Storagezilla

vSpecialists: Not so Mad Men

You’ll have to remember that I came up in an era when EMC was pretty much something out of Mad Men so I’m sitting here in a suit with a (non-alcoholic) beverage right now watching the vSpecialist loons dance around the place while thinking most of them look like Vanilla Ice wannabes.

But it’s fun. And it’s a good video produced well. Good job.

 

The New EMC is Pretty Impressive…

Impressed with the new EMC

So back to the meeting we had last week. We had discussions around BURA, Unified Storage, Encryption, RSA, and the list goes on. Just the amount of resources that EMC was able to bring to the table, on short notice, was impressive. Each individual was knowledgeable about their individual product, but also was able to spout the overall EMC messaging well. They were able to show how their individual product meshed with the other EMC products that were relevant to the customer.

Most impressive about the day were the products in the pipeline slated for the near future. In the past EMC was a company with a lot of products that didn’t really fit well together. Now they are a company that has brought things together and it feels like they have a true vision on where they want to go. Not just a vision but a plan on how to get to where they want to go. The end game is something that I always felt they lacked in the past. You can actually see how the differing products are going to come together to create an entire solution. Beautiful!

via blogs.rupturedmonkey.com

There are two things I find satisfying in my job.

1. A customer tells us we're doing a good job.

2. A partner tells us we're doing a good job.

Reading the above from Snig has made my weekend as I head out the door. It shows the direction EMC has moved in is the right one. I'd like us to have reached the destination by now, things never happen as quickly as I'd like them to, but they're happening at a speed that those who are really looking can see the destination.

Do you want to be hardcore?

The current SPEC benchmark submission for the Celerra VG8 running DART 6.0 has it performing 22.5K IOPS per processor core on the new Westmere system board.

That’s 135K IOPS on one DataMover.

You’ll recall the benchmark prior to that for the NS-G8 running DART 5.6 was 110K IOPS on two DataMovers

Glancing at the rest of the SPEC submissions I don’t think anyone else is that hardcore in IOPS per core. Have a look yourself.

Two benchmarks in one year. New EMC record I think. And I look forward to the pNFS functionality in DART. It’ll go along way to helping move pNFS into the mainstream.

I’ve a couple of notes on a few things from yesterday I wanted to get in print.

Here’s my quick Cheat Sheet for what just happened.

Then	Now
RAID Groups, MetaLUNs, HyperGroups	Storage Pools
Navisphere, Celerra Manager, RecoverPoint/SE Management Application	Unisphere
LUN Migrations	FAST
Buy faster disk drives	FAST Cache
Add more capacity or delete data	Compression

 

The Now column is how it is from here on in. None of the things in the Then column are going away or will stop working but as the slogan states “Next Starts Now”.

And if you want to be ready for what comes next, start now.

All lines are in listen only mode..oh wait they're not.

If you've ever been on an an internal conference call you'll have heard all of this and much much more.

 

 

Celerra File System Performance Parameters

Just a quick housekeeping note for some of our field folks or other interested parties. Yes me putting it here can ensure more people read it than they would if I blasted it through email, here I’m going to list the Celerra File System Performance Parameters.

Celerra File System Performance Parameters

Name	Description
syncInterval	Timeout between UFS log flushes
asyncThresholdPercentage	Total cached dirty blocks for NFSv3 async writes
cgHighWaterMark	Defines the system's high water mark for the number of CG entries in CG cache
inoBlkHashSize	Inode block hash size
autoreboot	Delay until autoreboot
updateAccTime	Disable access time updates
nFlushDir	Number of UxFS directory and indirect blocks flush threads
prefetch	Disable DART read pre-fetch
inoHighWaterMark	Number of dirty inode buffers per filesystem
thrdToStream	Number of NFS flush threads per stream
inoHashTableSize	Inode hash table size
maxStrToBeProc	Number of network streams to process at once
nFlushIno	Number of UxFS inode blocks flush threads
nFlushCyl	Number of UxFS cylinder group blocks flush threads
nthreads	Number of threads dedicated to serve nfs requests
withoutCollector	Enable NFS-to-CPU thread affinity
dnlcNents	Directory name lookup cache size
ofCachesize	NFS openfile cache size
cachedNodes	filesystem vnodes cache size
backlog	Connection limit for a given port
start nfsd	Number of NFS daemons

I’m not sure it gets geekier than than posting Celerra File System Performance Parameters but here they are. Back to the usual nonsense later.

It’s not about Android Phones, it’s about Android Everything.

Oracles case against Google is nothing less than a fight for relevance in the hottest sectors of the embedded market.

It’s about Android Phones today but it’ll be about Android Pads and Android TVs and Android eReaders tomorrow. All of those will be able to run applications written in Java but as things are today Oracle stands to make nothing from any of those devices.

When I went looking a while back to see if there was a Spring Framework Micro Edition for the development of applications on Android what I found instead was Dalvik, Androids Virtual Machine which really isn’t a Java Virtual Machine at all.

What Dalvik appears to do is take your Java program files modify them to bring the resource requirements down and then convert them into a Dalvik specific set of instructions. So lines of Java code go in but it isn’t Java Bytecode that gets executed, it’s Dalvik specific Bytecode on a Dalvik VM.

Your Java application becomes a Dalvik application at run time.

For Google this is great. They don’t have to pay any license fees but they get to tap all the Java developers out there. Oracle aren’t pleased that this could mean they’ve been handed their hat in the embedded space so they’ve gone after Google citing patent infringement. This is Oracle looking to be dealt back into the Android game. And the Android game is one of the fastest growing embedded games around.

The Oracle lawsuit looks like a strong move when looked at as an act of Java stewardship. But if you consider that Android is starting to show up in a whole host of devices, none of them kicking up any money to Oracle, you quickly see that it’s a last ditch defensive move designed to stop ongoing embedded Java revenue erosion. Every time a manufacturer who has been shipping Java chooses Android Oracle loses a revenue source.

I don’t read patents, you probably shouldn’t either but if Oracle lose this case due to a weakness in those patents they’ll watch as their embedded Java licensing revenue (A Billion Dollar business) vaporizes in front of them.

People have asked if Larry the Java Tax Collector might file suits against other parts of the Java ecosystem. Possibly, but I don’t see a similar threat from anyone in the desktop or server space.

And we’ve been here before. Back in 1998 while the Desktop and Enterprise vendors, except Microsoft, were moving in Java lockstep with Sun, HP had a Google moment and decided Suns embedded license wasn’t to their taste and the Sun embedded JVM was too bloated for their needs. Their answer was ChaiVM, a HP developed (This was back when HP used to have an R&D budget and not a piggybank for Mark Hurd) Virtual Machine to execute Java applications.

Chai was aimed squarely at embedded systems and had two aspects to it. One was to modify the Java code to bring the resource requirements down and the other was to convert the Java code into what they called a concentrated form using a patented technology known as FreezeDry.

So your Java Application became a Chai application at run time.

It seems that there’s never any new disputes in this business. Just modifications to old ones.

From textiles to Hyperscale Green

It's Mastex.

Gov. Deval L. Patrick, speaking at Heritage State Park this morning, named the former Mastex Industries Inc. facility as the chosen site for a proposed $100 million high-performance computing center that will support research at several top-tier Massachusetts colleges, including the University of Massachusetts and the Massachusetts Institute of Technology.

The Mastex site, which runs along the first-level canal on Bigelow Street, was chosen over Open Square, a massive renovated mill complex located between Lyman and Dwight streets. The two sites were named in March as the finalists to host the project. The field had been whittled from an initial list that included 85 potential sites.

via www.masslive.com

I blogged about Hyperscale Green previously and now we see some forward progress with ground being broken on the facility later this year.

I'm watching this one with interest, done right it'll be a powerhouse of Computer Science.

Fracture

Friday is a great day to dump the garbage which is why HP announced Mark Hurd was stepping down under a cloud of scandal while guiding revenue upwards after the market closed.

What with the press looking to enjoy their weekend too the quick fix of some commentary on HP’s spin control offensive will mean Hurd will drop off the radar quickly. I’m sure the Forbes and Fortunes of the world will go looking for the red meat but that’ll show up months from now when Hurd is a distant memory and will have a few quotes of wistful regret.

If it were I positioning this I’m positioning that next week is the start of “The Search”. Hurd’s relationship with a contractor was last week’s story why would you focus on that when instead you can watch CEO Idol play out in front of you?

This industry and it’s watchers love speculation, indulging in some of the most brain damaged fantasies it’s watchers can imagine so guessing who’s going to land in the big chair at HP is the bigger story in the long run.

What did I think of Hurd’s tenure? I think he reaped the rewards of HP’s acquisition of Compaq and knew when the wind was at his back. HPQ was a blockbuster deal that looked like a wreck while people were living it but is what awakened the slumbering giant that was HP.

It was the right deal.

It was a Rubicon crossed which allowed Hurd to engage in other mega-mergers without fear of HP choking to death. Going after EDS must have appeared pain free to the HP Board after what they had been through already.

EDS also looks like a mess right now but I think the next CEO will reap it’s rewards.

Looking backwards at it now HP-Compaq worked in a way that a mega-merger in the same decade such as Symantec-Veritas didn’t. HPQ was chaotic but stronger when a steadying hand grasped the wheel, Symantec-Veritas on the other hand looked orderly but was critically wounded and continues to diminish even to this day.

Hurd also was allowed skate away from the Reporter bugging scandal even after it was shown he approved the contents of the information leak to be transmitted with the malware HP planned to infect Reporters machines with.

At the time HP Chairwoman Patricia Dunn was offered up by the Board as the sacrifice while Hurd apologised to the crowd and left before her execution. The twist being most of the malware never did anything since the Mac using reporter masses was immune to it’s Windows only charms.

On balance as an employee of the stockholders he returned tremendous value to them and earned his salary. But as a man he was as flawed as anyone.

I can’t condemn him for having an affair, we don’t know what goes on inside the Hurds marriage nor should we but he filed false expense reports to cover an affair he was having, cost shareholders Billions when the stock dived on news of his termination and I just couldn’t stomach how clearly he was shielded by more powerful forces during the bugging investigation.

The fact he was the steadying hand on the wheel of HP’s great ship shouldn’t have negated the visibility of the ethical flaw at the centre of his thought process. He conspired to commit fraud twice in five years.

What type of man, paid $24,201,448.00 in 2009, decides it’s a good idea to bill the company for the hotel rooms of the woman he’s having sex with? He could have paid for those rooms out of his own pocket and would still have a job if he did.

It is this flaw in his character which fractured and destroyed him.

Enterprise Business Restart

The Problem

This customer has over 1,000 separate systems that are critical to ongoing operations and are located in a core data center.  The systems are a combination of Mainframe, UNIX variations (AIX, HP-UX, LINUX, etc.), and Windows.  They use SAN and NAS storage (the Mainframe CKD over FICON I include in the SAN), with over half a Petabyte of unique data and various additional copies.  They have SAP and other systems that have data flowing between different them, often with different operating systems, with no easy way to perform data isolation (picture an interaction of web services, shipping, billing, and warehouse data and you get the general idea).  The customer wanted to implement business restart, with clear business requirements.

via applyit.typepad.com

I like Vince's blog. He spends as much time on the problem as the answer.