Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Sun Delivers Extensive Scalability... Without Forced Migrations
This slide illustrates Sun's new Sun Fire Enterprise server product line based on the chip-multithreaded UltraSPARC IV processor as well as Sun's previous-generation Sun Fire V1280-15K server product line based on the UltraSPARC-III processor. Each of these UltraSPARC IV based systems offers up to twice the productivity and computational throughput of their predecessors thanks to Sun's innovative chip multithreading (CMT) technology.
Unlike other vendors, Sun accomplishes this scalability without breaking binary compatibility, delivering unprecedented investment protection. All of these servers share the same constant vision and are binary compatible to run the approximately 12,700 available Solaris applications.
The new Sun Fire Enterprise product line shares even more with a common system architecture, and common components such as the Sun Uniboard CPU/memory board. This architecture, and key components such as the Uniboard, power servers from the rack-mountable Sun Fire E2900 server all the way through Sun's flagship 72-processor Sun Fire E25K server.
Introducing the Sun Fire E2900, E4900, E6900, E20K, and E25K Servers
The new Sun Fire E2900, E4900, E6900, E20K and E25K servers from Sun are powered by new UltraSPARC IV CPU/memory Uniboards. Running across the Sun Fire Enterprise product line, each UltraSPARC IV Uniboard provides four processors (up to eight concurrent threads) and up to 32 GB of memory.
For high-end Sun Fire E20K and E25K servers, new I/O enhancements provide up to 60% improvements in I/O performance over previous-generation Sun Fire 12K and 15K servers. These I/O enhancements come from new Sun Fireplane expander boards and new hsPCI+ I/O assemblies.
Sun Fire E20K and E25K Servers: High-Throughput I/O
As mentioned, Sun Fire E20K and E25K servers provide enhanced I/O capabilities with up to 60% better I/O performance over the previous-generation hsPCI I/O assembly.
These improvements come both a new Sun Fireplane expander board and the new hsPCI+ I/O assembly. In addition to bandwidth improvements, the hsPCI+ I/O assembly supports three 33/66MHz slots and one 33MHz slot providing up to 54 66MHz slots in a single Sun Fire E25K server.
A Scalable Line of Binary-Compatible Servers
This slide illustrates the new Sun Fire Enterprise server product line based on the UltraSPARC IV processor. All of the Sun Fire Enterprise servers are configured to run with multiple CPUs, from a minimum of four processors on the Sun Fire E2900 server all the way up to a maximum of 72 processors on the Sun Fire E20K server.
With each UltraSPARC IV processor capable of running up to two concurrent threads, these servers deliver up to twice the throughput of previous-generation systems based on the UltraSPARC III processor (Sun Fire V1280-Sun Fire 15K servers). In addition to processor performance, these servers are all designed to support very large memory capacities and very high levels of sustained bandwidth across their system interconnects. As an example, the Sun Fire E25K provides up to 576 gigabytes of system RAM and provides up to 57 gigabytes per second of sustained bandwidth across its Sun Fire plane interconnect.
All of the Sun Fire Enterprise servers (except the Sun Fire E2900 server) also provide support for multiple Solaris Operating System domains, enabling multiple completely independent operating system instances to operate on a single hardware server platform.
And, of course, all of these systems are binary compatible and the robust Solaris Operating System runs across the product line. Innovative approaches to data center availability such as Sun Remote Systems (SRS) NetConnect are also supported across the product line.
Common Components
As mentioned previously, all of the Sun Fire mid-range and high-end servers share a common system architecture. In addition, many actual components are common across the product line as well. This slide illustrates the common components among Sun Fire Enterprise servers.
Common CPU/memory, I/O cards and even I/O assembles means that customers can easily move resources between different platform cabinets as needs dictate. This approach greatly simplifies the administration and maintenance of these systems and provides for simpler parts stocking requirements.
Note: The Sun Fire E2900 server is excluded from this slide because its components are physically different than those of the other Sun Fire servers, even though they share the same design
Transition: Understanding how this common architecture is leveraged across such a broad range of server products requires a brief history of interconnect technology at Sun.
Unprecedented Investment Protection: No Forced Application or OS Migrations
All of the new Sun Fire Enterprise servers deliver unprecedented investment protection through Sun's binary compatibility guarantee. Upgrading to UltraSPARC IV technology requires neither changes to applications nor forced operating system migrations as the new Sun Fire Enterprise servers support both the Solaris 8 and 9 Operating Systems.
Existing customers' investments are preserved with flexible upgrades that let them move to UltraSPARC IV technology without disruption. Board-level upgrades let customers install UltraSPARC IV Uniboards in their existing Sun Fire V1280-15K servers. Though additional hardware and software upgrades are required for these upgrades, existing systems can continue to operate.
Existing UltraSPARC III Uniboards can also be installed in new Sun Fire E2900-E25K servers. Individual Sun Fire data center servers can Sun's mixed-speed CPU support to run UltraSPARC III and UltraSPARC IV Uniboards in the same server, in the same or separate domains. With mixed-speed CPU support all processors continue to run at their rated speeds.
Customers can also elect to upgrade entire previous-generation systems to Sun Fire Enterprise servers to take advantage of full factory integration and testing of components.
Note: Sun Fire V1280 and Sun Fire E2900 servers share common CPU/memory boards but that these are not physically the same as the Uniboards that run in the rest of the product line.
Full Hardware Redundancy: Redundant, Hot-swappable Components
To enable high availability, as well as Dynamic System Domains and Dynamic Reconfiguration, Sun's mid-range and high-end server product line provides full hardware redundancy and key hot-swappable components. Two kinds of redundancy are provided in Sun server systems:
N+1 redundant subsystems are truly fault-tolerant in that the subsystem can suffer a failure and remaining components can continue to support the system without affecting operation. Examples of N+1 redundant subsystems in Sun Fire Enterprise servers include redundant system controllers, the system clock, power supplies, and fans.
Redundant components, in contrast, may cause a reboot of a domain in the case of failure but their redundancy allows the system to recover in a stable state following the failure. Sun's Automatic System Recovery (ASR) feature allows systems to reboot automatically after a failure, configuring around any failed components in the process. Examples of redundant components include CPU/memory Uniboards, MaxCPU boards, and I/O assemblies. Even the Sun Fireplane ASICs are redundant so that systems can recover from ASIC failures and continue to operate.
Sun Fire Enterprise systems also support both redundant network and disk interfaces along with software to manage them.
Fifth-Generation Dynamic System Domains
The discussion of Sun's Sun Fire Enterprise product lines begins with the key technology that sets them apart in the marketplace. Dynamic System Domains represent an industry-leading technology, now in its fifth generation that allows Sun's Enterprise servers to be divided into fault-isolated domains, each with their own system resources and separate running instance of the Solaris Operating Environment.
Because faults are isolated within each domain, failing software or hardware does not affect the stability of other domains running on the system. In addition to running a separate instance of the Solaris Operating Environment, each domain also has its own security, and assignable administrative roles. This approach lets multiple groups (or even companies) share a single physical hardware system without risk of accidental security breaches or other inadvertent corruption.
Dynamic System Domains can also be dynamically reconfigured in response to changing demands — either automatically or manually — literally allowing hardware resources to be added or removed from running domains on an as-needed basis. Processors and memory are configured independently from I/O so that administrators can make fine-grained resource adjustments.
Server Virtualization: Granularity at Every level: Hardware, Operating System, and Application
This slide illustrates how Solaris, dynamic reconfiguration, and fault-isolated dynamic system domains can be combined to help drive up utilization in a consolidated environment. Three domains are shown, each running a different version of the Solaris Operating System. Solaris Containers combined with Dynamic System Domains enable a new way to partition systems to achieve higher utilization. The Solaris Next Operating System will enable even more complete fault isolation with trusted containers (next slide)
Sun's servers already offer unmatched dynamic configurability that combines closely-coupled hardware resources with flexible resource management software to efficiently support:
Resource virtualization
Dynamic resource provisioning
Large, capacity workloads
Hosting of virtual horizontal servers
High availability requirements
By consolidating onto larger servers with these capabilities, customers can experience drastically-improved resource utilization.
Fifth-Generation Dynamic Reconfiguration
Also in its fifth generation, Sun's Dynamic Reconfiguration technology allows hardware components to be added and removed from running domains. CPU/Memory Uniboards, PCI cards, and even entire I/O assemblies containing multiple PCI cards can be added or removed, all without requiring the domain to reboot.
This key technology helps enable advanced features such as concurrent repair and maintenance, hot CPU upgrades, and mixed-speed CPU support. Because resources can be added or removed from running domains, system downtime is no longer required for simple upgrades or maintenance. As new or faster processors become available, they can simply be added to running domains without requiring additional down-time.
Sun also provides industry-standard WBEM-based Dynamic Reconfiguration (WDR) that allows system resources such as CPU, memory, I/O assembles, and I/O cards to be visible on a WBEM-aware network as manageable resources. (WBEM stands for Web-Based Enterprise management.) This technology enables centralized resource management through a programatic interface that can be used by remote management applications (such as BMC Patrol). Adding to Sun's previous automatic Dynamic Configuration (ADR), this capability enables automated resource management based on either scheduled or programatic controls.
Solaris Trusted Containers: Optimizing System Resource Management
Complementing Dynamic System Domains, Solaris Trusted Solaris Containers offer a new way to optimize system resources with arbitrary granularity:
Security and fault isolation: Applications are protected from error propagation or security intrusions since each Trusted Container is an isolated, virtual system. Each Trusted Container is shielded from other tenants on that server and each tenant can only see or touch processes in their Trusted Container
Taking Cost Out: Trusted Containers share a single instance of the Solaris Operating System, lowering administrative costs. IT departments can greatly accelerate their server consolidation efforts and reduce costs faster with the flexibility of Trusted Containers. Trusted Containers are designed to have very low system overhead focusing system resources on the workload, not the operating system.
Investment protection: Existing server settings are easily moved to Trusted Containers since each container allows its own IP address, network namespaces, users, disk, etc. Trusted Containers run on existing Sun systems as well as new SPARC, x86, and Opteron systems. Mission-critical applications run at predictable and protected resource levels, are protected from failures of applications in other Trusted Containers, and can be independently rebooted if necessary.
Increased flexibility: Trusted Containers and their resources can be easily configured and manage themselves dynamically via business rules that are set and easily tuned by the administrator. Sharing of unused resources is managed dynamically, automatically maintaining agreed-upon service levels. Usage data is tracked by the Trusted Container allowing IT to demonstrate, and even bill for system resources that are actually delivered.
Sun Cluster Solutions: Sun Cluster 3.0
Sun Cluster solutions delivers high availability through automatic fault detection and recovery along with increased scalability, helping to ensure that mission-critical applications and services are available when needed. Sun Cluster software allows up to eight nodes in a tightly coupled cluster — enough to handle growing numbers of simultaneous users and access to large databases. With Sun Cluster software, administrators can add or remove nodes while online, and mix and match servers to meet specific needs. Individual domains running on Sun Fire mid-range or high-end servers can participate in Sun Cluster 3.0 implementations.
With cluster functions integrated into the Solaris kernel, Sun Cluster software extends the Solaris Operating Environment, enabling core Solaris services — devices, file systems, and networks — to operate seamlessly across a tightly coupled cluster. Global cluster file services, device services, and scalable networking services enable full Solaris compatibility for existing applications.
Enhanced availability for core services such as file and network service allows existing Solaris applications to benefit from higher availability, more localized failure containment, and faster failover.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
This slide provides an overview of the key functions of Sun StorEdgeTM 3310 NAS. The top picture shows the front view with the bezel on. The middle picture show the front view, bezel removed, exposing the 12 hot swap disk drives. The bottom picture show the rear view, exposing the RAID controller, power supplies and connectivity ports.
Key points:
This is a new generation of storage array that was designed to meet a wide range of application demands. It can do this because it provides an incredible amount of flexibility. And what's unique about the Sun StorEdge™ 6120 array is that it's delivered with a suite of comprehensive software tools to help ensure easy manageability and high data availability.
This slide shows some detailed specs on the L25 module. The modules supports DLT 8000, SDLT 220, SDLT 320 and LTO Ultrium 1 & 2 tape drives.
The L25 use two removable magazines, each with 10 DLTtape or 12 LTO Ultrium cartridges (you cannot mix and match DLTtape and LTO Ultrium in the same library module). There are two fixed slots that can be used for data or cleaning cartridges. A single L25 module supports up to two tape drives and 21 DLTtape or 25 Ultrium cartridges. If you stack seven L25 modules in a Sun rack you get up to 14 tape drives and 147 DLTtape or 175 LTO Ultrium cartridges. Thus, you get up to 5TB native with one module and 35TB native with seven modules (using LTO-2) . Up to 60MB/sec native performance is achieved with two drives in one L25 module and 420MB/sec native performance is achieved with 14 drives in seven L25 modules (using LTO-2).
Additionally, magazines in the L25 are interchangeable with the L100 and there are no “right or left” side magazines. This eliminates human error and allows for easy bulkloading and removal of cartridge magazines to offsite locations for vaulting purposes or integration in other modules.
Many competitors in the L25's market do not come close to the density, performance, and simplicity offered in this product. Sun is clearly leading the market segment with the L25 tape library.
This slide shows some detailed specs on the L100 module. The modules supports DLT 8000, SDLT 220, SDLT 320 and LTO Ultrium 1 & 2 tape drives.
The L100 uses eight removable magazines, each with 10 DLTtape or 12 LTO Ultrium cartridges (you cannot mix and match DLTtape and LTO Ultrium in the same library module). There are four fixed slots that can be used for data or cleaning cartridges. A single L100 module supports up to five tape drives and 84 DLTtape or 100 Ultrium cartridges. With six tape drives, one magazine and fixed slot slot is deleted providing storage up to 73 DLTtape or 87 Ultrium cartridges. If you stack two L100 modules in a Sun rack you get up to 10 tape drives and 168 DLTtape or 200 LTO Ultrium cartridges. Thus, you get up to 20TB native with one module and 40TB native with two modules (using LTO-2) . Up to 180MB/sec native performance is achieved with two drives in one 100 module and 360MB/sec native performance is achieved with 12 drives in two L100 modules (using LTO-2).
Additionally, magazines in the L100 are interchangeable with the L25 and there are no “right or left” side magazines. This eliminates human error and allows for easy bulkloading and removal of cartridge magazines to offsite locations for vaulting purposes or integration in other modules.
Many competitors in the L100's market do not come close to the density, performance, and simplicity offered in this product. Sun is clearly leading the market segment with the L100 tape library.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Given these challenges, customers are asking for systems that optimize workloads depending on the applications they are running, at the best price points and at maximum performance.
Depending on the application, the best approach could be vertical or horizontal computing, or typically a combination of both.
This slide illustrates the reach of systems from the heart of the data center to the network edge.
The back-end of the data center will almost always require vertical scalability - large SMP servers – for what we call the data center and application tiers.
For workloads running from the edge of the network out to the client tier, horizontally scaled servers are often the most appropriate choice.
For high performance technical computing workloads, which span the spectrum from 64-bit graphics workstations to large SMP data base servers, both vertical and horizontally scaled servers come in to play depending on the specific need.
This slide illustrates the differences between vertical and horizontal computing.
Vertical servers are large SMP systems with more than four CPUs. There is only one instance of the OS covering the processors, memory and I/O components, generally contained within a single chassis or box. The interconnect is commonly a centerplane or backplane that is tightly coupled, low-latency, high-bandwidth and cahe coherent (which means it maintains information on location of all data regardless of cache or memory location). Resources are added inside the box by adding system boards into the system. Memory is shared, meaning that all processors and all I/O connections have equal access to memory. Memory appears to users as one large chunk.
Horizontal computing is the ability to grow the infrastructure for a given application set by adding discrete compute capacity to a pool of clustered, networked servers. Applications suitable for horizontal scaling demonstrate linear increase in capacity with the number of servers in the pool, each running its own OS image, and each with its own processor and memory. Capacity is increased by adding more servers or nodes.
The workloads, or kinds of applications running in this edge enviroment that we are optimizing the design of our horizontal systems to fall into 3 categories:
At the client we have traditional Tier 0 or network infrastructure services such as: load balancing, SSL accelerators and SSL/SSH encryption are located here. On the client side, you're looking at departmental Web file and print servers, servers for staging applications, servers for locally caching content and e-mail, some access mechanisms and security mechanisms such as firewall and VPN servers, file and print servers, group calendaring applications, and group collaborative computing.
At the Edge or Tier 1 you have web infrastructure services which provide rack and stack web services, caching services, DNS services, directory services and grid computing.
Tier 2 or the application tier includes services like Email, network management, portal computing, streaming media and grid computing.
There are various types of applications that are best suited for vertical systems. Applications that are stateful, that require significant amounts of data, users, and large-scale internal data communication are ideal for vertical servers. It typically depends on the workload size rather than the type of application.
In the Data Center tier, vertical servers are ideal for very large databases, online transaction processing, mission-critical applications of CRM, SCM, ERP, and datawarehousing and business intelligence.
In the application tier, vertical servers are ideal for applications like:
-E-mail and groupware
-Web application servers
-High-performance compute farms
-Electronic commerce applications
-Business integration applications (EAI)
In HPTC, vertical servers are ideal for those applications that have lots of threads that are dependent on and communicate with each other and/or require lots of shared memory.
There are many different benchmarks to demonstrate performance, this slide shows parts of a datacenter that they test. This is one of the reasons that Sun performs a wide variety of benchamrks.
It is important to note that when a benchmark tests lots of layers it really only stresses one tier -- for most benchmarks that is the database tier. For TPC-W it now looks like one type of Image servers in the application tier is stressed (so this benchmark might not reveal much useful information for a customer.
The TPC-W is so new and there is so much optimizations that are going on a people learn to use this benchmark it is difficult to say what tier is most critical for this benchmark.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.
The Ultra SPARC IIIi processor provides a competitive advantage through 100% binary compatibility, allowing for investment protection and simplification of software licensing maintenance, management of costs and flexibility in upgrades.
UltraSPARC IIIi processor is designed with features including integrated L2 cache, memory controllers and 16 GB of addressable memory, all of which lead to lower system cost , lower power consumption and real-world balanced performance with enterprise-class reliability, availability and serviceability (RAS).
Only Sun has maintained 15 years of binary compatibility across processor generations allowing customers to protect their huge investments in the UltraSPARC and SolarisTM platform and related costs.
The Sun Fire V210/240 and Netra 240 servers are powered by the UltraSPARC IIIi processor and were specifically designed to meet the high throughput demands of the web and application tier with Enterprise-level RAS, while holding the line on power consumption and price.
Welcome. My name is ____________.
I'm here to talk to you today about Sun's systems, and how our expansive server line addresses the need for both vertical and horizontal computing in the data center, and why both approaches to computing is important – the differences, the requirements, and the benefits.