Frank Lloyd Wright didn't see his structures as empty buildings, but as holistic environments.  That's why his designs often included not only the roof and walls but the chairs and tables as well.

Google is the Frank Lloyd Wright of data centers: they design their own storage systems that function as an integral part of the whole. 

You can see the power of this philosophy in their brutal efficiency.  Information Week points out that Google-designed data centers use nearly five times less energy than conventional data centers.  That correlates with the 4-to-1 ratio in cost per capacity that IDC found in comparing conventional data centers to Google-style "content depots".

Google has uncharacteristically given us a peek behind the curtain to encourage sustainable energy practices. Sure, being eco-friendlier is good.  All the more so when it aligns with Capitalism.

How can conventional data centers compete without this approach?

More learnings from the IDC Enterprise Disk Storage Consumption Model report: 

Since 2005, the Cloud's share of storage capacity has grown from less than 5% to almost 20%.  Yet the Cloud's share of storage revenue has risen to only about 5% in the same time period. 

How can "Content Depots", as IDC calls them, build data centers for one fourth the cost (storage-wise) of conventional corporate data centers?

They use more high-capacity SATA drives, but that's only part of the answer.  Rather than using off-the-shelf enterprise storage systems, they are often building their own storage system/server contraptions to strip away anything that doesn't add value as a big bucket for internet bits.  

Granted, these systems aren't going to support your everyday ERP system.  But these changes should be watched closely.  Innovation by these end-users will migrate back to the rest of the industry to give us all more bytes per buck. 

It already is - in many of the latest greatest storage upstarts' products.

Amazing german guys ! :D

Storage is growing everywhere, but nowhere more than in the Clouds.  Here's a link to my guest post on Sitepoint on storage technology and philosophies that can help battle the challenges of billowing Cloud Storage.

"Qualcuno dice che è inevitabile - spiega Stallman al Guardian - e quando sentite qualcuno dire così, è molto probabile che si tratti di una strategia d'affari per renderlo vero". L'opinione di Richard non è isolata nel panorama IT, ma appare senz'altro una delle più categoriche: non ci sarebbe alcun bisogno di ricorrere alla computazione remota di massa per tirare avanti, bastano e avanzano gli strumenti attualmente in circolazione. Meglio, ovviamente, se free.

"Una ragione per non usare le web application è la perdita del controllo" prosegue Stallman: i dati fluiscono liberamente tra qualsiasi postazione o thin client in giro per il mondo e i datacenter, ma a scapito della capacità del legittimo proprietario di disporne a suo piacimento. Basti pensare a cosa accadrebbe nel caso in cui un account venisse sottratto al suo titolare: da quello stesso account potrebbe partire una reazione a catena, che coinvolgerebbe tutti gli altri servizi ad esso collegati, stravolgendo le attività personali e lavorative di quello stesso individuo.

Ma c'è di più. L'avvento del cloud computing secondo Stallman non fa altro che riproporre l'antica questione del software free contrapposto a quello non libero: "È un male proprio come usare programmi proprietari". In ballo ci sarebbe persino la libertà personale: "Fate il vostro lavoro su un vostro computer con un programma che rispetti le vostre libertà: usando un programma proprietario sul server di qualcun altro si è senza difese. Vi state mettendo nelle mani di chiunque abbia sviluppato quel software".

Il rischio, prosegue Stallman, è che se all'inizio questi servizi possono apparire più economici (o addirittura gratuiti) rispetto alle abitudini attuali, nel lungo periodo possano invece rivelarsi oltremodo costosi. E soprattutto, l'intera mole di informazioni personali (foto, appunti, appuntamenti in agenda) o aziendali (budget, bilanci, piani strategici) sarebbe affidata alla onestà e alla solidità di una azienda, esponendosi a tutti i rischi di boicottaggio o incidenti che questo comporta.

Che si tratti o non si tratti di una semplice moda, le aziende impegnate nel settore si stanno rimboccando le maniche per trasformare il cloud computing in una realtà sempre più stabile e efficiente. Se poi tutto questo dovesse tradursi in una breve (ennesima) parentesi passeggera non è dato saperlo: di certo, alla fine, l'informatica avrà fatto degli importanti passi in avanti nella programmazione parallela e nella computazione su vasta scala.

Fonte: punto informatico

Daniel Dern's common-sense summary of what SMBs can do to reduce their storage energy costs and footprint rings true for companies of all sizes.  It's uniquely practical - he avoids the "silver bullet" approach common to this topic.  Like most things, it takes a lot of little steps to make a big difference. 

His recommendations:

1. Store less data. Good luck!  But maybe growth can be slowed.

2. Use fewer disks.  Fill up the ones you have, and combine multiple disks into today's high-capacity behemoths.

3. Turn off idle disks.  Like turning off the light when you leave a room, it's a little added work that can add up to real savings. 

4. Use the right disks.  Match storage with the performance and recall demands of the data.

5. Use more efficient disks.  Converting to 2.5" enterprise drives reduces power and speeds up I/Os. 

6. Use more efficient storage systems.  Some storage systems leverage disk-level energy efficiency features with system-level power savings.

7. Improve cooling.  Sometimes it's as simple as keeping things in proper working order.

8. Exploit "Green" rebates. Get full credit for your efforts with Utility and government incentives.

Who's got additional ideas to add to Daniel's list?

In their August Storage Consumption Model , IDC's Richard Villars talks about several intriguing trends that are changing the way storage is designed, distributed and used. 

The one that caught my eye is what he calls the "serverization" of storage platforms - the addition of computing power to storage systems, especially for mass-scale clustered storage data factories.

A parallel trend is that mainstream server storage capacity is growing incredibly fast.  Storage capacity (often measured in terabytes) now stands next to processing power as cornerstone specifications for most servers.   

It begs the question:  is the difference between a server and a storage system becoming insignificant? 

I think we need a new name for systems that blur the line. Let's call them Stervers.

Thoughts?  Is this a substantial shift, or just a reinvention of the same old processing/storage partnership?

No preguntes cuánto cuesta, pregunta cuánto gasta. Las máquinas van baratas, no así el consumo energético. Lo último, los servidores. Mientras que los costes de electricidad y de climatización para que funcionen los servidores se han multiplicado por ocho en los últimos años, y los de gestión y administración por cuatro, el precio de los servidores se ha reducido a una décima parte desde 1998. La mayor disminución de precio se ha producido en los últimos cinco años.

En 2003, un servidor costaba alrededor de 30.000 euros. Ahora, una máquina con mucha más capacidad de proceso se puede comprar por 5.000 euros. Así, el coste de la energía eléctrica y de la climatización necesaria para mantener funcionando en perfectas condiciones los servidores comienza a sobrepasar el precio del servidor. La consultora Gartner estima que en 2008 el 48% de los presupuestos de tecnología se destinará al pago de las facturas de consumo eléctrico.

Soluciones verdes
El 59% de la energía que se consume en un centro de proceso de datos (CPD) se destina al funcionamiento de las máquinas. El 41% restante se emplea en la iluminación de la sala, en los sistemas de alimentación ininterrumpida (SAI) y, fundamentalmente, en la climatización. Para lograr un buen acondicionamiento ambiental se recomienda que la temperatura de la sala se encuentre entre 21 y 23 grados. La humedad relativa debe ser de entre el 45% y el 55%. Estas temperaturas -que deben medirse en el frontal del rack- son las que establecen los fabricantes de equipos.

Desde hace un par de años, las soluciones verdes, que llegan a consumir hasta el 50% menos de energía, figuran entre las propuestas de todos los fabricantes. Para algunos analistas especializados, los CPD ecológicos son uno de los temas estrella del sector. A la hora de valorar las tendencias de futuro, lo relacionado con el consumo de energía despierta mucho interés entre los directores de tecnología. No resulta extraño que, en muchos casos, se opte por soluciones drásticas, como mantener las salas a temperaturas inferiores a 20 grados, con el aumento en el coste de la factura eléctrica que comporta, antes que por realizar inversiones en equipos e instalaciones que, según los fabricantes, se pueden amortizar en muy poco tiempo por el ahorro en la factura eléctrica.

Hasta hace poco, se tendía a instalar los CPD en sótanos para no malemplear el espacio más noble. Las cosas están cambiando. Empresas como Yahoo!, Microsoft y, por supuesto, Google invierten millones de euros en lo que se conoce como granjas de servidores, que crecen de forma exponencial.

Se calcula que Microsoft puede estar instalando 20.000 servidores por mes para acercarse a Google, cuya red supera el millón de servidores, según explicaba The Economist hace unas semanas. A juicio de la revista, estos nuevos gigantes tecnológicos pueden ser comparados con las fundiciones de aluminio por la energía que necesitan. Como si se tratara de una vuelta al pasado de la industrialización, a la hora de elegir el emplazamiento, los promotores no dudan en buscar lugares adonde se pueda hacer llegar la fibra óptica con facilidad y que se encuentren próximos a las fuentes de generación de energía eléctrica.

Con el objetivo de progresar en el mundo de los servicios Web, Microsoft ha desarrollado un plan de granjas de servidores en los últimos 18 meses en el que figura un CPD en el pueblo de Quincy (Washington), que disfruta de una tarifa eléctrica a mitad de precio de la media nacional, una buena razón para elegirlo como granja de ordenadores. El nuevo CPD de Microsoft consumirá la misma energía eléctrica que 40.000 viviendas.

Fuente: El País

What did a year buy us? (along with a substantial sum involving many zeros and commas invested) 

Fully virtualized, bicoastal, server environment...check

Centralized SAN consolidating all data with auto-replication to secondary site...check

Enterprise-wide MPLS WAN deployed for increased and flexible network connectivity...check

Enterprise grade remote access platform...check

Conversion of our secondary data-center on the east coast to a fail-over site...check

Well, this is the makings of a great DR plan. I mean, what more could I ask for, right? ;-)

Heh..heh..heh...my director should have thought twice before asking me that! I mean what else could we possibly need to execute a controlled and orderly fail-over to the secondary site should our west coast data-center take a dump, or vice versa in a time-efficient manner?

Currently, we use several monster scripts and additionally have to manually do some setting and network configuration changes to affect Tier 1 systems and full network failover to the secondary DR environment. Maybe 8-12 hours total of work depending on the severity of the disaster including initial disaster and site impact assessments. Pretty good compared to the nightmare that was our DR plan before in the traditional server environment. In those days, we would have been looking at about a week at least to get our tier 1 applications fully restored and up and running along with network operations failed over. Not a pretty sight at all. I shudder to think if we ever had to go through that with the original setup.

 Enter VMWare SRM (Site Recovery Manager)

I have started looking into this baby and let me tell you, this seems like just the ticket for what we are looking for.

So what is VMWare SRM? Well, here's the blurb from their fact sheet:

"VMware Site Recovery Manager (SRM) is a new VMware 

Infrastructure 3-based solution that is tightly integrated with Dell 

EqualLogic PS Series storage to offer centralized disaster recovery 

management, automation, and testing for a virtualized datacenter. 

The integrated solution  utilizes the PS Series’ native Auto- 

Replication feature, integrated directly into Site Recovery Manager 

using specialized storage adapter software developed by Dell.  SRM 

enables non-disruptive, automated testing of recovery plans and 

automates the recovery process, while the PS Series supplies cost 

effective and easy-to-configure replication over existing IP networks. 

The core server virtualization platform that provides IT environments 

with the tools to help reduce the cost and complexity of managing 

their IT infrastructure now expands to simplify disaster recovery as 

well.  Storage and server virtualization solutions from VMware and 

Dell provide the infrastructure for the fully virtualized data center, 

enabling a dynamic, highly automated computing environment."

Sales-speak eliminated, what this add-on should do is effectively cut our fail-over times in the event of disaster to at least half of our original estimates. Depending upon the severity, they claim even shorter times for automated execution of the disaster plan!

The beauty of this solution lies in its tight integration of the VMWare platform as well as the SAN platform we are currently using in production. This means that implementation is as simple as drop-in placement and configuration because it leverages our existing infrastructure technology. Financially, it provides a compelling argument in relation to overall long-term ROI versus our initial investments. In fact, it actually enhances the ROI overall in relation to our total virtualization/SAN/DR project as the last puzzle piece.

Now I'm the type of person that needs to see to believe so you can bet your bottom dollar I'm gonna be testing this sucker out before we even think of spending a penny. Until I see it work in our environment, it's just a lot of pretty words on paper. So until then, the jury is out. Stay tuned.

El aumento de los data-centres, necesarios para cubrir el abundante cauce de información generado en la web, ha provocado que las compañías informáticas tomen soluciones radicales con el fin de reducir costes. Las supercomputadoras alojadas en los data centres, que pueden tener el tamaño de un campo de fútbol, usan enormes cantidades de electricidad para evitar el sobrecalentamiento de las máquinas.

En un intento de solucionar el problema, ya que no es muy ecológico, Microsoft ha investigado la construcción de un data-centre en climas muy fríos como Siberia, mientras que en Japón la empresa tecnológica Sun Microsystems planea enviar sus supercomputadoras a una mina de carbón abandonada, usando el agua del suelo como refrigerante. Sun asegura que pueden ahorrar hasta 9 millones de dólares en electricidad por año.

Los expertos declararon que el proyecto de Google Computer Navy es una "inesperada pero inteligente solución". Rich Miller, autor del blog sobre el conocimiento de los data centres asegura que el proyecto es "realmente innovador".

Google no ha confirmado cuando tendrá lugar el desplazamiento de las supercomputadoras a los barcos, aunque la compañía ha afirmado que están archivando las ideas que surjan sobre la variedad de aplicaciones de la patente, "algunas de estas ideas se convertirán en productos reales, servicios o infraestructuras y otras no".

Fuente: El País

Facing increased pressure from rivals Microsoft and Citrix Systems, VMware will announce what it calls a 'virtual datacenter operating system' at VMworld this week to extend virtualization beyond servers and into all corners of the datacenter, including storage and network equipment. The products are scheduled for release in 2009. The 'VDC OS' is not a product but a set of capabilities that will appear in VMware Infrastructure 3 updates and other products.

There is no information about the pricing or a delivery timetable yet, but the list of new features is promising.

The new products can be broken roughly into two categories: software that works at the virtual machine level for improving application performance and availability and infrastructure products for managing the wider datacenter.

List of modules in VMware 2009:vNetwork - Virtual switch for a pool of virtualized servers vStorage - Efficient "thin provisioning" for storage allocation to virtual machines vCenter - An updated Virtual Center management suite. New modules include CapacityIQ, ConfigControl and Orchestrator. vCloud - A set of technologies that enables hosting providers to act as cloud environments. vAPP - A development tool for software vendors. vClient - A new client virtualization layer for desktop, laptops and smart phones

VMware will be releasing APIs to increase the visibility to many of its modules. VMware's initiative to open its architecture to other vendors is a promising step in the right direction.

Seems people at Google are desperate to find new ways to save energy / money and to reduce pollution. They have recently filed a patent that shows us how a datacenter in the middle of the sea would be feasible.

Using cold water to cool down the servers and wave movement to generate electricity, this would mean that the datacenters could be self sufficient.

I have to ask: ''Why not add some solar panels to that?''

Times Online

Analysis: In 1994 I moved from San Jose to San Francisco, and began working for Frank Jordan, Mayor of San Francisco, doing a variety of things - writing his speeches ("Fellow citizens, I can report to you that the State of our City is strong"), running his communications office over at the reelection campaign, and leading a small strategic staff to devise comprehensive policy proposals on future transportation, homelessness, and economic growth.

I had done the same thing before elsewhere. Many of the ideas were ones we borrowed for San Francisco from the emerging city to the south, San Jose - where I had previously held the same multifaceted job for Mayor Tom McEnery, the legendary "Mayor of Silicon Valley."  The advisor who knew both mayors well, Clint Reilly, had recruited me to move north not because of my political leanings - but for my lack of leanings, and preference for solutions.  So in San Francisco our new approach was pragmatic and non-ideological, with a heavy tech focus. 

To foster growth, we focused on policies designed to spark San Francisco's then-still-embryonic technology startup culture, which through a combination of incentivizing enterprise zones, hands-off tax policies, goosed-up entrepreneurial gunslinging - and geographic good fortune - was to grow into a powerhouse of technological activity that threatened to shift Silicon Valley's geographic center of gravity several notches north over the next fifteen years. The Multimedia Gulch and South Park neighborhoods became known for emerging Internet-based startups such as Organic (now part of Ominicom Group), Macromedia (now part of Adobe), and even the iconic WIRED magazine itself, and before the bust came, SoMa gave birth to over a thousand dot-com businesses in the city. 

(Oops, sorry - "The City," as Herb Caen would correct me - he was my upstairs neighbor in the Brocklebank for his last years.)

In the spring of 1995 we set up a pioneering political website for the Mayor's reelection campaign that year, surprising the local "old media" world with our heavily hyperlinked pages and near-real-time reposting of ads, speeches, and news coverage.  Despite all that, the Mayor wound up losing reelection, and while the winner (Willie Brown) began to plagiarize our tech policies immediately - a satisfying conclusion from one perspective - he didn't really know what he was doing, nor did the political hacks he scattered throughout his administration.

Here's a synchonicity that Herb Caen would have loved: that same election year of 1995, while we were toiling away at reinventing government's use of online information, just down the street an unknown 30-year-old man named Terry Childs was convicted on a weapons-possession charge, just one in a string of criminal convictions in his checkered life.

Usually with that kind of story you'd ask, "Whatever happened to Terry Childs?"  In fact, within a decade he was one of the most important people in San Francisco's municipal technology world, holding the relatively minor post of network administrator in the city's Department of Telecommunications and Information Services, but more importantly holding all the keys and passwords to the city's WAN and datacenter. 

If you haven't been following this affair, well you should be for entertainment purposes alone.  Among the allegations in the bizarre case, as summed up by InfoWorld:

[Childs] was arrested July 12 on charges of network tampering after he refused to provide his superiors with administrative access to the city of San Francisco's network, which he had managed for the past five years.

Initially Childs refused to hand over administrative passwords to the city's routers, which had been configured to wipe out all configuration information if they were reset.

After a dramatic jailhouse meeting with San Francisco's mayor one week after his arrest, Childs handed over the data, but DTIS Chief Administrative Officer Ron Vinson said Wednesday that the city now expects to spend more than $1 million to clean up the mess."

There's a lot of interesting reading from the government IT perspective on this case, not least some telling observations on the lack of governmental understanding of routine IT terms and practices (see for example Paul Venezia's technical dissection of the case on his blog).  Internetnews.com has a disturbing background piece on the overall case ("Rogue Sys Admin Still Haunts S.F."), and InfoWorld keeps an updated "Special Report" on the case.

What led to this idiotic cock-up? A million factors, I'm sure - but a central one was the ham-handed and increasingly bureaucratic approach to urban tech policies during the eight-year administration of Mayor Willie Brown, who followed my boss in office.  I'll let City Journal's words sum it up from their definitive 1998 article, "Willie Brown Shows How Not to Run a City" - and keep in mind this was written early on in the deterioration of the job machine that had been painstakingly built:

Revenues of interactive media firms in San Francisco increased 150 percent last year, and Multimedia Gulch now employs 35,000 workers, up 70 percent from 1995. The highly mobile industry is worth approximately $2 billion to the city's economy. Multimedia executives could probably live with the city's inflated rents, but they grumble ominously about the long wait for required permits that makes expanding a successful business time-consuming and costly. A study by the Coopers & Lybrand accounting firm reports that close to a third of these innovative companies are considering leaving San Francisco. Should the industry trickle out of town, San Francisco's economic horizon will darken further. San Francisco depends on its rich cyber-yuppies, its growing local high-tech sector, and its booming tourism industry to fund a grossly bloated municipal welfare state. Mayor Brown has fattened the city budget by over $1 billion in three years—a whopping 30 percent increase."

Those policies, and the dot-com bust, led to a widely chronicled exodus of talented tech workers, and a related decline in the talent pool for San Francisco's government/tech policymakers and administrators.

The chickens are still coming home to roost. Willie Brown's successor, Mayor Gavin Newsom, a friend of mine back in the halcyon early '90s, has his hands full today with the wreckage wrought by Terry Childs, and this is no time to point fingers. I'm hopeful that Gavin will push forward with an aggressive reform broom and clean out the bureaucratic muck in the city's tech stables.

Here's a link to the latest news on the Terry Childs case.

 

Email this post to a friend

Parascale's got a new angle on the Cloud: a software-based clustered file system that can form a cloud across hardware systems from different vendors.  Their sweet spot seems to be companies that need their own internal clouds.  Here's the release from Byte and Switch. 

DCIG wonders if companies are ready for this atmospheric responsiblity.  

I'm a bit skeptical of hardware independence, based on first-hand experience at the turn of the millenium trying to get a heterogeneous storage pooling product to fly at StorageTek.  Anyone remember the SN6000?

It's harder to achieve than it seems.  But the value is tremendous if Parascale can pull it off.

As the number of Cloud varieties grow, I think we need a Web Meteorologist.  What's your forecast for DIY clouds?