c0t0d0s0.org - Sun

Adam Leventhal about the storage business in the light of the HSP

nospam@example.com (Joerg Moellenkamp) — Tue, 02 Dec 2008 20:52:16 +0100

Adam Leventhal wrote an really insightful piece about the potential effect of the Hybrid Storage Pool to the storage business in "Casting the shadow of the Hybrid Storage Pool":

What we have today though integrates flash in a way that changes the landscape of storage economics and delivers cost efficiencies that haven't been seen before. If the drives manufacturers don't already, it can't be long until they hear the death knell for 15K RPM drives loud and clear

When the performance of a storage architecture is ensured by SSD there isn´t really a need for low capacity 15k discs ... 7200 rpm (or even lower) drives with high capacity are more than enough.

IBM benchmarketing - now with 100% more nonsense

nospam@example.com (Joerg Moellenkamp) — Tue, 02 Dec 2008 19:12:25 +0100

There must be a standard course at the IBM marketing school. The tricks of the benchmarkting are the same again and again. At least it doesn´t get boring to read at the IBM website. Now i found another benchmarketing stunt at their website. It´s called "Move Up to IBM Power Systems".

It was really fun to read it. IBM Benchmarketing - now with 100% more nonsense. One important disclaimer: The following article just contains the stuff that was obvious at my first read. I have a daytime job that isn´t dissecting IBM marketing stunts

Continue reading "IBM benchmarketing - now with 100% more nonsense"

This made my day

nospam@example.com (Joerg Moellenkamp) — Tue, 02 Dec 2008 12:56:40 +0100

From a discussion about NAT in IPv6:

the only good NAT is a dead NAT

NAS at home

nospam@example.com (Joerg Moellenkamp) — Mon, 01 Dec 2008 13:40:39 +0100

While preparing the next part for the configuration walkthrough of the Sun 7000 my mind was playing with an desire: I want this interface on an soho device. I played around with many NAS boxes but there wasn´t a single interface in the class of the Fishworks interface. Intel Atoms are cheap enough to build real soho NAS devices from it ...

About all this "What Sun should do?" articles

nospam@example.com (Joerg Moellenkamp) — Sun, 30 Nov 2008 17:05:10 +0100

I´ve thought about linking to all this people writing articles about "What Sun should do now?". But at the end most of this articles are written with no background knowledge or with an hidden agenda. So i simply decided not to link to this articles at all.

But there is an article i want to comment. Rich Sharples writes in What Sun Should Do:

#3 Solaris
See above. Linux has won. Whatever technical merit Solaris has today will be commoditized next year. The innovation around Linux is relentless.

Next year ... really? ZFS is available for years now. I don´t see it´s commoditization by any technology in Linux at the moment. Same thing for DTrace. No SSD support besides of simply using it as a faster disk.

And i know, all those Linux people will cry "Foul" but i do not really see innovation in the Linux kernel. Having many hardware drivers in the kernel isn´t an innovation. Commoditization and innovation are diametral concepts. Linux is good in commoditization, it´s the master in the field of good enough. But it fails in the field of innovating. And to get interesting technologies (besides of drivers) Linux need strong competitors. We don´t stand still. There are other interesting technologies in their creation that will gives us a good technological lead.

By the way. To understand Mr. Sharples standpoint you should know, that he is "Director, Product Management at Red Hat". What a strange luck that he suggests to drop products that could hurt his employer. I tend to say this is a kind of hidden agenda.

From my perspective, Sun has just to learn one thing: Learning to kick the butts of our competitors with their good enough products and learning to make more money from our innovations. But: You don´t make money by dropping them.

Virtualbox 2.0.6 available

nospam@example.com (Joerg Moellenkamp) — Tue, 25 Nov 2008 14:44:59 +0100

A new version of Virtualbox is available at the usual location. Version 2.0.6 fixes some bugs as described in the changelog. Thanks to Constantin for the hint.

SPARC Enterprise T3120 - you´ve read it at c0t0d0s0.org first ;)

nospam@example.com (Joerg Moellenkamp) — Mon, 24 Nov 2008 18:39:48 +0100

TPM adds many speculations to the bits of truth in an article at the Register (fueled by this Flag Day announcement). No confirmations or declines about the stuff in the article.

Wikipedia uses Sun equipment

nospam@example.com (Joerg Moellenkamp) — Thu, 20 Nov 2008 15:42:06 +0100

We announced in the press release "Wikimedia Selects Sun Microsystems to Enhance Multimedia Experience for Expanding User-Base Worldwide and Accommodate Explosive Growth of Rich Content"(who uses such fscking long headlines ?!? ), that the Wikipedia will use Sun hardware and software for it´s service:

The Wikimedia Foundation, the non-profit organization that operates Wikipedia and other free-content projects, announced today that it has selected Sun Microsystems, Inc. (NASDAQ:JAVA) [...] to help it build a secure and scalable open source web infrastructure using Sun Fire(TM) x4500 and x4150 servers and Sun StorageTek(TM) storage arrays.
[...]
By implementing Sun's open source MySQL(TM) database and Sun Open Storage infrastructure, we are in a position to create a better and faster user experience.

Hardware teaser: SPARC Enterprise T3120

nospam@example.com (Joerg Moellenkamp) — Thu, 20 Nov 2008 14:31:32 +0100

Just a teaser. There is an interesting putback in Opensolaris Build 104:

PSARC 2008/665 Tenaya Platform Software Support
6766865 Solaris support needed for SPARC-Enterprise-T3120

adds support for Tenaya platform to Nevada

No, i will not tell you more And this putback doesn´t disclose so much stuff like the M3000 support putback

My thoughts about the future of Sun ...

nospam@example.com (Joerg Moellenkamp) — Thu, 20 Nov 2008 00:08:15 +0100

I´ve stopped to think about all the articles from analysts or newspaper articles about Sun. Reading them leads to the instant urge to ask your doc for Prozac or Effexor. Mostly they are underinformed, heavily biased to other computer companies or just outright sensational. I don´t want to tell you that all is well in the sunny kingdom. But my outlook for the next 2 years looks really positive.

Continue reading "My thoughts about the future of Sun ..."

Immutable Service Containers

nospam@example.com (Joerg Moellenkamp) — Wed, 19 Nov 2008 16:32:49 +0100

Glen Brunette describes at the Sun Wiki the concept of the Immutable Service Containers (ISC):

Immutable Service Containers (ISC) are an architectural pattern used to describe a platform for highly secure service delivery. While the actual instantiation of an ISC can differ based upon customer and application requirements, an ISC embodies at its core the key principles inherent in the Sun Systemic Security framework including: self-preservation, defense in depth, least privilege, compartmentalization and proportionality.

A really interesting concept! By delivering the service by a zone in conjunction of denying write access to the binaries or configurations and vastly locking down the privileges you can build a really secure environment.

Sun slashes 6000 employees

nospam@example.com (Joerg Moellenkamp) — Fri, 14 Nov 2008 19:17:26 +0100

Not a good way to start the day as i´ve found a mail from Jonathan and some Google Alerts mails not much later. Sun will cut costs in the verge of a recession of the worldwide economy. eWeeks write for example in "Sun Flames 6,000 Workers in Cost-Cutting Bid":

The company says that falling demand for high-end hardware is behind its losses and forcing budget and staffing cuts. The layoffs, which will eliminate as much as 18 percent of Sun's workforce, will save the company $700 million to $800 million. The company will incur a charge of as much as $500 million as a result of the layoffs and restructuring its software division.

A walkthrough to the Sun Storage Simulator - Part 1: Initial Config

nospam@example.com (Joerg Moellenkamp) — Fri, 14 Nov 2008 01:30:05 +0100

I hope that you won´t say after this walkthrough, that Sun can't develop a decent GUI. I´ve used the VMware based storage simulator for the Sun Storage 7000 series to create a short walkthrough to the installation of an Fishworks based system.

Continue reading "A walkthrough to the Sun Storage Simulator - Part 1: Initial Config"

Sun Storage 7000 in the words of Kent Brockman

nospam@example.com (Joerg Moellenkamp) — Thu, 13 Nov 2008 19:20:26 +0100

Paul Venezia wrote an article about this opinion about the new 7000 series and comes to an interesting conclusion:

So, in the immortal words of Kent Brockman, I, for one, welcome our new storage overlords.

BTW: He will publish a test of the 7410 soon.

A different modus operandi

nospam@example.com (Joerg Moellenkamp) — Thu, 13 Nov 2008 00:30:09 +0100

Adam Leventhal brings it to a concise point where the new Sun Storage 7000 systems are really different from other systems besides of this cool administration interface:

Further, you can design the specific system you need with just the right balance of write IOPS, read IOPS, throughput, capacity, power-use, and cost. Once you have a system, the Hybrid Storage Pool lets you solve problems with targeted solutions. Need capacity? Add disk. Out of read IOPS? Toss in another Readzilla or two. Write bogging down? Another Logzilla will net you another 10,000 write IOPS. In the old model, of course, all problems were simple because the solution was always the same: buy more fast drives. The HSP in the 7410 lets you address the specific problem you're having without paying for a solution to three other problems that you don't have.

At the end, one of the core differentiators are the hybrid storage pools. ZFS gives us with the HSP an important advantage. Furthermore it makes difficult questions simple. For example mirroring write caches in RAID controllers: Many of our competitors use special interconnects to mirror the cache from one controller to another. When your write cache is a drive in a JBOD you don´t have to mirror the cache, you just import the drive to the surviving filer node.

Many problems become simple, when you think about them from another perspective ... you just have to start to think.

This gives for Sun and the Opensolaris community an important lead: It may be easy to replicate the GUI, it may be easy just to take commodity hardware, but it isn´t easy to copy dtrace and it isn´t easy to copy ZFS.

Interesting links about the Sun Storage 7000 series.

nospam@example.com (Joerg Moellenkamp) — Wed, 12 Nov 2008 18:24:07 +0100

There are two interesting articles about the new storage from Sun: Amitabha Banerje wrote about an benchmark with filebench on the 7410. Another good article is Josh Simons' article about his experiences with the Sun Storage 7000 Unified Storage System simulator in "Unified Storage Simulator: Too Fun to be Legal". I´ve tried it, too and it´s really great. I will write about my experiences when i´m back in Germany.

E25K - the most expensive blade rack of the world

nospam@example.com (Joerg Moellenkamp) — Wed, 12 Nov 2008 01:32:29 +0100

There is another important new product: Sun Fire USBRDT 5240 Uniboard. To get the expectations right: The unibord gets just power and cooling from the E25K chassis, it´s not about building an 36 socket monster CMT server. That would be cool, but that´s outside the specification. When you put 18 of this boards into you E25K you would get the mentioned "most expensive blade rack of the world".

The usecase for this board is a different one. Perhaps you have an perfectly running installation on a uniboard capable system (Sun Fire 4800, E4900, 6800, E6900, F12k, F15k, E20k, and E25k servers are supported) but parts of you appliaction could need some additional horsepower. and you have some free uniboard slots. With such a board you could add it within an existing chassis. Believe it or not ... sometimes it´s easier for a customer to explain an extension of an already existing system to his or her financial deparment than to explain a new system.

Back to the new uniboard: The Sun Fire USBRDT 5240 Uniboard offers 2 UltraSPARC T2+ procs, up to 128 GB of memory, 4 PCI Express module slots, 8 GigabitEthernet-Ports and some other really nifty features ...

Nice side blow ...

nospam@example.com (Joerg Moellenkamp) — Tue, 11 Nov 2008 03:20:23 +0100

David Blasingame writes in his blog:

What do you get when you mix Sun Servers, ZFS, DTrace and Commodity Storage?
A: Nervous competitors.

I tend to agree

Sun Storage 7000 Unified Storage Systems

nospam@example.com (Joerg Moellenkamp) — Mon, 10 Nov 2008 19:46:02 +0100

The Sun Storage 7000 series storage just appeared on the Sun website today (albeit the lauch event here at CEC will only start in 3-4 hours). There are three systems available at the moment: The 7110, the 7210 and the 7410.

The smallest variant is the Sun Storage 7110. This is a network storage system with up to 2 TB.

Sun Storage 7210 is essentially a filer based on the X4540, up to 46 TB storage, but we won´t sell them in such a configuration.

The system is available in 11.5 TB, 22.5 TB, 44 TB raw capacities. The 22.5 and 44 TB devices are equiped with write-biased SSD.

The biggest system (at the moment) is the Sun Storage 7410.

Filer heads with up to 4 quad-core opterons. Up to 6 x 100GB Read Flash Accelerator per controller node). Up to 288 TB (576 TB with free software upgrade). Clustering is availble for this system.

You may have recognized that this systems are really similar to our servers. That´s intended. We take commodity components to build our network storage. But what´s the difference to this servers? It´s the software. Albeit it´s a kind of Opensolaris, we´ve put much effort in building an user-interface to manage those devices in an appliance-like manner. Just have a look to this website to get some screenshots. As usual pictures telling more than 1000 words.

Or even better: As the operating system is commodity and the the hardware is commodity Sun was able to prove a Sun Unified Storage Simulator to get some deep insight into our network storage systems: Download the the VMware image here and run a Sun Unified Storage Similator on you notebook or workstation.

New storage systems

nospam@example.com (Joerg Moellenkamp) — Thu, 06 Nov 2008 19:30:39 +0100

One of the events of the upcoming CEC: We will announce the OpenStorage systems in Las Vegas. Will write about them directly at the event.

OpenSPARC Internals

nospam@example.com (Joerg Moellenkamp) — Thu, 06 Nov 2008 18:09:36 +0100

A really good source of in-depth informations about the UltraSPARC Tx processor line: OpenSPARC Internals. It´s available as a pdf-book for free after registration.

Mysql on SSD benchmarked ...

nospam@example.com (Joerg Moellenkamp) — Wed, 05 Nov 2008 15:46:52 +0100

Roger Bitar did some really interesting benchmarks with mysql on SSD. In his benchmarks he artificially constrained the availability of memory for the mysql buffers (aka cache):

We choose to test MySQL, the open source database (DB), using a simple MySQL benchmark called Sysbench. We populated the Sysbench table with 112 Million rows (around 26GB size) that fit on 1 SSD drive . We executed read-only queries while varying the buffer size. We mounted the file system in DIRECTIO mode to disable file system caching. We perfrormed the tests with regular HDDs and repeated them with SSDs.

There results were somewhat astounding. Of course a fully memory saturated database is the fastest way to implement a database (buffer cache at 24 GB). But with a database on SSD Roger was able to yield 96 percent of the performance of the in-memory database even with heavily constrained RAM space (buffer cache limited to 8 GB). The latency at memory constrained situation is vastly better with SSD than with rotating rust.

This leads me to an interesting conclusion: When you just compare the power consumption of SSD and hard disks (aka rotating rust), you forget a part of the equation. Let´s assume a FB DIMM takes 10 watts (it´s a little bit more, but 10 is easier for calculating). In the benchmark you were able to have almost the same performance with SSD (2 watts) at 8 GB as at 24 GB. Let´s assume 4 GB DIMMS. You would save 18 Watts (10*2-2) by using the SSD. Okay 18 watts isn´t a big deal, but now think about a 240 GB database.

You can design your database server in quite a different manner, when you take SSD into your considerations.

Jonathan about Q1FY09

nospam@example.com (Joerg Moellenkamp) — Thu, 30 Oct 2008 22:19:21 +0100

Jonathan wrote an interesting article about his perspective on the first quarter of the current fiscal year: Understanding Sun's Business - Q1 Results. Really worth a read.

Ouch ....

nospam@example.com (Joerg Moellenkamp) — Thu, 30 Oct 2008 21:17:37 +0100

On our first quarter of this fiscal year:

Net loss for the first quarter of fiscal 2009 on a GAAP basis was $1.677 billion, or $(2.24) per share on a diluted basis, as compared with a net income of $89 million, or $0.10 per share, for the first quarter of fiscal 2008. GAAP net loss per share includes a $1.445 billion non-cash charge for goodwill impairment. It also includes a restructuring charge of approximately $63 million pursuant to the restructuring that commenced in the fourth quarter of fiscal 2008.

On a non-GAAP basis, net loss for the first quarter of fiscal 2009 was $65 million, or $(0.09) per share on a diluted basis, as compared with a non-GAAP net income of $285 million, or $0.32 per share, for the first quarter of fiscal 2008.

Memory prices for SPARC Servers.

nospam@example.com (Joerg Moellenkamp) — Wed, 29 Oct 2008 10:42:55 +0100

One of the critics about the M3000 isn´t the system itself. It´s the price for the memory. But as alway, this isn´t about making Sun rich. There are technical reasons for this: Let´s assume you buy an x86 system. Most of the systems are phased out within 3 years or so. Now think about Suns: I know several customers, that still uses E250 or E450 for certain tasks. Without any problems. This systems were current system 8 years ago. But they still work. The reason for this: Extreme quality standards for components.

It´s important to know, that electronics scatter vastly in the fulfillment of their specifications. This is the reason, why there is a frequency number on your processors, on your memory. Because the fulfillment of specification may vary with the frequency or with the temperature or the age. So you test your electronics and print the most expensive frequency in accordance to your specifications on the chip casing (okay, there are some problems with a matured manufacturing, sometimes you don´t produce enough low specification modules, so you have to downgrade better parts). This scatter in quality lies in the nature of mass producing electronic parts.

When Sun wants to sell memory, the standards are really rigid. At first, we don´t take the memory from the spot market and sell it. At first Sun defines exact specs for the memory. If the memory module doesn´t fullfil this specs within a really thin margin ... back to the drawing table for the manufacturer. When the manufacturer fullfil this specs, Sun takes modules from the last few months from all fabs of the manufacturer and check the modules by aging them artificially: A month or two at a temperature above the specs running at the threshold frequency. In a month you can simulate a several years lifetime of the memory module.

When more than really small number of modules fail in this time ... well ... back to the drawing board for the manufacturer. Just when all tests are fulfilled within the really narow test specification, the memory modules are used for the our SPARC servers.

So: Why do you have to spend so much money for the memory. Well ... if you want DIMMS within +/- 10 percent of the specification you can choose from a vast amount of modules. If you want memory modules within 0.5% of your specification, the choice gets really thin. At 0.1% they are hand selected (this are not the current numers ... numbers just choosen to give you an impression). The same for long time stability: If you want a series of memory modules with a lifefime of 3 years, you get choose from a vast amount of memory modules. At 10 years the story looks really different. So we want the cream of the crop of memory modules in at least two dimensions. And when you want to have the best quality, you have to pay for it. So we can´t simply take this el-cheapo DIMMS from the market and put it into a SPARC server. We have to buy the expensive modules from the manufactures, as the manufacturer know as well, that they sell their top quality to us.

To answer the overarching question: Does high-quality memory modules really matter? Yes ... definitly. Perhaps not for your PC at home. But surely for systems running your business for the next years.

Sun SPARC Enterprise M3000 announced

nospam@example.com (Joerg Moellenkamp) — Tue, 28 Oct 2008 14:08:58 +0100

Sun and Fujitsu jointly announced the system codenamed Ikkaku today : Sun SPARC Enterprise M3000.

The M3000 is a single-socket system for the SPARC 64 VII CPU, thus it´s a 4 core system, executing up to 8 threads. Up to 32 GB memory. 4 PCI-Slots. So it´s an entry-level system. Starting at $14.795,00.

But before you compare it with your favourite brand of el-cheapo x86 systems. This system shares many RAS features of his bigger brothers: Instruction Retry on processor level ... ECC for memory ... ECC for the interconnect between system controller and the CPU ... ECC for integer registers ... parity for floatingpoint registers ... fault isolation ... a single defective core doesn´t keep the system from starting up, etc. ... well ... just read the architecture whitepaper on your own.

SAP Benchmarks revisited

nospam@example.com (Joerg Moellenkamp) — Mon, 27 Oct 2008 19:51:02 +0100

There is a new benchmark from Sun for SAP on the SAP website. SAP certified the benchmarking of the Sun Fire X4600 with quadcores. This benchmark result is especially interesting when you compare it to another 8 socket system - the HP ProLiant DL785. Both systems have the same basic characteristics:

The Sun hardware:

Sun Fire X4600M2, 8 processors / 32 cores / 32 threads, Quad-Core AMD Opteron Processor 8360 SE, 2.5 GHz, 128 KB L1 cache and 512 KB L2 cache per core, 2 MB L3 cache per processor, 128 GB main memory

The HP hardware:

HP ProLiant DL785, 8 processors / 32 cores / 32 threads, Quad-Core AMD Opteron processor 8360 SE, 2.5 GHz, 128 KB L1 cache and 512 KB L2 cache per core, 2 MB L3 cache per processor, 128 GB main memory

The Sun system uses Solaris 10 and MaxDB 7.6, the HP system uses Windows Server 2003 Enterprise Edition and SQL Server 2008.

The Sun system yields 5,800 SD Users and 29,670 SAPS (SAP Benchmark Certificate: 2008061). The HP system yields 5,230 SD Users and 26,180 SAPS (SAP Benchmark Certificate: 2008026). That´s quite a difference, but not the end of the story: The Sun result is a 6.0 unicode result, the HP is a plain 6.0 result. You may remember: You loose at least 15% (depends on the systems) performance by using unicode. That´s a really impressive benchmark result. It´s the fastest 8 socket x86 SAP benchmark result at the moment.

As both system uses different operating systems and different databases, you can´t say if it´s a bad idea to use SAP on HP or a bad idea to use SAP on Windows/SQL Server

Andy will work for Arista Networks as the Chief Development Officer

nospam@example.com (Joerg Moellenkamp) — Thu, 23 Oct 2008 11:05:00 +0200

Hmm ... there is a lot of buzz around the leave of Andy. I was shocked in the first moment. But as usual nowadays, there are often missing parts in reports. I don´t know what i´m allowed to say here, thus there will be only one cryptic comment from my side to this: Don´t take all this stories at Forbes or Businessweek for granted. The real story is pretty much different from the stories in the most of papers so far. The NYT got it fairly right:

But he said he would retain a part-time advisory role at the company. Its my baby, Mr. Bechtolsheim said. I will always be associated with Sun.

New blades from Sun

nospam@example.com (Joerg Moellenkamp) — Wed, 22 Oct 2008 13:00:33 +0200

Sun announced a number of new blades for the Blade 6000 chassis yesterday. From my perspective two of them were especially interesting. The first one is the Sun Blade 6000 disk module. It´s an 8-slot disk module based on SAS (you will find in-depth informations in the Sun Blade 6000 Disk Module Configuration Guide manual).

The other important was the Sun Blade T6340 server module. This blade is equiped with two UltraSPARC T2+ procs. Thus you have 128 threads in one blade in a system and up to 256 GB of memory. When you look at the blade, you will recognize that a large part of the circuit board is occupied by memory slots. At the moment this is the blade with the biggest memory capacity in the industry.

In addition to this both blades we´ve announced a dual-proc Opteron based blade with the X6240 for the Blade Server 6000 chassis. The Blade Sun Netra CP3250 ATCA. This isn´t a standard blade for the 6000 or 8000 blade chassis. It´s a Xeon server blade for usage in telco blade servers.

Is iSCSI really dead? The Register thinks so ...

nospam@example.com (Joerg Moellenkamp) — Wed, 15 Oct 2008 18:49:25 +0200

The Register thinks that FibreChanneloverEthernet(FCoE) will take over the datacenter and proclaims "iSCSI: Game over". Okay, no problem .... there is already a FCoE implementation for Solaris. Been there - done that. There is only one simple problem: I´m not sure, that FCoE is the way to go.
Continue reading "Is iSCSI really dead? The Register thinks so ..."