Aspen Systems Direct Blog

Data Capacity Demand Driving New Storage Servers

robd — Thu, 23 May 2013 16:50:46 +0000

Traditional storage, tiered storage, ZFS storage solutions and JBOD sales are growing in record numbers. Aspen Systems continues to get high volumes of requests for higher-data capacity storage solutions. The Storage servers we are building today are a reflection of the times; with the utilization of larger, 3 and 4TB traditional spinner drives. But, it is not only the demand for greater data capacity storage that is growing at a staggering rate. The development and manufacturing of improved SSD drives can barely keep up with the latest storage models requiring higher iOPS and longer durability.
>
High-endurance, enterprise-class SSD drives are no longer an added luxury, but quickly becoming a hard and fast rule for building highly scalable, long-lasting storage solutions. In the past couple of months a number of new eMLC (extended multi-level cell) SSD drives have been announced, launched and released from Intel, Micron, OCZ, Kingston and other prominent manufacturers. IT storage administrators have made it a priority to use only these drives for caching, ZIL and OS boot drives. The reason is obvious: eMLC drives like the Intel DC S3700 series, for example, can write to itself up to 10 times every day for up to 5 years. Other drive manufacturers durability ratings vary, but the one thing they all have in common is a much higher price tag. The need for longer SSD life has also driven demand for higher SSD capacity in general. Even in cases when 60 or 100GB SSD drives might offer enough capacity, larger drives bring value to the overall life of any given storage solution.

The recent demand for high-durability technology, has unfortunately caused some delays in obtaining these products. More than likely, this has only delayed any reductions in cost. But, the industry moves on. The overall market for Hard Disks and SSD drives reportedly grew by 73.4% in the first quarter of 2013. We can only imagine how quicker things will grow when these newly announced products become more readily available.

For storage, the time is now and through the close partnership of Supermicro, Aspen Systems can put together storage solutions for any and every type of need and budget. Let us help.

Aspen Systems Direct
888-227-1286
sales@aspensystemsdirect.com

Intel 4th Generation Core Processors Won’t Change Server Market Anytime Soon

robd — Tue, 21 May 2013 21:08:48 +0000

Since Aspen Systems gets a fair number of questions on new processors before they are released, it makes sense to clarify what we know about Intel’s pending launch of the new, 4th-Generation Haswell, 22-nanometer processors. The first and most important thing to understand is that the Intel Xeon E5-2600 is still the number one performance choice for servers and that is not likely to change any time soon. Intel is expected to launch the new, Haswell (4th-Gen) processors at the 2013 Comdex show in Taiwan on June 4th. Versatility, might be the word that best described Intel’s new, 22-nanometer CPU architecture.

Understandably, Intel is anxious to follow up with a strong new product to meet the trending demand for ARM and mobile computing as well as more powerful, energy-efficient desktops. The Haswell follows up on the Ivy Bridge architecture with increased awareness in the growing mobile, smartphones and tablet markets. With Haswell, they hope to capture a larger share of the mobile computing market space. Currently, Intel is not specifically targeting servers with the successor to Ivy Bridge. Rather, they are expected to address three product areas with the 4th-gen processors: Desktops, Mobile/Laptops and BGA.

While a little late to the game, Intel aims to take-over the low-powered, ARM marketplace in the next year and Haswell will play a mighty role in achieving that goal. Intel also pins its hopes on a revitalized, re-invented desktop market due to the touch-screen interface of Windows-8, all-in-one, PC popularity. Improved power and performance for new, consumer-oriented form factors has greatly driven the 4th-gen Core CPU.

Intel boasts vastly improved graphics performance with up to 18% speed improvement over the previous Ivy Bridge, Hitman. Reports suggest this should beat NVIDIA and AMDs edge in graphics in the mainstream graphics market.

For IT Admins and engineers interested in highly intensive computing tasks, there is currently no compelling reason to wait and see about Haswell. An advanced new Xeon server architecture sprung from Haswell will come soon enough. In the server and enterprise storage markets, the future is now.

Supermicro Releases Data Center Optimized Series of High Performance Servers

robd — Thu, 16 May 2013 22:35:43 +0000

Today’s data center is about more than just performance alone. Servers and storage solutions in our thriving, growing data centers must meet the ever-increasing demand for improved space utilization, cooling, density and scalability. When it comes to adding necessary technology hardware, IT departments face more challenges besides monetary budgets. Additional servers and storage must be able to fit into the data center’s space, energy and cooling budgets as well. While it is true that rackmount servers are getting faster, denser and more energy efficient, they are also expected to do more than ever before. For IT administrators in data centers and cloud computing environments, acquiring the proper equipment has become more strategic and challenging than every before. Supermicro has released a series of new servers which are specifically optimized to run cooler in high-temperature environments, as well as save space and energy costs.

In order to accomplish this task, Supermicro has built features, and cloud ready solutions into servers which are specifically optimized for data center tasks. These highly flexible servers are built on the Intel’s E5-2600 processor family and range in size from 1U to 4U Rack and workstation form factors. Built to operate at temperature ranging from 0-C to 47-C, these systems are highly suited for free-air cooled data center environments. What specifically makes Supermicro’s line-up of Data Center Optimized servers suited for these environments?

The Supermicro Data Center server line in all of the various form factors are built with the following, common building blocks: WIO + 768GB Memory support in 24 Dimm Sockets; 4 x GbE Lan Ports and/or 10GbE. Data Centers need expansion capability. Standard Features on Rackmount Data Center servers include up to 6 x PCI-E 3.0 Slots, redundant power supplies, Hardware RAID, IPMI 2.0 dedicated LAN with Virtual Media/KVM over LAN.

For extreme, Cloud environments, Supermicro offers the 5037MR-H8TRF, a 3U MicroCloud computing with 8x Single-Processor Hot Nodes. For extreme processing power, 4-Way, cluster-ready compute nodes are available in 1U and 2U Form Factors. These powerful, Supermicro servers are designed and optimized for data center/cloud environments as well as Virtualization, Simulation, Automation, Enterprise Database applications and more.

Would you like help selecting the most effective hardware for your Data Center or Cloud environment. Please give us a call.

Aspen Systems Direct
888-227-1286
sales@aspensystemsdirect.com

World’s Most Versatile GPU Workstation

robd — Thu, 16 May 2013 17:10:31 +0000

The Supermicro 7047GR-TRF continues to be our best selling GPU workstation to date. The capabilities and features of this highly expandable unit, make it extremely flexible and suitable for a wide variety of applications and hardware configurations. Today’s GPUs and MIC accelerators are available in a wide variety of form factors, meaning not all chassis and workstation motherboards will support them. Additionally, not all server-class units area built with the power and space requirements to handle the increasing demand for multiple GPU configurations. Today’s Tesla’s, GTX Cards, Quadro, Titans and Xeon Phi’s open up a world of configuration possibilities for research and application engineers wishing to maximize the performance of their computational demands. The 7047GR-TRF makes it possible to explore a virtually unlimited number of high-performance computing solutions using the latest GPU/MIC Accelerators.

The 7047GR-TRF uses the Supermicro X9DRG-QF motherboard which supports the fastest, dual, Intel Xeon E5-2600 processors currently available. The X9DRG-QF has 16x Dimm sockets can be populated with up to 512GB of DDR3-1600MHz ECC Registered memory. Powered by the CSE-747TQ-R1620B chassis with redundant 1620 Watt Power Supplies, the 7047GR-TRF can be turned into an impressive workstation with support for up to 4 x GPU/MIC accelerator cards. The chassis will handle up to 8 x 3.5″ Hot Swap SAS/SATA Hard Drives. There is onboard SATA 3.0 6Gbps support for RAID’s 0 and 1 and SATA 2.0 3Gbps support for RAID levels 0, 1, 5, 10. If you would rather add a hardware RAID controller there are still 2 x PCI-E x8 slots available even after fully populating the 4 x x16 PCI-3 3.0 slots with GPU Cards.

&nbsp
There is one thing this unit does better than any of them. Not all GPU Servers are capable of handling the heftier sized GTX-690 and Titan cards from NVIDIA – the 7047GR-TRF will accommodate them quite nicely. At Aspen Systems, we have fully loaded the 7047GR-TRF up with GTX-690 cards and 8 x 3.5″ Hard Drives and have never had a problem with the power requirements. Considering the computational capabilities of the latest Tesla, GTX and Xeon Phi products, the 7047GR-TRF offers the greatest range of possibilities. If you’d like some help with a GPU server configuration, please let us know.

Aspen Systems
sales@aspensystemsdirect.com
888-227-1286

Benchmarking the Xeon Phi

robd — Tue, 14 May 2013 21:31:33 +0000

While the Xeon Phi continues to gain momentum from a growing base of enthusiastic users and developers, standardized measurements of performance and stability are a critical component in helping the art of parallel processing meet its fullest potential. Popular, generic computing benchmark software lacks the specific, relevant tests for evaluating benchmarks of Xeon Phi’s unique parallel co-processing hardware. Intel has addressed this by porting over the Scalable Heterogeneous Computing Benchmark Suite (SHOC) for full support of the Intel Xeon Phi parallel co-processor. The benchmark uses offloaded programming sets from the compiler included in Intel’s Composer XE 2013 Software package. The HOC Benchmark software tests the Xeon Phi through three levels:

Level 0 performs three different tests: Bus speed download and bus speed readback bemchmarks for measuring data transfer speed between the system and the Xeon Phi co-processor on data of various sizes; GDDR3 Memory read/write speeds from the co-processor core and MaxFlops which measures floating point rates of double and single precision arithmetic. Intel says to make note, however, that this particular test does incur some errors due to the specific compiler optimization. Still, the MADD8-SP and MADD8-DP numbers should give you a good idea of the overall computation speed.

Level 1 Benchmarking includes four tests: Gemm performs general matrix measurements of single and double precision numbers based on the Xeon Phi’s Math Kernal library. MD benchmarks the Lennard-Jones potential computations from Molecular Dynamics. The Reduction test benchmarks sum reduction operation of floating point calculations and Scan is a benchmark of parallel prefix sums for floating point numbers.

The final, Level 2 benchmark is for measuring the performance of real application kernals. This is accomplished through the benchmarking of S3D combustion simulation application which provides the computational rate of chemical reactions within a 3D grid.

Benchmark numbers are only useful with the full support and adaptation of the user community. You can download this SHOC benchmark tool here: Intel Xeon Phi Benchmark Download The GitHub repository also includes user files and various other information and instructions. This is the way to gain better insight on the capabilities of the Xeon Phi and share.

When 4 GPUs are not Enough

robd — Tue, 14 May 2013 17:32:16 +0000

There is no such thing as ever reaching the point of diminishing returns when it comes to adding extra computational capabilities to power-intensive applications. If and when 4 GPU’s are not enough, Tyan offers the FT77A-B7059 GPU server which supports up to 8 x x16, full-length, PCI-E Gen 3.0 GPU’s.

The Tyan FT77AB7059 is a 4U Rackmountable Barebones GPU Server supporting dual, E5-2600 Processors and up to 8 of the very latest GPU’s including the NVIDIA Kepler K10/K20/K20X and Intel’s SC5110P Parallel Co-Processor. Tyan has put a tremendous amount of innovation in getting such unprecedented computational power into a single 4U platform. As we’ve seen from ASUS, Supermicro and other GPU server brands, fully loaded GPU servers are requiring every last ounce of their 1620 – 1800 watts of power for uninterrupted operational performance. That’s with only 4 x GPU’s. What kind of power is required to run 8?

Tyan has outfitted the FT77AB7059 GPU Server with 2 + 1 2400 Watt Redundant Power Supplies. Tyan confidently claims that this is more than enough power to run anything you can throw into this system. The chassis is deep and roomy with (5+1) Hot Swap, 12cm fans and plenty of air flow to keep a fully loaded unit running cool. The chassis also supports either 6 x 2.5″ Hard Drives or 4 x 3.5″ Hard Drives.

The FT77AB7059 motherboard supports the latest, E5-2600 Processors up to 150W. 24 Dimm Sockets will handle up to 768GB of System Memory or 384GB Per CPU. The Tyan Onboard RAID has support for 2 x 6Gb/s and 4 x 3.0 Gb/s RAID levels 0/1/10/5. Of course, there are extra slots for hardware RAID. The system includes 8 x 3.0 slots for GPU’s and 2 additional 8 x slots for your choice of additional RAID or Networking Cards. Onboard are 2 x 10GbE Intel X540 Networking cards (1 port is shared for the IPMI control). Tyan also offers recommended QDR-2 and IB Mezzanine cards operating at 40Gb/s and 56Gb/s.

The features and specifications of Tyan’s 8 x GPU server only begin to explain the computational possibilities.
What do you think? When you’re ready to put your thoughts and imagination into action, we’d like to help.

Aspen Systems
888-227-1286
sales@aspensystemsdirect.com

ASUS Delivers Quad-Socket Dual-GPU Hybrid Computing

robd — Tue, 14 May 2013 16:24:59 +0000

ASUS has a reputation for building high quality, rock-solid computers that don’t compromise performance, features and innovation. There is probably no better example of this than the RS926-E7/RS8 Quad-Socket, Dual GPU server. The concept of hybrid computing is about more than just super computing power. By combining 4 x Romley-EP 4S Series processors with Dual GPU’s, ASUS is able to consolidate unprecedented super computing power within the confinements of a 2U space. The RS926-E7/RS8 is a mere 29.5″D x 17.5″W x 3.5″H.

IT Admins, End-users, re-sellers and scientists familiar with the ASUS brand, have a high level of trust in its reliability. Such an ambition server requires careful thought and innovation. ASUS has obviously put a great deal of thoughtful engineering and innovation behind the RS926-E7/RS8 design. With such great amounts of processing and GPU computing power inside a single 2U box, energy and cooling might cause concern for some who are not familiar with the brand. A quick overview of the RS926-E7/RS8 should answer any skepticism. The unit is powered by Dual 1620 Watt, platinum-level power supplies and 5 x 80mm x 38mm fans to keep things cool. If capable Processing and Graphics Computing power are important, only advanced connectivity and features can fully harness the usefulness of the machine. ASUS has managed to over-deliver in the area of functionality and features.

With 32 Dimm Sockets, each CPU can can be populated with 256GB of 1600MHz Registered ECC Memory for a total of 1TB System RAM. The RS926 comes with 4 x 1GbE Ethernet Ports + 1 IPMI KVM Management Port and a Mellanox ConnectX-3 FDR. There are 2 x SATA 3 6Gb/s Ports and 4 x SATA 3Gb/s ports. If you need advanced, hardware RAID, ASUS offers their optional Pike 2008/2108 SAS2 6GbE RAID controller at a reasonable, additional cost. The well-made ASUS chassis comes with 8 x 3.5″ Hot-Swap Hard Drive Bays for storage. ASUS has certified Operating Support for CentOS 5.6, Windows Server 2008 R2/Enterprise/32/64 Bit, RedHat Enterprise Linux AS5.6/6.0 32/64 and VMWare ESX4.1/ESXi4.1. Other operating systems may work, but this is what has currently been tested.

If you have questions about the RS926-E7/RS8 or would like us to help you with a configured quote, please don’t hesitate to contact us today.

Aspen Systems
888-227-1286