Your SlideShare is downloading. ×
A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures
A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures
A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures
A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures
A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures
A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures
A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures
A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures

587

Published on

Learn about Flex System, offering superior manageability, faster communications, more storage capacity, better storage management, and broader/better physical/virtual system management than all of …

Learn about Flex System, offering superior manageability, faster communications, more storage capacity, better storage management, and broader/better physical/virtual system management than all of today’s leading blade competitors. For more inofrmation on Pure Systems, visit http://ibm.co/J7Zb1v.

Visit http://bit.ly/KWh5Dx to 'Follow' the official Twitter handle of IBM India Smarter Computing.

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
587
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
18
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Research Report A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures Executive Summary In this Research Report, Clabby Analytics takes a closer look at IBM’s Flex System converged architecture. We compare the new Flex System advanced blade offering to its closest rival: traditional blade architecture. And what we find is that Flex System offers superior manageability, faster communications, more storage capacity (using up to eight internal solid state drives [SSDs] per compute node) and better storage management — as well as broader/better physical/virtual system management — than all of today’s leading blade competitors. Introduction For over a decade Clabby Analytics has been a big fan of blade architecture. We like blade designs (the ability of blades to share a chassis and common components such as power supplies and fans — reducing energy and real estate requirements). We like blade flexibility (the ability to run multiple operating environments such as Windows, Linux, and Unix); and the ability to run different processors (for instance, x86, POWER, and field programmable gate arrays). We like blade extensions in reliability/availability/serviceability (RAS). And we like the way that some blades handle virtual I/O (this simplifies the assignment of network addresses). However, today’s blade architectures have some hardware and software limitations as compared with new “converged systems” designs:  Siloed system management — most blade environments offer monitor/management tools focused on managing the physical blade environment. But, when it comes to managing virtual (logical) machines as well as network/storage subsystems, many blade vendors rely on 3rd party management/infrastructure products. To manage blade environments most efficiently, systems administrators and managers need well-integrated tools that can manage both physical and virtual resources across the entire systems environment (compute nodes, storage, and networks) in a cohesive, integrated manner.  Greater Latency— Cisco blades and Hewlett-Packard blades that use Cisco’s FEX architecture flow traffic north (up) to a Layer 3 network Top of Rack (TOR) switch (that provides security, intrusion detection, and other services), then south (down or out) to communicate with other compute nodes or storage devices. This slows the system down. A more efficient approach is to offer a networking architecture with the ability to route traffic between nodes (in an east-west) as needed.  Chassis design — most of today’s blade chassis rely on a hardwired mesh of copper connectors that will make it difficult (if not impossible) to support speeds beyond 40 Gb without a chassis/midplane redesign. So, a major redesign of the market’s leading blade architectures will need to take place to handle higher communications speeds (which means that some blade environments may not yield the expected return-on-investment [ROI] because the blade chassis may need replacing sooner than planned). Additionally, today’s converged systems chassis have advantages in airflow, power consumption, and cooling — plus room to support advanced communications adapters such as fourteen data rate (FDR) InfiniBand (some competing blades cannot support this technology).
  • 2. A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures September, 2012 © 2012 Clabby Analytics Page 2  Internal storage capacity and proximity — with a few exceptions, blades offer little internal storage capacity (usually limited to two blade-attached disk drives). (Exceptions include IBM’s BladeCenter S chassis and HP’s c3000 with HP storage blades). Further, storage attached to the leading commercial blades is not sharable; it lacks reliability features (such as flash copy and thin provisioning); and it lacks advanced functions such as external virtualization, tiering, real-time compression, and clustering. Finally, because internal storage is limited, most blades must exit the chassis to access the data that they need. Placing storage outside of the blade enclosure means that several hops may need to take place to access/retrieve data — creating potential bottlenecks and latency issues.  Memory limitations — Hewlett-Packard’s workhorse BladeSystem (the BladeSystem BL460c G8) tops out at 576Gb; and Dell’s blade environment can address 640Gb (but at 3 memory DIMMS per channel, using all 640 Gb of memory may result in unbalanced performance). These architectures do not take advantage of the maximum capability that Intel Xeon E5 processors can support and may, accordingly be leaving the ability to create more virtual machines (VMs) on the table. The amount of memory available affects the number of virtual machines that can be launched within a given blade environment. The more virtual machines that a system can launch, the higher the systems utilization rate and, therefore, the better the return-on-investment. Background: IBM’s New PureSystems Family and Associated Flex System Architecture In April, 2012, IBM announced a new family of systems known as “PureSystems”. The underlying architecture of these PureSystems is known as Flex System (a group of compute nodes, storage nodes, network switches, adapters, and chassis that, when blended together, form the entry point for IBM’s PureSystem family). The difference between PureSystems and Flex System is that IBM preconfigures PureSystems into what they call “expert integrated systems” — configurations that use IBM’s deep systems design and tuning expertise to fine-tune PureSystems environments. With Flex System, IBM enables customers who may wish to utilize some non-IBM components to design and tune their own systems by mixing and matching with Flex System components. Both blade servers and Flex System architecture allow buyers to mix and match compute and storage blades — and each environment offers a variety of fabrics and interconnects. But, unlike competing blades, Flex Systems offer:  IBM’s cross-system management appliance known as the Flex System Manager (FSM). Flex System Manager uses a “single-pane-of-glass” management interface to manage across compute storage and networking resources. The ability to manage across an entire systems environment results in greater manager/administrator productivity — helping to significantly lower operational costs related to systems/storage/network management.  Flex Systems Manager also offers a rich collection of advanced, integrated, cross-system virtual and physical systems management services. IBM’s VMControl module, for instance, can manage open source KVM, Microsoft Hyper-V, EMC VMware and PowerVM environments, all from the same management interface. Further, IBM’s SAN Volume Controller can be launched from within Flex System Manager, and can be used to virtualize and manage existing storage, regardless of vendor.  Communication within Flex System architecture can be north/south as well as east/west. By being able to pass traffic through an internal Layer 2/3 switch, latency can be reduced by up to 50%, resulting in significantly improved performance for inter-system communications.
  • 3. A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures September, 2012 © 2012 Clabby Analytics Page 3  Flex System architecture offers access to large amounts of enterprise-class internal storage (where blades typically do not). In its Flex System, IBM offers access to up to eight SSDs (solid state disks) located within each compute node — giving each compute node access to up to 1.6TB of local data. These SSDs can be used like extended, fast memory — and are also positioned to provide “hot data” rapidly to compute elements. This is a unique IBM innovation that IBM has dubbed Flex System Flash.  IBM’s Storwise V7000 storage array can be mounted outside a Flex System chassis today — attached by a Fibre Channel switch (which is similar to how most blades attach to external storage today). But IBM has also issued a statement of direction regarding its Flex System storage nodes, stating that it will deliver a Storwize V7000-like node that can be directly mounted within a Flex System chassis and attached to the system midplane. The first benefit of putting this large storage array within a Flex System rack is that compute nodes will need to take fewer hops to get to data. The second benefit is that the Storwize V7000 will be easier to manage (described in the next bullet point).  The Flex System chassis/rack midplane connects all system components to each other. Using the Chassis Management Module (CMM), information technology managers and administrators can control power utilization; perform fan management, initialize the chassis and compute nodes, manage switches, perform diagnostics on the chassis and input/output (I/O) devices, work with security policies, perform resource discover and inventory management, issue alerts and perform monitoring/management functions. IBM’s compute nodes and network switches currently attach to the midplane — and IBM is expected to attach a large storage array (the Storwize V7000-like node described previously) in the near future (as stated previously, this is a statement of direction). When this happens, all major systems components (compute nodes, network switches, and storage nodes) as well as ancillary components (such as fans and power supplies) will all be connected in a unified, converged system architecture that can be managed from a common management point.  Finally, Flex System architecture offers more memory than the leading blade competitors. As stated earlier, HP and Dell Blades lag Flex Systems in terms of total addressable memory. IBM’s new x440 compute node supports up to 1.5TB of memory. What this means is that Flex Systems managers can configure more virtual machines — delivering higher utilization and superior ROI to the enterprise. Market Positioning: Blades, PureSystems and Flex System IBM offers traditional blades (found in its BladeCenter portfolio). IBM also offers a family of converged servers known as PureSystems (these are based on IBM’s Flex System component architecture). At present, PureSystems are offered in two configurations: 1. IBM PureApplication systems (integrated and tuned with the IBM middleware stack that exploit “patterns of expertise” to exponentially improve performance while greatly simplifying deployment). This design fits into a class of servers known as “Platform-as-a- Service” (PaaS) servers. For more information on this environment, visit this Website; and, 2. IBM PureFlex systems (integrated infrastructure environments with built-in expertise, to help clients reduce deployment time and costs). This design fits into a class of servers known as IaaS (“Infrastructure-as-a-Service”). For more information on PureFlex, see our report here). In the PureApplication and PureFlex configurations, IBM adds system integration value using “patterns of expertise” (deployment and tuning patterns developed by IBM technical experts) to accelerate deployment and significantly improve system performance. IBM’s Flex System is
  • 4. A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures September, 2012 © 2012 Clabby Analytics Page 4 positioned to offer IT buyers the ability to build their own systems environment and populate that environment with the infrastructure of their choice, with their own custom applications, and/or with 3rd party packaged applications. All of these systems use the same underlying systems architecture (chassis, compute nodes, storage nodes, switches) — but PureApplication and PureFlex systems are tuned by IBM, while Flex System is tuned by the customer. What Is So Special About the Flex System Design as Compared with Traditional Blade Servers? Traditional blades are made up of compute blades and network components — and some offer on- blade PCIe mezzanine card extensions. These components are inserted into a blade chassis. By contrast, Flex Systems are comprised of components that make up a whole system environment: compute nodes, storage nodes, network switches, and PCIe expansion nodes. Further, Flex Systems also support an optional management appliance in a standard compute node form factor known as the Flex System Manager (pictured in Figure 1 as the “management node”). Figure 1: IBM’s Flex System Elements Source: IBM Corporation — April, 2012 When directly comparing blade architecture to Flex System, several major differences can be found. These differences can be found in:  Features/functions of the compute nodes;  The availability of an advanced management node that manages physical and virtual systems as well as storage nodes and the network fabric;  An advanced networking design that features high-performance and networking flexibility  The chassis/rack design;  The amount of memory that can be addressed by each compute node;  The amount of solid state storage available to compute nodes;  The capacity of storage nodes; and,  The physical location of storage nodes. Each of these points deserves a closer look.
  • 5. A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures September, 2012 © 2012 Clabby Analytics Page 5 The Compute Nodes Flex System compute nodes are available in standard- and full-width form factors. These nodes use Intel Xeon (x86-based) processors — the most recent of which is IBM’s new Flex System x440 Compute Node. This compute node includes 4-socket Xeon E5 processors, access to 48 LP DDR3 DIMMs (memory modules) offering access to up to 1.5TB of memory, a 10Gb LAN-on- Motherboard (LOM) — and 2 hot-swap 2.5” SAS/SATA SSDs or hard drives. This node also offers a dual enabled hypervisor as well as a VMware ESXi on flash key option. Flex System Manager and the Chassis Management Module IBM’s Flex System Manager (FSM) and CMM highlight the strongest difference between Flex System and traditional blade architectures. The focal point of IBM’s Flex System architecture and associated chassis/rack design is to build an integrated, “converged systems” environment where all resources (systems/storage/networks) can work together in an integrated, orchestrated fashion. According to IBM, FSM was designed from the very beginning to be the common-pane-of-glass centralized management environment for systems managers. In the FSM world, systems managers have control over all of the resources needed to ensure that applications can run optimally. This is to be contrasted with the blade approach where a systems manager may use one management product set to manage physical servers, then turn to VMware or Microsoft or to open source KVM to manage virtualized servers. In the blade world, these same systems manager may turn-over responsibility for storage management to another group of specialists — and the same may hold true for networking. IBM’s FSM is designed to be a cohesive, cross-system management environment, whereas blade management software tends to be compartmentalized. From a systems management perspective FSM provides simplified system setup by self-discovering the resources within a Flex System environment. Tools are available that help monitor resources — providing health summaries, issuing alerts, enabling thresholds to be set, making updates easy to accomplish, and streamlining service and support. The FSM environment is highly-visual, enabling managers and administrators to intuitively understand the relationships between physical and virtual devices. Using FSM tools, administrators and managers can view a topology map of a systems environment, drill down for further details, and easily perform troubleshooting. Troubleshooting responses can be automated — saving time should problems recur. And Flex System Manager allows policies to be established and automated (such as policies to track and automate firmware updates and software compliance). All of these functions are integrated as part of Flex System Manager. From a storage perspective, FSM can manage the pooling and virtualization of external storage. And FSM enables disks to easily be defined and attached to virtual servers. Storage can be dynamically provisioned as part of any image deployment (meaning that the amount of storage needed by a virtual machine can be associated with that virtual server or virtual appliance image). Policies can be established to place storage where it belongs within the storage hierarchy (known as storage tiering). Storage can follow virtual servers using a dynamic zoning/masking approach to virtualized storage management. And FSM storage management can be linked with other advanced storage management facilities available in IBM’s Tivoli product set. FSM network management allows the network switches of various leading switch providers to be managed from one common interface. FSM’s network management facilities allow network components to be discovered and inventoried — and the status of these components can be constantly monitored. IBM offers support for KVM, pHyp, and VMware virtual switches as well as for physical switches as well. From a network management perspective, administrators and managers can see not only the server activities, but also the related network activities. So administrators and managers can see virtual machine activities per virtual machine — and gather performance statistics to gauge network performance and to locate network trouble spots. Further,
  • 6. A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures September, 2012 © 2012 Clabby Analytics Page 6 because FSM provides logical views of servers and network resources grouped by subnet and VLAN (virtual LAN), administrators and managers can easily visualize network behavior and make performance adjustments accordingly. Finally, the FSM provides the ability to automatically provision and move VLANs — enabling VLANs to be configured to support the networking demands of virtual machines. Virtualization management is impressive within the Flex System Manager, as is “built-from-the- ground-up” security. FSM provides virtualization facilities to manage virtual servers and hosts, virtual server lifecycle management (so unused virtual servers are returned to the resource pool), topology maps, and more. Editing and relocating virtual resources is straightforward to accomplish. Virtual images can easily be created, deployed, imported, or captured. And numerous advanced virtualization features are offered as part of Flex System management (see Figure 2). Figure 2 — IBM Flex System Manager — Advanced, Integrated Virtualization * IBM has released a statement of direction for the energy automation features displayed here. Source: IBM Corporation — September, 2012 Flex System Manager also features centralized user security management controls using secure communications protocols, centralized user management, and simplified security policies (includ- ing low, medium, high security profiles) that can be centrally managed from the FSM or CMM. The CMM module offers administrators and managers a secure chassis environment that provides secure boot facilities, secure protocols, user-level management, signed firmware — and IBM will soon support TCG Dynamic Root of Trust Measurement (DRTM). Further, the CMM provides support for power management, compute node initialization, switch management, and it can perform diagnostics on the chassis and input/output (I/O) devices. Early customer feedback indicates that IT executives really like the combination of FSM and CMM. With these offerings, managers and administrators can understand what is going on in their information systems environments across entire systems as opposed to understanding what is going on in system/storage/network silos. With FSM and CMM tools, managers and administrators gain a better understanding of how their information systems are being utilized; they can more easily find trouble spots;
  • 7. A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures September, 2012 © 2012 Clabby Analytics Page 7 and they can produce hard-evidence that their systems environments are running optimally (this is useful when it becomes necessary to go to the chief financial officer’s office to ask for more funding to support new initiatives or to serve growing capacity). Managers and administrators also like the ability to model application behaviors using FSM — a feature that helps in capacity planning. Switches and the Communications Subsystem One of the biggest differentiators between traditional blade architecture and Flex System architecture can be found in the switching facilities. Blades employ a top-of-rack switching approach that sends all traffic through a layer 3 switch (various management and security functions happen at layer 3). This creates processing overhead — and this overhead creates latency. Because Flex System can avoid layer 3 switching (if so desired) by using a Layer 2/3 internal switch, the amount of latency in program-to-program communications, or disk read/writes within a Flex System chassis can be cut in half. For IT buyers who need high performance, this is a major advantage. The Chassis/Rack Design IBM bills its Flex System chassis as “the chassis of tomorrow” and claims that this chassis has been designed to handle compute, storage, and networking requirements for the next decade. Implied in this positioning is that IBM will constantly update the modules that fit into this chassis — but the chassis will not change (thus protecting customer investments in the chassis itself while enabling customers to constantly upgrade to the latest/greatest compute nodes, storage devices, and switches). The chassis itself takes up 10U worth of space in a standard rack or the Flex System rack (a 19-inch T42 rack with network components and a built-in IBM Storwizev7000 storage subsystem). This chassis can hold up to 14 half-width compute nodes. From a networking perspective, this rack offers much faster communications speeds than traditional blades with high performance 10Gb Ethernet switches with 40Gb uplinks, 8/16Gb Fibre Channel, FDR/QDR InfiniBand as compared with the 10Gb Ethernet, 4/8Gb Fibre Channel, and QDR InfiniBand of blade architectures. At Clabby Analytics, we believe that all major blade vendors will need to rebuild their blade chassis environments in order to support high speed 100Gb Ethernet and faster connections in the future. And this belief is supported by the research of Daniel Bowers of Ideas International who wrote an excellent research report entitled “Is Your Blade Chassis Obsolete?” in which he describes the hardwire connections within various blade designs. As he describes it “most of today’s blade enclosures have one thing in common: a roughly rectangular circuit board midplane built into the enclosure that connects the individual blade servers to I/O devices through a hard-wired "mesh" of copper connections”. Bowers contends that the “style and number of connections in this mesh defines how much I/O bandwidth the blade chassis can handle”. All of today’s leading blade servers can handle 10Gb Ethernet — and even converged Ethernet — but cannot handle higher speeds such as the 40Gb Ethernet and faster speeds available today in IBM Flex System. Blade Storage vs. Flex Architecture — Proximity Is Important Typically, traditional blade servers offer slots for two disk drives located on the blade. IBM’s Flex System architecture provides access to up to eight solid state drives within a single compute node. As IBM states it: “this places hot data closer to the processor”. The proximity of data is extremely important because the closer data can be placed to a processor, the faster it can be executed. We also observe that IBMs eight internal, on-the-compute-node SSDs can act as extended memory that can accelerate the processing of applications that benefit from high IOPS (input/output per second) performance. Applications that should perform extremely well within a Flex System environment include various data mining and database applications, multimedia streaming and
  • 8. A Comparison of IBM’s New Flex System Environment to Traditional Blade Architectures Clabby Analytics http://www.clabbyanalytics.com Telephone: 001 (207) 846-6662 © 2012 Clabby Analytics All rights reserved September, 2012 Clabby Analytics is an independent technology research and analysis organization. Unlike many other research firms, we advocate certain positions — and encourage our readers to find counter opinions —then balance both points-of-view in order to decide on a course of action. Other research and analysis conducted by Clabby Analytics can be found at: www.ClabbyAnalytics.com. video-on-demand, a wealth of financial services applications (that rely on results for quick decision making), surveillance and security applications (especially for real time security checks against reference materials), and video rendering. Summary Observations IBM’s Flex System architecture has many technical hardware and software advantages over traditional blades. But this does not mean blade architecture is dead. From a hardware perspective, traditional blades may use hardware components not yet available on Flex System architecture (such as cell processors or FPGA or other components) to execute specialized workloads. Or traditional blades may run custom software that has not yet been certified to run on Flex System architecture. Or there may be consistency reasons to continue to deploy traditional blades (such as an established management environment, skill sets, and even sparing considerations). On the other hand, Flex System architecture can:  Significantly reduce management costs because Flex Systems are managed as systems instead of discrete components. Managing entire systems (systems/storage/ networking) enables managers and administrators to build and manage systems more efficiently — reducing management costs related to human labor. Further, by managing blades as part of an integrated systems environment, IT executive management can gain a better understanding of how efficient their systems are running — and IT managers can model new workloads (assisting in capacity planning). These are strong run-the-business benefits of a centrally managed, integrated systems environment;  Reduce network latency by up to 50% using an internal switch;  Accommodate future growth with a chassis/rack built for future designs (such as greater than 100Gb switching);  Considerably improve processing speed by placing data closer to processors (in larger memory than offered by most blade vendors — and on internal high-speed solid state disk drives); and,  Deliver superior ROI through increased system utilization. Because Flex System can support more memory, users can load more VM’s onto the same number of CPU’s. Based upon these management and performance benefits, we suggest that IT executives perform an analysis that weighs the acquisition and management costs of existing blade architectures against these new Flex System converged systems architectures. We are certain that IT executives will find that Flex System architecture can tremendously reduce operational costs related to cross system management — while offering higher utilization and significantly better performance than traditional blades (and this will result in much stronger return-on-investment).

×