In my previous post Fall IDF: Is Italian Pasta the Actual Inspiration for Server Virtualization? I talked about the evolution of Server I/O virtualization. I mentioned a few demos and invited you to check them out. But I didn’t give any details about the demos…

Well, it's the 3rd and last day of IDF today and the demos have running now for 2 days. You have one moreday to check them out! So let me describe them quickly.

Dell has been a great partner for these demos. We’re showing 2 demos together in booths 709 and 711 in the Virtualization Community, using Dell’s R710 servers, based on the Xeon 5500 platform, and using Intel 82599 (Niantic) with Virtual Machine Direct Connect (VMDc). The 1^st demo is with VMware and their Network Plug-In Architecture (NPA) technology.

When you visit the demo, check out the great CPU utilization as well VMotion* among heterogeneous server configurations!

The 2^nd demo is delivered with Citrix, showing scalable direct assignment by using XenServer with VT-d and SR-IOV support. The overall performance is really great and live relocation of virtual machines is working nicely.

Another demo (booth 707) is delivered by Red Hat, featuring RHEL 5.4 with KVM (shipping SW with VT-d and SR-IOV support!), Neterion with their 10GbE NIC, all running on an Intel Xeon 5500 server. Look out for the performance value shown for scalable direct assignment!

And finally, a storage (not LAN) demo! Using the same combination of VMM and server, in booth 517, LSI is showing the value of scalable direct assignment for a RAID controller. The performance boost is fantastic!

Check out these demos at the IDF Showcase... I’d love to hear your impressions!

ENERGY STAR compliance is becoming an increasingly important factor for servers and workstations.

Accurate interpretation of the ENERGY STAR specification can be quite challenging with the ever-expanding scope of rules and regulations that govern platform and component-level design considerations.

Intel recently developed a semi-automated test tool to greatly simplify ENERGY STAR testing of servers and workstations. This tool has the following capabilities:

1. Ability to identify the relevant ENERGY STAR criteria per configuration

2. Automated assignment of ENERGY STAR test conditions and system testing

3. Preparation of system test data for analysis and ENERGY STAR submission

Feel free to stop by my booth at IDF in the Eco-Technology Community for a hands-on demonstration and to learn more.

Jennifer

In that case, iSCSI or FCoE (Fibre Channel over Ethernet) could be used for the SAN connection, still using the same high-speed ports. Standards like Data Center Bridging (DCB) could add a lossless character to the 10GbE link to make it friendlier to FCoE.

Few new solutions though come without new challenges. The common way for VMs to share I/O devices in a today’s environment in through mediation of the hypervisor, using emulation or para-virtualization. That reduces the effective I/O bandwidth. It also becomes a fairly significant overhead to the server in its own right, reducing the available server capacity for application processing, and it adds latency. With the growing trend in IT to treat virtualization as a default deployment mode for any application, these issues become quite limiting.

We at Intel have thought that the best way to overcome these issues is by using “direct assignment”. Using the Intel® VT-d technology (launched in the Xeon 5500 platform), a VM can be assigned a dedicated I/O device. This nearly eliminates the overheard related to the hypervisor mediation I mentioned above. A side benefit is that it increases the VM to VM isolation and security. But assigning an individual I/O device to one VM is not very scalable…

This is where the PCI-SIG’s SR-IOV (Single Root I/O Virtualization) standard comes into play. This standard allows a single I/O device to present itself as multiple virtual devices. With SR-IOV, each virtual device can be assigned to a VM, adding scalability to the direct assignment model, effectively allowing the physical I/O to be shared yet with greater security and reliability.

Another challenge with the direct assignment model is related to live migration. Hypervisors have typically assumed the SW mediated IOV model. As a result, hypervisors need to be modified to adapt their live migration solutions to direct assignment.

These technologies span many different components of the server platform. Intel® VT-d is necessary, so Xeon 5500 must be used (or later platforms). SR-IOV capable I/O devices – NICs or Storage controllers, are required. BIOS must be modified, as well as hypervisor software. This is pretty heavy lifting.

So you can only imagine how excited I am to be able to showcase 4 different SR-IOV demos at IDF next week! The demos involve 2 server vendors, 3 VMM vendors – 3 different vendors implementing 3 different hypervisor architectures, and 3 different IHVs representing 2 different I/O technologies. We show the performance improvements, as well as VM live-migration. It works!

Come and see it (Booths 517, 707, 709, and 711 in the IDF showcase)!

I finally finshed the content for the talk that I am scheduled to deliver next week at IDF on Sept 22 (TCIS001 10:15 am – Room 2004).  The content covers examples of optimizing for multi-core using our software tools to accelerate performance,  and more importantly the seamless use of the same software base with minimal or no changes in next-generation architectures (what we call scaling performance forward). Personally, I am excited about the potential of multi-core optimizations with today’s architectures. When I was a graduate student in parallel computing from 1988-1994, it was extremely difficult to take any algorithm and map it to the parallel architectures since most of the algorithms were not very efficient once you took the communication delay’s into account.  The key is to get total delivered performance at an application level, not at the kernel level. However, given the architectures of today which are better balanced, and the availability of multi-core, the memory bandwidth, software tools that work, and the faster interconnects, the number of algorithms that can be parallelized and that actually benefit with accelerated performance (Delivered total application time)  is huge, pretty much every industry vertical is taking advantage of multi-core architectures, software tools, and clusters.

John Gustafson, from our Intel Labs, an industry HPC veteran, is my co-author, and I am thrilled to have him speak about Balanced Computing. Wes Shimanek, a colleague of mine at Intel introduced me to John and after listening to his explanation of balanced computing, and his views on what works and what doesn’t, we immediately  knew that John’s expertise will be greatly valued by the IDF audience, and invited John to be part of the talk. John graciously accepted to participate, and I hope that folks interested in computing architectures, especially in the HPC world will make the time to come listen to John’s talk.

I will also be giving you a high level view on the challenges that drive our products and briefly introduce you to the various aspects of our strategy. I will be followed by 3 more talks that will cover the key aspects of what we do at Intel in HPC, Software Tools for Scaling Application Performance Forward (TCIS002), Delivering more to HPC than just Performance (TCIS003), and Intel® Cluster Ready (TCIS009).

I am looking forward to IDF next week. See you all at the developer forum

Nash Palaniswamy (Intel)

If you would like to learn about a new power supply technology for reducing server energy usage there is an upcoming IDF session that may interest you. The title of the session is “Cold Redundancy – A New Power Supply Technology for Reducing System Energy Usage”. As you can probably guess from the title, we are calling this new technology “Cold Redundancy”. There has been a lot of research done to figure out ways of reducing the input power of a server when it is in an idle state. After all, an idle server is just a very expensive space heater sitting there doing nothing other than consuming energy and producing lots of heat. This is important because some utilization studies have shown that servers can sit idle for a considerable percentage of the time. So anything that reduces the input power of an idle system will have a very significant effect on the overall yearly energy usage - and ultimately save on operating costs.

This presentation will describe, and demonstrate, the cold redundancy technology we have been working on here at Intel® to reduce system idle power.

One great thing about this new technology is that everything can be kept inside the power supply. No changes to the system software will be needed so the only additional requirement to implement this would be using a power supply that has cold redundancy technology inside it. This will make it easy to integrate into systems in the future because it could become a “plug ‘n play” power supply upgrade option.

Since cold redundancy is a power supply technology, I’ll cover some basic concepts to get things started.

There are two different types of power supplies called redundant and non-redundant which are used in computers.

A non-redundant supply has only a single module which provides all the power needed to keep the system operating. This means there is no backup so if the power supply fails, the system shuts down until the supply is replaced. Desktop computers typically have a non-redundant supply in order to keep the costs low.

Most servers on the other hand, have a redundant type of power supply. That means there are extra (redundant) power supplies in the power subsystem so if one supply fails the server will continue working normally. This is for applications where getting maximum system uptime and reliability are worth the additional cost of putting in the redundant supplies. In a case like this though, the redundant supplies are not really needed until one of the supplies actually fails. The drawback with having the redundant supplies turned on until needed is that the supplies still use a lot of power which increases the system operating costs.

Cold redundancy reduces system idle input power by putting these redundant supplies into an almost off (standby) condition or “cold redundancy” mode, as we call it here at Intel®. Because of how cold redundancy works, the more redundant supplies there are in a power subsystem the more effective it is and the more energy that can be saved. The general idea of powering down redundant supplies is not new, but the problem has always been how to turn the supplies back on fast enough so that system operation is not affected in case of a failure. We have come up with a solution to this problem by developing cold redundancy technology. Cold redundancy has the ability to put the redundant supplies into a standby state to save energy at system idle while still being able to turn them back on fast enough in case of a failure to keep the system operating normally. It really is the best of both worlds, saving energy while maintaining the same system uptime and reliability as conventional redundancy where all the power supplies are running all the time.

If you are interested in learning more, the session number is ETMS001 and will be presented at the fall Intel Developer Forum in San Francisco on September 22^nd at 10:15 AM in room 2006 of the Moscone Center. This session will be a combination of lecture and live demonstrations. We decided that having a couple of demos during the lecture would help make the concept more understandable and the presentation more interesting as well. One thing to keep in mind about this session is that we will not be discussing theoretical possibilities or projects planned years in the future but real products that will be available soon. The live demonstrations use a production ready cold redundancy enabled power subsystem that is being integrated into a product Intel® plans to release in the 4^th quarter of this year. It doesn’t get much more real than that.

The demonstrations will show how the control logic works and what power and energy savings are actually possible. This will be done by measuring the AC input power to a four module power subsystem and running the same output load profile with and without cold redundancy enabled. By comparing the two input power graphs the advantages of implementing this new technology can be immediately seen and quantified. I think you will find this to be a very interesting and informative session but then I’m probably biased just a little bit. J

Hope to see you there,

Andy

Presenters:

Viktor Vogman – Power Architect

Andrew Watts – Test Automation

I wanted to share my excitement and some details about an upcoming demo at IDF 2009 (San Francisco, Sept 22-24, 2009) that will demonstrate advanced trading on Nehalem-EX.

Using the power of the 8-Core Intel® Xeon® Processor codenamed Nehalem-EX, NYSE Technologies will demonstrate a complete Smart Order Routing System in a single box that has a total of 32 cores. The demonstration will process the entire OPRA feed and all north American equities feeds, apply rules to decide when to trade and convert the order information into FIX format for delivery to the trading venue. The demonstration will show NYSE Technologies’ Market Data Platform V5™ feed handlers processing raw market data at a rate of over 1.5 million updates per second; NYSE Technologies’ Data Fabric™ messaging platform will pass those messages via its Local Direct Memory transport to a mock Smart Order Routing program which will use Data Fabric again to pass orders to a NYSE Technologies’ Market Access Gateway. Typically, these processing tasks are designed as a three tier model with two latency inducing network hops. Deploying this solution in a single server provides for an order of magnitude reduction in latency.

Nash Palaniswamy (Intel)

Have you ever had one of those MacGuyver moments? You know – you have a problem to solve and a collection of items at your disposal, and if you use can figure out how to use those items, you can save the day.

Earlier this year Intel, HP, IBM, Lawrence Berkeley Labs, Emerson Network Power & Wunderlich-Malec collaborated on a California Energy Commissions sponsored MacGuyver-ish project call Advanced Cooling Environment (ACE) to solve an issue that many data centers face – overcooling their data centers beyond the needs of the servers and other IT equipment running inside of the facility.  You see, IT equipment is designed to run within a temperature envelope. If the air coming into the server is warmer than the envelope, you run the risk of overheating. If the air coming into the server is colder than the envelope, you are spending too much money on cooling the air, which does nothing other than needlessly increase the cost of operating the data center and reduces the energy efficiency as well.

The team surveyed the items at its disposal and determined that they could link data from the front panel temperature sensors (server instrumentation) on the servers to the control systems of the computer room air handlers (CRAHs – essentially air conditioners) via standard data center management communication protocol.   The CRAHs could then dynamically adjust the speed of the fans and the temperature of the air to the requirements of the servers. The results: servers received the appropriate temperature air, power costs for cooling went down and the energy efficiency of the data center went up.  Problem solved….and they didn’t even use a paper clip or shoestring.  The real beauty of the project is that all of the items used are commercially available today for you to instrument your data center and improve the energy efficiency of your operation.

To learn more about ACE at the upcoming Intel Developer Conference in San Francisco, check out the “ECOS003 Advanced Cooling Environment (ACE) Technology: Controlling Data Center Cooling with Servers” or stop by the ACE demo in the Eco-Tech Zone of the Tech Showcase.

In spite of significant gains in server energy efficiency, power consumption in data centers is still trending up.  At the very least, we can make sure that the energy expended yields maximum benefit to the business.  A first step in managing power in the servers in a data center is having a fairly accurate monitoring capability for power consumption.  The second step is to have a number of levers that allow using the monitoring data to carry out an effective power management policy.

While we may not be able to stem the overall growth of power consumption in the data center, there are a number of measures we can take immediately:

· Implement a peak shaving capability.  The data center power infrastructure needs to be sized to meet the demands of peak power.  Reducing peaks effectively increase the utilization of the existing power infrastructure.

· Be smart about shifting power consumption peaks. All the watts are not created equal.  The incremental cost of generating an extra watt of power during peak consumption hours is much higher than the same watt generated in the wee hours of the morning.  For most consumer and the smaller commercial accounts flat rate pricing still prevails.  Real time pricing (RTP) and negotiated SLAs will become more common to put the appropriate economic incentives in place.  The incentive of real time pricing is a lower energy bill overall, although the outcome is not guaranteed.  In pilot programs residential consumers have complained that RTP result in higher electricity costs.  With negotiated SLAs the customer can designate a workload to be subject to lower reliability; for instance, instead of 3 9’s, or outages amounting to about 10 hours per year, the low reliability workload can be designated as only 90 percent reliable, and can be out on the average of two hours per day.

· Match the electric power infrastructure in the data center to server workloads to minimize over-provisioning.  This approach assumes the existence of an accurate power consumption monitoring capability.

· Upgrading the electrical power infrastructure to accommodate additional servers is not an option in most data centers today.  Landing additional servers at a facility that's working at the limit of thermal capacity leads to the formation of hot spots, this assuming that electrical capacity limits are not reached first with no room left in certain branch circuits.  Hence measures that work under the existing power infrastructure are to be preferred over alternatives that require additional infrastructure.

For the purposes data center strategic planning it may make economic sense to grow large data centers in a modular fashion.  If the organization manages a number of data centers, consider making effective use of the existing data centers, and when new construction is justified, redistribute the workloads to the new data center to maximize the use of the new electrical supply infrastructure.

Intel has built into its server processor lineup a number of technology ingredients that allow data center operators optimize the utilization of the available power system infrastructure in the data center.

Newer servers of the Nehalem generation are much more energy efficient, if only because of the side effect of increased performance per watt.  These servers also have a more aggressive implementation of power proportional computing.  Typical idle consumption figures are in the order of 50 percent of peak power consumption.

Beyond passive mechanisms that do not require explicit operator intervention, the Intel® Intelligent Power Node Manager (Node Manager) technology allows adjusting the power draw of a server and trade off power consumption against performance.  This capability is also known as power capping.  The control range is a function of server loading.  For the Intel SR5520UR baseboard on the 2U chassis, the server will draw about 300 watts at full load and its power consumption can be rolled down to about 200 watts.  The control range tapers down gradually until it reaches zero at idle.

For power monitoring, selected models of the current Nehalem generation come with PMBus specification compliant power supplies allowing real-time power consumption readouts.

The Node Manager power monitoring and capping capability apply to a single server.  To make this capability really useful it is necessary to exercise these capabilities collectively to groups of servers, to add the notion of events and a capability to build a historical record of power consumption for the servers in a group.  The additional capabilities have been implemented in software through the Data Center Manager Software Development Kit developed by the Intel Solutions and Software Group.  An additional Software Development Kit, Cache River allows programming access to components in servers and server building blocks produced by the Intel Enterprise Products Server Division (EPSD), including the baseboard management controller (BMC) and the management engine (ME), the subsystems that host or interact with the Node Management firmware.  EPSD products are incorporated in many OEM and system integrator offerings.

Data Center Manager implements abstractions that apply to collections of servers:

·  A hierarchical notion of logical server groups

·  Power management policies bound to specific server groups

·  Event management and a publish/subscribe facility for acting upon and managing power and thermal events.

·  A database for logging a historical record for power consumption on the collection of managed nodes.

The abstractions implemented by DCM on top of Node Manager allow the implementation of power management use cases that involve up to thousands of servers.

If this topic is of interest to you, please join us at the Intel Development Forum in San Francisco at the Moscone Center on September 22-24.  I will be facilitating course PDCS003, "Cloud Power Management with the Intel(r) Xeon(r) 5500 Series Platform."  You will be the opportunity to talk with some of our fellow travelers in the process of developing power management solutions using Intel technology ingredients and get a feel of their early experience.  Also please make a note to visit booths #515, #710 and #712 to see demonstrations of early end-to-end solutions these folks have put together.