attack;
• Alternatively, if there were offsite backups – if they were physically isolated, they were not sufficiently secured;
• Retention policies seem inappropriately small; why could they not recover from say, a week ago, or two weeks ago? The loss of some data even under a sustained hack should be somewhat reversible if longer-term backups can be recovered from. Instead, we’re told: “we have been advised by the recovery teams that the chances for recovery beyond the data and files so far retrieved are slim”.

It’s also worth noting that this goes to demonstrate a worst case scenario about snapshots – they’re typically reliant on some preservation of original data (either running disks, or ensuring that the amount of data deleted/corrupted doesn’t exceed snapshot capacity).

via nsrd.info

1st: Ouch!

2nd: Yet again Snapshots are not Backups. You do need a Dedicated Protection Storage Infrastructure for Backup.

3rd and finally: Ouch!

Isilon Movie Premiere and Product Launch Solution Series

YOU ARE INVITED! JOIN US TO SEE HOW ISILON MAXIMIZES PERFORMANCE FOR BIG DATA.

The Big Data era has arrived and it is impacting our daily lives in a variety of ways, from the pursuit of personalized medicine to social networking to movies and entertainment.

Come see BIG DATA on the BIG SCREEN and join us for an exclusive movie premiere and launch event! Find out why Isilon is leading the way in Big Data storage by simplifying management, saving time, and maximizing performance.

Isilon has just launched new generations of its scale-out NAS hardware and software platforms that will enable you extract greater business value from your information, more quickly and collaboratively than ever before.

via www.isilon.com

Listen up Poozers, VFX for Green Lantern?

It's on Isilon.

Wired's pick, the Iomega Home Media Network Hard Drive, Cloud Edition

This sturdy and svelte monolithic slab looks good without being pretentious. Of the products we reviewed, it’s the only one with a metal case (versus a plastic one). Lay it flat or use the included stand to position it vertically. Software setup is quick and painless, and then the fun begins. The Iomega, it turns out, is loaded with great features yet never seems bloated.

The basics work well—the best of any drive we tested when it came to sharing files with DLNA devices and other computers running iTunes—and Iomega's management app offers a range of intuitive, pre-set shared drive letters to play with.

But deeper digging reveals many more tricks. You can share your "personal cloud" with up to 250 users over the web, though each has to install the same management app to see the drive.

via www.wired.com

As I keep telling my friends at Iomega, when they own the software we win. When we suffer at the hands of a software OEM we don't.

Good job Iomega and timely since I've a small capacity drive or two coming up for replacement.

Nonsense, Direct From Dell.

Michael Dell, the man who famously told the public were it his decision he’d shut down Apple and sell it off as Steve Jobs was turning into printing press for Apple shareholders, led his company from a top spot in the x86 Blade Server market to number 4 behind newcomer Cisco as of last month and 25% of who’s shareholders withheld their votes to show their opposition to his most recent election, is at it again.

This time however he’s gazed into his crystal ball to look at the future of Backup & Recovery and to him the future of Backup is limited to only what Dell storage systems can do today. Tier & replicate.

You always know who doesn’t own any real Backup IP as they say you “don’t really need any of that Backup stuff” just copy it to the same priced tin you’re storing it on today in a different room somewhere else.

Unlike Array based storage software products Backup Applications are Information Management products. They manage the location and life cycle of individual copies of your information throughout what should be a Protection Storage infrastructure.

When it comes to Backup there are those who can and those who cannot. It’s easy to tell who from who if you listen to what they’re saying and how they say it.

As for the rest of Dell’s portfolio while they’re busy chasing everyone who doesn’t sell monitors, printers and USB sticks I won’t be chasing them for a replacement for my aging Dell corporate Laptop.

IT sent me a note earlier this week saying I was due for a replacement, but long ago I used Iomega v.Clone to image it onto an eGo Drive and have been using that on the road with my other devices.

I’ll swing back and look again when the Laptop selection doesn’t scream “Dude, you’re going to Hell”.

Avamar Storms The Backup Software Market

“Among midrange products, Avamar netted an unprecedented 100% of users who said they would buy the product again today -- not surprising given the consistently high marks the product received across all categories.”

Storage Magazine end user survey.

Data Domain integration with Avamar allowing for increased backend storage scale is going to leapfrog Avamar over many of it’s competitors this year.

Backup is transforming and it’s transforming from running some software on some servers with some storage to deploying an Application Aware Protection Storage Infrastructure.

That’s the primary design direction of EMC Backup Recovery Systems, we design, build and deploy Application Aware Protection Storage owned and managed by the Backup Administrator. It’s not an after thought of the Server and/or Storage Administrators and not subject to their day to day operational whims.

100% is a number you can’t argue with. When you help Backup Administrators get their proper seat at the table they’ll want to keep you around.

Dedupe with Snapshots, things you can read.

Allegedly I have NetApp’s dedupe architecture all wrong when it comes to running dedupe with Snapshots.

But hey, I’m only reading it from NetApp’s own materials.

Lets read along together!

Snapshot copies lock blocks on disk that cannot be freed until the Snapshot copy expires or is deleted. On any volume, once a Snapshot copy of data is made, any subsequent changes to that data temporarily require additional disk space, until the Snapshot copy is deleted or expires. The same is true with deduplication-enabled volumes.

Volumes using deduplication do not see the savings initially if the blocks are locked by Snapshot copies.

This is because the blocks are not freed until the lock is removed.

Some best practices to achieve the best space savings from deduplication-enabled volumes that contain Snapshot copies include:

• Run deduplication before creating new Snapshot copies.
• Limit the number of Snapshot copies you maintain.
• If possible, reduce the retention duration of Snapshot copies.
• Schedule deduplication only after significant new data has been written to the volume.
• Configure appropriate reserve space for the Snapshot copies.

Publication date for this? Q1 of this year.

And then weeks later the NetApp PR wagon rides in calling Snapshots “backups”. So if you’re looking for a solution where deduplication requires you limit the number of Snapshot “backups” you keep and if possible reduce the amount of time you keep them then SnapProtect is the product for you.

Of course then we have the whole snapshot replication thing going on. Since the Snaps are intertwined with the primary data if you delete the snaps on the source they go away on the replica. But think of it this way, your storage administrator has just reclaimed 2x the storage by deleting all that worthless stuff the backup administrator said she/he just had to hold on to.

But the Backup Admin doesn’t want that, so the answer you’ll get here is SnapVault! Except SnapVault requires all the data be copied to the destination at native size and then be deduplicated all over again.

There are better ways.

They involve heterogeneous backup applications and real protection storage with deduplication, compression and bandwidth optimised replication at all times.

Snapshots aren’t backups. They never were. And the more they try to retrofit modern functionality onto them the larger the mess it becomes for the Backup Admin.

Snapshots are not Backups

Since I have invented a button I can press which makes the attention seeking Lotus Eaters come over all frothy and then simper for my attention in public, I’ve decide to press it once again.

Pointer based Snapshots are not backups.

Why not?

Because they depend on the data they’re a snapshot of existing.

That’s not a backup, a backup is “A collection of data stored on (usually removable) non-volatile storage media for purposes of recovery in case the original copy of data is lost or becomes inaccessible; also called a backup copy.”

A pointer based snapshot fails in being a backup as if the volume the snapshot is inaccessible then so is the snapshot.

Lets take the case of a large social networking site (And NetApp user) who had a shelf or four get blown away during what they described as routine upgrade, the result being that 15% of their users photos were vaporized.

Snapshots of those volumes were worthless as they were on the same storage which was inaccessible. What did they restore from? Tape. Soon after, The Social Network went off and introduced an in house developed storage system for all such photos.

So much for “Integrated Data Protection” it’s so integrated that failures cascade from primary to backup as their concept of backup is intertwined with the primary data.

But what else is wrong with building your backup infrastructure the SnapProtect way?

NetApp systems can only take 255 snapshots of a volume. Sounds large, not when you’re dealing with backup and recovery it isn’t. It’s less than 37 weeks if you were taking one snapshot a day. The phrase “long term retention” doesn’t come into play.

What’s the answer?

Get the data off the array and onto external storage. This is what SnapMirror to Tape was designed for, used it myself to unclog backup data out of primary storage to Data Domain Protection Storage, and that’s what they’re trying to do by reselling CommVault.

NetApp are coming up with ways to get backup data off of primary storage.

Of course there’s a couple of other issues with what they’re trying to push.

Weren’t the SnapManager products supposed to “replace” backup? That failed idea now appears to have been shot in the head.

With SnapProtect you put a whole bunch of VMs on a volume and want to backup two of them as part of a group, you end up quiesicing all of them. That’s really fun when different VMs are in different backup groups with different backup policies.

Snapshot was designed 20 years ago when each server had it’s own volumes. In the grand server virtualised present Virtual Machines share storage, treat it as a commodity and bounce around between volumes. Using SnapProtect what you inflict on one of them you inflict on all of them.

Who’s dedupe should customers be using? The PR people appeared to do everything they could to tap dance away from answering that question.

Since it’s a matter of record NetApp dedupe runs as fast as a glacier working it’s way across a flat landscape if you’re using regular snapshots for your backup you’re now clogging up your primary storage with duplicate blocks as NetApp dedupe can’t eliminate duplicates if they’re part of Snapshots.

Now that you’re filling your storage array with backup data, causing ever decreasing deduplication ratios in their wake, are storage admins expected to buy new disk shelves or just start deleting backup data when their applications start running out of storage?

Deduplicate before you Snap? So much for high speed zero impact snapshots if you’re waiting for post process fixed block deduplication to churn Array CPU and finish.

When asked about heterogeneous backup NetApp tell you to buy a V-Series. So you too can watch snapshots calcify your primary storage while their Dedupe burns CPU cycles and does nothing.

The bottom line is while SnapProtect does everything to get the data off the primary storage, something NetApp said you didn’t need, it’s not a heterogeneous backup solution and shouldn’t be thought of as.

It’s just another silo they want a Backup Administrator to spend time baby sitting independent of a heterogeneous backup and recovery design.

Their future. Your past.

Today NetApp bought the SnapProtect trademark from CommVault (Screwing the premiere showcase partner on the CommVault website, Dell Equallogic, in the process) and agreed to resell Simpana.

From a technology standpoint the only thing they’re adding to that Banana is a bruise since CommVault already did everything they’re banging the drum about yesterday and the value of a backup application is in it’s application awareness to which NetApp contributes nothing.

Bottom line, CommVault has grown beyond NetApp’s reach so having missed the opportunity once again to get into the backup market their only option is to resell.

There’s a long history here of licensing technology from BakBone, SyncSort and now CommVault, painting a blue stripe across it and palming it off as their own.

At this stage they’re running out of backup application vendors to sign a distribution agreement with.

One might ask where’s Symantec in all of this? I’d say still trying to sell themselves to Huawei. Having been burned by NetApp on the original OpenStorage development project I think they’re not returning too many NetApp phone calls these days anyway.

Point In Time Copies on the array. This is supposed to be the new new thing, right?

Indeed you can already take application consistent array based point in time copies on EMC Arrays, restore from them and roll them over to external media.

It’s so much of a new new thing that functionality shipped in the NetWorker PowerSnap module.

In 2003.

The first release even supporting backup of remote devices in an SRDF group on a different Symmetrix.

Today PowerSnap supports application & catalogue consistent point in time backup using TimeFinder, SnapView, SnapSure and RecoverPoint CDP.

You can even backup NetApp systems with NetApp Snapshots with the EMC NetWorker PowerSnap for NAS Module. 

The question becomes though with the ubiquity of software snapshot solutions in server virtualisation products, such as VSS and what VADP can give you, why bother locking yourself into an array based technology at all?

While there will always be corner cases with physical workloads, virtual workloads are moveable workloads and can end up anywhere, on any storage at any time.