“Physicalization” uses mobile tech to offer an alternative to virtualization

The server space has run into a problem. The continued operation of Moore’s Law is delivering greater amounts of computing power, storage, and bandwidth every year, and the demand for these capacities is escalating as well. However, our ability to efficiently make use of this capacity under one OS instance isn’t increasing as rapidly. This is aggravated by the trend toward multicore computing, which enables a level of parallelism that modern software still hasn’t fully accommodated.

The answer to this dilemma on the software side is the same as it is on the hardware side: increase parallelism by introducing a greater number of OS instances per processor, just like we’ve introduced more cores per processor. One technology for doing this is virtualization, which uses large, multicore, multisocket servers to run many OS instances. Virtualization has been on a massive upswing in adoption in the last few years, as this problem becomes more and more pressing. But management overhead is one of the by-products of virtualization, and great care is required to make virtualization work properly.  So some small datacenters may be looking for an alternative to or a complement to a full virtualization deployment.

Several server vendors are pushing an alternative solution that could be a better fit for some types of workloads; this solution often goes by the name of “Physicalization.” Physicalization involves building smaller servers out of low-power processors, tailoring them to the amount of computing resources which can be efficiently used by one OS instance, so that a one-to-one correspondence between hardware and instances can be maintained, thereby bypassing the management overhead and licensing costs of virtualization. As a side benefit, low-cost, low-power hardware, like the Intel Atom designed for netbooks, can be employed, rather than expensive server hardware.

A number of physicalization solutions have been announced in recent months, and while these solutions can’t offer greater density or PPW, they do represent a workable alternative to virtualization for some users. For instance, web hosters can offer their clients dedicated hardware, while using commodity hardware made cheaper by consumer economies of scale.  They also allow centers to harness the “free first gigabit” of ethernet over copper, and potentially avoid expensive 10GigE and Fibre installations. For some applications, physicalized servers may be the best option for using Moore’s Law to meet the ongoing challenges posed by increasing demands and limited resources in the datacenter.

It remains to be see, however, whether physicalization will actually catch on. The challenge that physicalization faces is that, for all its benefits, it simply doesn’t make as efficient use of Moore’s Law as virtualization does. Let me explain.

The problem with physicalization

Moore’s Law is most often cited as a statement about “computer power” or “the number of transistors on a chip” doubling within a specific timeframe (typically 12 or 18 months). This explanation is serviceable enough for the popular press, but it’s not entirely accurate. In its original 1965 formulation, Moore’s Law was about the cost-per-transistor dropping as a result of increased integration. In other words, as you pack more transistors into the same-sized area of die space, the cost per transistor drops dramatically given a fixed cost-per-chip.

Chipmakers take advantage of this declining transistor cost in two distinct ways: 1) they pack more transistors into the same-sized chip and charge customers roughly the same amount, but for more computer power, or 2) they make more chips that are smaller and cheaper, so that customers get the same amount of computer power but for less money.

The first option is typical of the server space, and the second is typical of the mobile space. But what physicalization tries to do is to take the second option into the server space. Unfortunately for the proponents of physicalization, if you’re going to put more hardware into the same machine, it’s always cheaper and more power-efficient to integrate that hardware onto a single processor die. This is because, as an iron-clad rule of computing, die-level integration is always cheaper than board-level integration.

So by taking multiple smaller, cheaper chips that are intended for separate mobile products, and using board-level and network-level integration to cram them into a single server product, physicalization vendors are using Moore’s Law in an odd, non-standard way that forfeits one of its major advantages, i.e., the advantage of die-level integration over board-level integration. …
Read More »

Think Like a Startup

Many companies practice zero-based budgeting. The idea is that you do NOT start with last year’s budget. You don’t say, “We spent $xx in this category last year, so how much more or less do we spend this year?” You start with a clean sheet. This has the effect of getting rid of those costs that have always been there just because…well just because they have always been there. And that frees you to spend on what you really need, not on what you needed or thought you needed several years ago.

You can take the same approach to IT. Think like a startup, even if you have been in business for decades. Startups do almost all their IT in the cloud and using software as a service (SAAS). That is the default. A startup would think really, really hard about doing anything in-house or custom. Unless something is absolutely core (which means you should be the best in the world at doing it) it is best outsourced to a firm that does that as a core competency.

It is agility that really matters today. Huge companies are crumbling before our eyes. Watch Citigroup and General Motors and your daily newspaper as they go from invincible to suffering or dead. Huge companies have huge overhead. But medium-sized companies cannot afford that overhead.

In 1955, Fortune 500 companies accounted for one-third of the GDP in the US. In 2000, that had doubled to two-thirds. Within that cold statistic lies thousands of human stories of family farms, Mom & Pop stores and other small businesses trampled by Wal-Mart, Agribusiness and other large companies.

That is a massive shift. But it is not written in stone that large companies should control two-thirds of the economy. We may have seen the high water mark of this trend. It maybe reversing. The demise of huge companies is a massive, historical opportunity for smaller companies.

Mid-sized companies need to decide: Are you are agile like a startup or a smaller version of a large company? The latter may give you all the problems of a large company without their advantages of scale.

So, think like a startup. Only do some IT functions internally if you are really (no, really) convinced that those functions are totally core to your mission and you can be the best in the world at them.

That is easier said than done. You do have legacy IT and processes. The most important thing to recognize is that people’s jobs are at stake. If you make the process of moving to cloud/SAAS a genuine win for your people, those objections (”It’s OK for some companies, but its way too risky for us for this reason and this and oh this one as well.”) will disappear like morning mist.

Somebody who is maintaining/managing in-house systems needs to see one of two future career paths:

1. Within your company, moving from context to core, from cost center to profit center.

2. Outside your company, working for cloud/SAAS vendors, trading on the success of the migration of your systems.

Either is a good result, and it will vary by individual.

…
Read More »

Introduction to Server Virtualization

What Is Virtualization? What Are the Benefits?

A typical data center contains many servers. Traditionally, since hardware was cheap and applications were complex, many administrators installed one application per server. Though this provided their applications with security and isolation from other processes, it resulted in server sprawl. Workload on these servers would vary widely depending on the time of day and user activity. Many of the servers would spend a large percentage of their time idling away, doing practically nothing. Average server utilization levels are typically about 10%. Server virtualization increases each physical machine’s utilization by dividing physical servers into multiple virtual servers. Each virtual server contains its own operating environment and applications; it looks and acts exactly like a real server.

There are three ways to run virtual servers: full virtualization, para-virtualization and OS-level virtualization. In all three, the physical server is called the host and the virtual servers are called guests.

Full virtualization uses a special kind of software called a hypervisor to supervise the guests. It interacts directly with the host’s CPU, disk and network access. The hypervisor keeps each of the guests independent of one another and unaware of the other servers running on the same machine. Full virtualization requires the most overhead, and thus allows the least number of virtual servers per physical server.

With para-virtualization, the guest servers are aware of each other’s existence. Though it also uses a hypervisor, para-virtualization doesn’t require as much overhead as full virtualization because the guests are each aware of the load that they’re placing on the host. This method requires that the host be running a specially modified version of the operating system to cooperate with the hypervisor, but it offers a considerable performance improvement over full virtualization.

An OS-level virtualization doesn’t have a hypervisor at all. The host’s operating system performs the hypervisor’s duties and interacts with the guests. The biggest drawback of OS-level virtualization is that all of the virtual servers must run the same operating system as their physical server, but the performance in the guests can be similar to an actual physical machine!

Which is best? It depends on the needs of the data center. If all the servers run the same operating system, the OS-level virtualization would be the fastest and allow the most virtual servers per host. If you need to combine a number of servers running different operating systems, both full virtualization and para-virtualization can do the job, but full virtualization is the older and more mature technology for doing so.

When to use Virtualization?

Virtualization is perfect for small to mid-size applications and services that don’t fully tax the hardware resources. Since virtualization is dividing a server’s resources into fractions, applications that are too resource-intensive can overwhelm the server. When the physical server can’t meet the demands of one of the virtual servers, all of the virtual servers will bog down.

Virtualization allows for server consolidation. If there are many applications that each only use a small amount of processing power, the administrator can consolidate several servers into one virtual machine. In a data center with hundreds or thousands of computers, the power and cooling savings would be significant.

In the past, when you upgraded your server hardware because it was obsolete, you were usually required to upgrade, or at the very least, extensively test all of the software running on it as well, because the old version of the software might not run the same on the new hardware. With virtualization, though, it is possible to maintain the same operating environment regardless of any changes to the physical infrastructure of the server. This has far-reaching implications for server upgrades, because upgrades are typically incredibly expensive, not just because of the cost of the actual hardware but because of the associated costs of testing to make sure that that everything behaves as it should. …
Read More »

Recap of Recent Webinar: Managing for Growth

Did you miss our recent webinar, “Managing For Growth And Change In A Tough Economy”?

If so, you can still watch the full webinar here. Meanwhile, let me give the highlights of some of what you missed 

The goal of the webinar was to outline strategies and tips for growing your business by leveraging information technology.  I was the moderator and one of the speakers. I was joined by two extremely knowledgeable people:

• Chris Peters, End User Product Marketing Manager, Server Product Group – Intel
• Mac McConnell, Global Program Manager, Systems Group at Sun Microsystems

Business Outlook

First we covered the current business and economic environment. Our collective view is that although the economy is still in the midst of a recession, some cracks are starting to emerge in the dark facade. We noted the small signs of improvement here and there. For instance, U.S. Treasury Secretary Timothy Geithner has reported that the global recession has stabilized. A number of analysts and economists are now predicting that the recession will bottom out and end in the second half of 2009.

But perhaps the most interesting data point about recovery came from the webinar participants themselves. When asked about their budget expectations for the rest of 2009, 41% of the webinar participants reported that they expected budgets to increase. Compare that to only 18% who expected budgets to be cut. That was mildly surprising to us, given the economic environment we are living through, but also encouraging to see.

No matter what, innovative growth-oriented SMBs know that technology gives leverage and can support and even accelerate growth.  For instance, research by Six Disciplines found that SMB companies strategically using technology outperformed their competitors by 100%.

Strategies and Tips

Among the specific strategies and tips, we discussed:

Setting goals and understanding the challenges you are under. Communicate closely with all the functions in your company, including the Sales Department and Finance, especially during times like these.  You may be in a “no growth” mode but find that it is actually costing money to defer expenditures.  Intel discovered that it could save $19 million (PDF) in its own internal operations, by actually spending money to refresh servers.  Also,  you may be called upon to turn on a dime and gear up quickly to address increased sales growth as the economy opens up again.

Use open source software and existing tools to save money. You can get a software package off the shelf, and customize the software internally to suit your needs.

Using SaaS (software as a service) and cloud computing, avoids having to make large up-front investments for software licenses. It allows SMBs to manage cash flow better.  Also, as many SaaS contracts are month-to-month or short term, you gain flexibility by not being locked in.  However, Mac pointed out that by placing so much reliance on an SaaS vendor, it can have a chilling effect on your ability to switch vendors, in effect giving you less flexibility.

Also discussed were a number of strategic hardware management techniques, including trade-ins, server clustering, virtualization, and the computing performance increases you can gain in your organization.

Chris Peters pointed out the “Server refresh savings estimator.” This is an online tool that enables you to evaluate the benefits of replacing server technology with the latest generation of servers.

One key point was illustrated several times:  by doing a strategic technology refresh, the benefits multiply and spill over beyond just the cost of any hardware itself.   Example:  by consolidating servers, you not only end up with fewer servers that you have to buy, but you also increase computing capacity.  At the same time, it  reduces power consumption, reduces demand on staff time to administer and manage the servers, etc.

Customer Case Studies

Mac and Chris covered 3 customer case studies involved Sun and Intel, showing how SMBs (including one non-profit) had benefited from a strategic approach to technology and had put in place some of the strategies discussed.  The case studies are online:

• Practice-IT, an online training technology provider, utilized VMware and was able to significantly expand its capacity to support increasing workloads without increased cost as the company added new customers.

• NaviSite, a medium-sized hosting company, has 17 data centers around the globe. The company has a large VMware environment and needed to consolidate this environment to save costs while increasing the agility to respond to their customer’s needs.

• Catholic Diocese of Boise, a non-profit providing services and support for 54 parishes, 33 missions and chapels, 14 K–12 schools, a library, and 40 offices, has 38 employees who work at a central administrative office.  They were able to reduce 28 servers down to 4, by implementing Windows Server 2008 Hyper-V to minimize costs and increase server utilization.

There was some livetweeting going on behind the scenes, using the hashtag #MIDMKT. Follow the Twitter discussion  here.

This recap only scratches the surface of what we covered.  To get the full context, all the details, and much more advice, please listen to and watch the full webinar:   “Managing For Growth And Change In A Tough Economy“?

EPA finalizes Energy Star for Servers

The first round of requirements for Energy Star compliant server hardware is out, and the EPA is already in the planning stages for a more rigorous set of requirements set to be released late next year.

The Energy Star program, a joint project of the Environmental Protection Agency and Department of Energy, rolled out its first set of standards for server products. The new standards, which were developed through consultation with major players in the industry, will apply to single and dual-socket servers. Manufacturers that have a product that meets or exceeds these energy efficiency measures will be allowed to place an Energy Star label on it, providing purchasers with a convenient way of picking hardware that won’t bust their power budget.

The relevant documents have been posted on the relevant portion of the Energy Star website. They include a memo that describes the production of the new standards, a spreadsheet for companies interested in certifying a product, and a document that describes both the energy use allowed in a compliant server, as well as plans for an upcoming iteration of that will expand the range of hardware covered.

The hardware that’s not included in the server-room standards is pretty substantial. This time around, the hardware left out includes blade systems, fault-tolerant servers (those with duplicate hardware), server appliances, and networking and storage hardware. Also omitted is any server that has more than four sockets for processors (the number of cores per processor never enters the equation). A lot of that is expected to change with the release of the next standards, termed Tier 2, which are expected to be released late next year. Things listed as under consideration for Tier 2 include blades, servers with more than four sockets, and various server appliances. There’s also consideration of developing a standard based on the energy use per unit of work performed.

All of that is in the indefinite future, however; the new standards are here now. These include specific efficiency numbers for the item that’s often the biggest power draw, the power supply itself. These vary based on the load of the server and the capacity of the power supply, but start at 70 percent energy efficiency, and rise above 90 percent in some circumstances.

For single and dual socket servers, the key item is idle power. For a base system with 4GB of RAM and a single hard drive, a single-socket server should burn 55 watts or less when idling; a dual socket system is allowed to go up to 100 watts. Servers with dedicated management hardware are allowed to go a bit higher. For standard hardware beyond this minimum system, every component increases the power allotment: a hard drive is worth eight watts, each additional power supply is worth 20 watts, each Gigabit ethernet port is worth two watts, and so on.

For four socket servers, there’s nothing like these hard numbers. Instead, the server must have any hardware power management features turned on in the BIOS and enabled by the operating system they ship with. They also need to have the ability to provide feedback on power use to their owners, including power consumption, air temperature, and CPU load.

So far, the standards are pretty limited, but they’ve undoubtedly been useful, both in terms of starting to give Energy Star a presence in the server room, and in terms of starting the industry-government cooperation needed to continue and extend the program.

Come On Techies, Gimme My Cloud Now!

For the biz guys, moving from a server that you had to buy and manage internally to hosting was a great move and a major hassle reduction. The cloud looks like the next logical step. The benefits — no upfront costs, no internal management, no capacity planning — are really compelling for guys who think about the business. But the guys who have to make it work in practice, occasionally say, “Woah, not so fast. You want this to work, right?” In this post, we look at the tension between business logic driving to cloud and the technical hurdles of implementing “apps that matter” in the cloud.

Questions For The Techies

So we decided to put some questions to some smart techies that we know. The role playing here is the CEO needing to make some decisions on where and how to host the service that will shortly get the team fame and fortune.

The first person who rose to this challenge is Tony Bain. Tony Bain is an expat Kiwi, father, entrepreneur, angel investor and blogger who occasionally writes for ReadWriteWeb. He kicked up quite a storm recently with his post “ Is The Relational Database Doomed.”

Apps That Matter

We use “apps that matter” instead of “mission critical” because it sounds less pretentious. These are apps that, when they go down or suffer brownouts bring on a storm of outrage in the blogosphere. Well, at least they do when you have traction, and you are planning to have traction, right?

Apps that matter usually involve write as well as read. They may also involve, at some stage, transactions that include money getting transferred. …
Read More »

Cost of corporate desktop power mismanagement: $2.8 billion

Over the past several years, the companies that run large data centers have become extremely energy conscious, as the price of electricity has forced them to measure and reduce the amount of power they burn through. Server farms are obvious targets for efficiency measures, as the energy use occurs in one place and the hardware is relatively easy to manage. But data centers are far from the only place where organizations host their computing equipment; although individual desktops may be widely distributed, they’re often quite numerous. A new survey attempts to put a price tag on the lack of energy-management practices in the desktop computing segment and comes up with a figure of roughly $2.8 billion wasted in the US alone.

As in many cases where there’s an eye-popping dollar figure involved, the report is using the figure to try to convince its readers to take the matter seriously. In this case, the data was crunched by 1E, a company that sells energy management software for corporate PCs. The company is using it as part of an energy awareness campaign that it is undoubtedly hoping will increase the use of its products. Most of the figures, however, like average price for power in the countries involved, are pretty easy to check, so 1E has little to gain by massaging its numbers. The report is liberally sprinkled with footnotes that describe where its authors obtained various figures as well.

What is more likely to be problematic is the source of employee-behavior data when it comes to desktop computer use. Those numbers come from an online poll performed by Harris Interactive in the US, the UK, and Germany. Harris focuses on this sort of polling, and presumably does it well, but the method will inevitably involve a degree of self-selection and demographic skewing within the population that responds. With that caution out of the way, none of the self-reported behavior derived from the polling looks unreasonable, or even unexpected.

The most basic numbers indicate that a significant majority of employees use a computer at work–just under three-quarters in the US, and nearly 80 percent in the UK and Germany. Of these users, only half shut the computer down when they leave work at night (the rates are slightly higher in Europe). Based on these figures and typical energy use and prices, 1E generated typical numbers for a 10,000-employee company in each of the three countries it surveyed: $260,000 for a US-based company, £168,000 ($244,000) for the UK, and euro-285,000 ($386,000) in Germany. That’s a lot of potential savings left on the table. …
Read More »

Measuring the Benefits of an IT Hardware Refresh

All businesses, whether small or large, depend on strategy. Even not having a strategy is a strategy. Goals are an essential part of those strategies. To achieve goals, and therefore follow the strategy, businesses need and rely on processes. Many people scoff at processes and think of them as nothing more than bothersome “red tape” that gets in the way of doing their jobs. The reality, though, is that clearly defined and communicated processes help create order, reduce chaos and enable consistency, which means results can more easily be measured, checked and audited. Another part of the reality, on the other hand, is that some processes are convoluted, unnecessary and/or inefficient.

So, how does a business objectively determine if a process is necessary or unnecessary? Take the process of refreshing IT hardware as an example. Most companies seem to refresh (or replace) their hardware, including servers, desktops, laptops and workstations, every three to five years. Many IT types and hardware suppliers can reiterate the potential benefits of such a refresh strategy:  improved performance, increased efficiency, reduced support costs, reduced energy consumption, etc.  To truly know if strategies and processes have worthwhile benefits, companies need an objective way of regularly scrutinizing those strategies and the processes used to carry out those strategies. …
Read More »