Successfully creating and maintaining high performance networks that can sustain enterprise-wide, mission-critical applications requires strategic thinking on the part of the IS manager. In seeking to better manage limited IT resources, the IS manager must simplify and control complexity while adding the new high performance technology required to meet ever increasing user demand.
Strategic thinking requires long-term planning that includes such issues as capacity planning, computing lifecycle costs, resource management streamlining, simplifying the ongoing processes of repair, service and upgrading, etc. Strategic thinking puts the focus on achieving cost-effective, long-term solutions in addition to short-term technology management. As compared to "fire fighting," strategic thinking seeks system-level solutions, not short-term problem avoidance.
Figure 1 - Strategic Thinking vs. Fire Fighting
Choosing and implementing an "industrial strength" high performance enterprise server can be the best way to meet the strategic goals of the IS Manager. An enterprise server system can provide an intelligent network building block, with a framework for integrating advanced technology within a controlled, easy-to-manage resource. But, in addition to selecting a high-performance enterprise server, selecting the right vendor and support team is also an essential part of a successful IT management solution.
Given that in selecting an enterprise server the IS manager is entrusting the computing and connectivity demands of many users to one device, the IS manager must be convinced of several things. First, that the architecture of the server can sustain a high level of availability, even when using inexpensive, off-the-shelf components. Secondly, that the vendor and support team is capable and willing to support the customer in both normal and atypical operating modes. Finally, the IS manager must be convinced that implementing an enterprise server solution will ultimately drive down the cost of operating the computing resource and supporting the end-user.
In Tricord's view, an industrial strength enterprise server must support six fundamental values:
* It must be a IS workhorse
* It must be an end-user racehorse
* It must provide high RAS (reliability, availability, serviceability)
* It must be part of a "Downtime is downtime" philosophy
* It must support a "Fix it FAST" approach
* It must be supported by a customer aware support team
Tricord produces high-performance, building-block server systems that support the IS manager in the integration of advanced technologies, in managing multiple levels of performance, and in minimizing the effort required to service and maintain the system. In short, Tricord seeks to make strategic thinking a reality.
An IS Workhorse From an IS perspective, nothing beats a system that installs easily and then simply runs. A system that takes a regular pounding and doesn't flinch; that grinds out the work with little glamour, but is there day after day -- this is a workhorse. It scales well with end-user workloads, and because it performs so well, applications are continually being added. The Enterprise Server should be the trusted, reliable staple of the processing infrastructure.
The End-User Racehorse Like the power company, the enterprise server should be noticed only when it isn't there, or when it isn't performing. The end-user expects, or more accurately, demands, a consistent, responsive system, regardless of what other users may be doing at the same time. To the end-user, there can be no excuses, the system must be there at the exact time when it is needed, and it must execute the given task responsively. In other words, the enterprise server must provide racehorse like performance - consistently, on demand.
High RAS While fast technology is always impressive, the IS manager is more impressed with dependable, continuous operation than with lighting fast specifications. It is ironic, but true that a technology-oriented organization like IS is driven more by stability than innovation. This is because IS is ultimately tasked with maintaining continuous operations of the IT engine, and must employ reliable, dependable, high-availability systems above all else. This means that mundane, often embedded features such as ECC memory, data integrity checks throughout the entire data path, low maintenance passive backplanes, dual interface buses, and hot swappable power supplies, while not as glamorous as super-fast clock speeds, pipelined CPU architectures and screaming data throughput figures, are ultimately more important in meeting key IS objectives like high availability.
Downtime is Downtime Whether the cause is hardware, software, lightning strikes or power outages, when a server is down, it's down. Regardless of the cause, downtime is unacceptable. Users are out of commission, some portion, if not all, of a company ceases to operate, and costs begin to mount -- immediately. When the downtime clock starts ticking, there is no time to point fingers, assign blame or run for cover. An enterprise server vendor should know this, anticipate emergencies, and be prepared to bring whatever resources are required to quickly isolate the problem, dispatch well-trained assistance, and bring the system back on-line -- immediately.
A "Fix it FAST" Approach Since high availability is of paramount importance to the enterprise server "appliance," when repairs are necessary, it is incumbent that the server design support fault prediction, fault tolerance and fast turnaround in swapping out failed components. Since the fastest fix is one that requires no system power-down, the ideal server would feature redundant components and sophisticated systems management tools that performs automatic, unattended fault detection and fault recovery without any system downtime. This requires several integrated hardware/software components as well as a modular hardware design -- all backed by trained, responsive support personnel.
There are several levels of responsiveness to Fix it FAST:
* Alert notification prior to failure. Give management the option to proactively address the problem before failure occurs.
* Redundant and hot-swap components. Failure in a redundant element will not bring down the system. Hot-swapping eliminates downtime during the replacement process. Both elements contribute to a fix it before it breaks approach.
* Local diagnostic tools. Should a shut-down occur, local diagnostic tools can help quickly isolate the problem area. Comprehensive system logs can pinpoint the nature of the problem allowing system managers to quickly resolve the problem and plan for future avoidance of similar situations.
* Remote diagnostic support. A "Fix it Fast" oriented enterprise server system should support both manual and automatic triggers for calling remote support centers or personnel. These remote system experts should be able to dial into the ailing system, perform diagnostic tests, and isolate the problem. Ideally, software problems can be diagnosed and repaired without the need for physical intervention. Hardware problems may require the use of a spare, or the dispatch of a local service technician. Finally, the remote staff must be equipped with a problem tracking system that monitors the progress of each incident, and also contain an encyclopedia of common problems and their resolutions.
* On-site service. This is the final piece of a highly effective service delivery program. It should be the last resort, one required no more than once in every 20 event incidents. Response time is a function of the travel time required to reach a site, and may be variable depending upon traffic, weather, or resource availability. As the most expensive form of availability assurance, it should also be readily available and time efficient.
A Customer Aware Support Team
In addition to being technically competent, the members of the enterprise server support team should also be experts in the IS environment in which their customers are operating. This means a knowledge of the customer's network environment and applications. Since every customer will apply enterprise server technology in a unique way, the more that is known about the customer, the more safeguards can be built into the system, the way it is operated, the skill level of the support staff, and the procedures to follow. The more that is known about the specific hardware, software and network connectivity of the customer, the fewer problems will occur, and the those problems will be resolved faster.
Now that we have defined some key characteristics of an ideal enterprise server, how does the Tricord PowerFrame Enterprise Server compare?
Tricord was founded in 1987 by a team of engineers from Control Data, IBM, Data General and Stardent -- companies who were leaders in the field of advanced computer architectures. At that time, the MIS battle cry was "Open Architecture, Interoperability, and Client Server technology." Using that as a guide, Tricord has invested considerable research and development in designing a new type of Intel-based server systems that would provide industrial-strength performance in a configuration that supported strategic IS planning.
By incorporating architectural concepts common in the mainframe world, but applying them with existing standards for processor, network bus and operating systems technologies, Tricord succeeded in creating a new type of product - the superserver that combines the best of both worlds: standardized, low cost, easy-to-maintain components in an integrated system with outstanding performance.
Since 1990, when it shipped its first server product, Tricord has invested even more time and effort in building a corporate environment and field sales and support team that makes Tricord PowerFrame Enterprise Servers a powerful solution for today's demanding IS environment.
Tricord's original system architecture was designed to permit the three basic computer subsystems (CPU/memory, Storage I/O, and Network I/O) to operate independently. This allows the addition of multiple resources at any subsystems node, independently of other subsystems. This "functional multiprocessing" results in a system with a very balanced resource to load equation.
Figure 2-- A Functional Multiprocessing Architecture
This original engineering concept resulted in Tricord's logo mark -- a triangle. The Tricord triangle represents the Central Processing Unit and Memory, Storage Input/Output, and Network Input/Output subsystems. These three subsystems in the form of a triangle represent the basic concept that any architectural engineer or mathematician will instantly understand. The triangle is a form that will evenly distribute structural loads evenly between its three points - irrespective of what plane the load is imposed. Likewise, the Tricord PowerFrame Enterprise Server can distribute application loading evenly across its CPU/memory, Disk I/O and Network I/O subsystems.
With commodity PC-based servers, even those with symmetrical multiprocessing, an excessive load on any of the three components will reduce the overall throughput of any standard server.
Figure 3 - PC-based Symmetric Multiprocessing
By comparison, the Tricord combination of functional and symmetric processing makes the Tricord server a higher throughput solution under the variety of application work loads that is typical in an IS environment.
Figure 4 - Tricord Provides Both Symmetrical Multiprocessing and Functional Multiprocessing
Design Improvements and IS Specific Features Through years of continuous improvement and design evolution, Tricord has reworked its basic architecture to improve every aspect of the original design. At the same time, Tricord has accommodated the demands of our IS customers by combining the best features found in PC servers with those previously found only on mainframes.
Such features include:
Ease of Setup and Configuration Although Tricord servers are architected in a high performance mainframe type fashion, they utilize industry standard components, similar to PC-based servers, to simplify administration, compatibility and service. Use of extensions to a standard BIOS instead of developing a proprietary one, for example, allows automatic configuration of the Tricord high performance components via EISA/PCI configuration utilities that anyone who has managed PC servers will find familiar. Exhaustive testing of hardware, operating systems, drivers and third party products in the Tricord laboratories help reduce the workload complexity of the on-site installer.
Investment Protection Tricord's current, second generation PowerFrame Enterprise Servers incorporate a backplane that allows for modular growth, improvement and serviceability. Each subsystem, whether CPU, PCI/EISA, memory, SCSI, or network communications can be independently upgraded. Flash BIOS (local or remote) on each subsystem simplifies in-system upgrades.
Technology Inclusion (upgrade within the existing system) The same techniques Tricord uses to deliver investment protection allows modular refresh of technology as it is developed, and makes it possible for Tricord to keep pace with the demands of our Enterprise customers. This philosophy of upgrading technology within the customer's existing investment contrasts with the typical PC server manufacturer's mentality of discontinuing old product lines with each new processor speed upgrade.
Workhorse Summary Strategic minded IS managers have embraced the Tricord product approach because only Tricord can offer the combination of system and support features that give the IS manager the confidence to know that the computing resource will be there, operating day after day, twenty four hours a day. The IS manager can trust Tricord to understand their needs as an enterprise level customer, to provide an IS workhorse product, and to support that product with the best in service and support.
Ultimately, everything that Tricord does, and everything that our partners in IS do, distills down to one thing: end-user productivity. That is where the proverbial rubber hits the road. IS bases its credibility on the availability and performance of the computing resources it provides to its user community. Tricord and the PowerFrame Enterprise Server provide assurance to the IS group that when a user requests some activity from the server, whether requesting a database to be mounted, a Windows program file, or simply to send an email message, the user won't be disappointed. The enterprise's end-users are allowed to focus on their work, while using the enterprise computing resources as a tool towards that end, seamlessly, all of the time.
Tricord's part is providing a system built for speed. Using a combination of symmetrical multiprocessing PentiumŪ-based CPU/cache architecture with a balanced functional multiprocessing, intelligent subsystem-based design, and very high-speed internal bus structures, Tricord PowerFrames provide outstanding performance. Just as importantly, performance in any subsystem (CPU/cache, disk I/O, or network I/O) can be independently scaled upwards to meet increasing user demand, satisfy more complex applications, and support dynamic network workloads.
For raw processing power, the Tricord PowerFrame Enterprise Server features up to eight Intel Pentium CPUs. Each CPU is supported by its own 2 Mbyte second-level cache subsystem. CPU/Cache Subsystem features such as distributed interrupts for multithreaded I/O, MESI cache coherency, read for ownership, cache-to-cache reads/writes, write-back cache, address pipelining, write-back buffers, separate snoop tags and cached lock cycles squeeze the maximum performance available from the Symmetrical Multiprocessing subsystem. A fast, page-mode dynamic RAM-based main memory system is shared by all processors.
The PowerFrame systems architecture is based on a Functional Multiprocessing balanced-load design that uses separate intelligent subsystems for CPU/cache, disk I/O and network I/O. Linking each component is the PowerBus, a very high speed (267 Mbyte/second) 64-bit bus that provides excellent internal data moving performance. The tight coupling of high speed processors, memory and I/O devices on the same internal bus enables the highly efficient use of computing resources. Independent address and data buses support address pipelining, overlapping the address and data phases of bus transactions (like that employed in the cache subsystem), further increasing data bus throughput.
The disk I/O subsystem (Intelligent Storage Subsystem or ISS) includes its own advanced 32-bit processor system to manage disk I/O processing. It acts as a bus master on the PowerBus, transferring data directly between memory and disk. It off-loads approximately 90 percent of I/O processing typically handled by the main CPU. It provides fast direct connection to 16-bit, Fast/Wide SCSI buses, bypassing the usual PCI/EISA type bus interfaces that can slow data access from SCSI devices. Performance is enhanced by features such as a fault-tolerant cache, elevator seek, overlapped I/O, read priority, scatter/gather and PowerRAID features. These features combine to maximize disk-oriented data access.
Network I/O is supported by a PCI/EISA Bridge Subsystem (PBS). It provides shared bus master slots for connecting industry standard network interface cards (NICs). Data buffering and high speed bursting of data allow concurrent operation of multiple buses and assemble data into efficient data packets. This results in a high performance utilization of network bandwidth.
The advanced design of the PowerFrame Enterprise Server results in the highest performance of any system in its class, even those using similar off-the-shelf components.
An example of the End-User Racehorse concept is a midwestern bank that uses Tricord's PowerFrame product to support over 700 diskless Microsoft Windows users in a 7x24 (seven days a week, 24-hours a day) environment where tens of millions of dollars per hour are at risk if the system is not available or performing to specification. Tricord's unique system architecture, previously described as the Triangle, allows users to enjoy availability and performance regardless of what their computing neighbors are doing. In many organizations, the presence of a Tricord Enterprise Server seems to be secret. The users rarely have the occasion to ask about its existence -- they just benefit by its presence.
Tricord has evolved products to match the changing needs of our customers. Evidence of this evolution can be found in the direction of both Tricord's product development ad marketing organizations. In earlier years, performance was king. As the enterprise customer placed more and more critical operations on these servers, the performance almost became a given, next focus was management. For some time management was the industry buzzword and IS found that knowing when the server was down, or even better, about to go down, was great information, but didn't really solve the problem. This realization prompted focus on the next issue -- availability. IS must serve their customers, period. This service requires available computing platforms with minimal interruptions from any component in the system.
Figure 5 - The Tricord High Availability Triangle
Tricord has designed its systems with both embedded high-availability features (redundancy, hot swapping of components, ECC memory, data parity checks throughout the system) and system-level availability features (intelligent system monitoring software, independent fault prediction, monitoring, and automatic fault recovery procedures). Tricord is continuing to improve its availability performance through its commitment to corporate-wide quality and support programs that result in high-availability systems.
Tricord's architectural modularity, highly integrated diagnostic and management systems, as well as a robust list of fault avoidance and recovery mechanisms all contribute to the IS contract with its customers/users. This combination of architectures and features provided using standard proven technologies allow rapid discovery of incident causes, quick and modular component replacement and the support options to ensure that the people and parts are mobilized appropriately.
Tricord ensures system availability from its twist-lock power cord to avoid accidental power outages all the way to CPU standby/failover capability and rapid customer support. While many of the PowerFrame Enterprise Server's availability features are standard, the modularity of components allows an IS team to measure the cost of downtime and invest in just the amount of high availability features that this cost exposure warrants.
The following list of availability features is offered standard on the Tricord PowerFrame Enterprise Server. It sets the watermark of the Enterprise Server label.
* Intelligent Management Subsystem plus battery backup for IMS and modem
* ECC memory - scrubbing and configuration mapping
* Parity protection on CPU, and all data paths
* CPU standby and failover
* Intelligent Storage System (ISS) duplexing and electronic bus isolation
* ISS RAID and duplexing/quadraplexing/octaplexing
* ISS fault tolerant disk controller cache with ECC
* Disk hot standby
* Differential SCSI drive
* Protocol level protection for network I/O
* Parallel PCI buses
* Redundant NICs
* Redundant power supplies
* Passive backplane - no mother board
Features such as these ensure that the Tricord Enterprise Server products will exhibit high RAS and meet both IS and end-user requirements consistently and dependably.
While product features and innovative new technologies all work to reduce system downtime, Tricord recognizes that IS is dependent upon their enterprise server vendor's people to respond when they call. Tricord's experience in the mainframe world has taught Tricord to accept the tenet: if a customer is down, there is severe impact on the people dependent upon that enterprise, and therefore, it is imperative that the vendor's people work relentlessly to get the system running again as fast as possible.
Downtime need not be attributed to hardware, software or force majeure; when the system is down, it's down. While this may seem like a truism to most, it encompasses a philosophy intended to reduce the finger pointing that frequently occurs in the PC server world. Tricord's goal is to fix the customer's problem and worry about who/what is responsible later, once the system is back on-line and operational.
To aid Tricord's efforts in resolving customer downtime issues, much groundwork has been prelaid. Deeply integrated hardware management systems allow Tricord diagnostic teams to access a customer's system in the field and derive information down to the board level.
Tricord's status in the industry as a high-end player with a very impressive customer list in the Fortune 1000 lends greater status with operating systems vendors. For Novell and Microsoft, Tricord has full OEM status and is given dedicated resources within these organizations to act as liaison for problem resolution. For other products that Tricord integrates or recommends, Tricord has established solid relationships and defined responsibilities to respond to customer situations.
Lastly, as an umbrella, Tricord is a member of the industry support organization known as Technical Support Alliance Network (TSAnet). TSAnet enforces a process of escalation between vendors working collaboratively to resolve customer issues. If there is disparity in arriving to problem resolution, the TSAnet serves as an impartial arbiter between the vendors with end-customer satisfaction as the primary concern.
So, for Tricord, maintaining high RAS is much more than just integrating advanced hardware and innovative software, it is a core value, and an understanding of the pressures on IS to provide dependable service.
Tricord holds a philosophy of downtime avoidance, first and foremost, and failing that, rapid recovery -- in many instances without user intervention. Tricord designs hot swappable componentry into the PowerFrame Enterprise Server, but prefers to use this as a maintenance feature since human intervention and an untested part from a shelf needs to be employed to effect repair. Tricord's innovative hot standby of disk drives, power supplies and network interfaces provide IS with an even more robust solution than ever before. In some cases, the standby components may even be used to aid in performance, but still maintain the ability to take over system operation, if a primary device fails.
In order to fix a problem fast, one must be able to rapidly identify the root cause of the symptom. Tricord's Intelligent Management Subsystem consists of hardware and software that does this and much more. Architectural modularity allows a very efficient Mean Time To Repair (MTTR) as well as efficient upgrade time. A disk drive can be replaced in mere seconds, and background rebuilding of the data onto the new drive is largely non-impacting to normal operation due to localization of the operation to the Intelligent Storage Subsystem.
Standard on every PowerFrame Enterprise Server are features that are designed to expedite the restoration of system operation:
* Intelligent Management Subsystem plus battery backup for IMS and modem
* ISS disk drive hot swap
* Power supplies and cooling fans x3 hot swap
* Remote administration console
* 7x24 rapid response service agreements
* Spare parts located at customer site
While the hardware is highly maintainable, this is for naught if the trained personnel, information, and spare parts cannot be found. Tricord offers a wide variety of service programs ranging from contract maintainers to Value Added Resellers to Self Maintenance programs. Internal IS staff may be sent to the manufacturer service training at the customer's option. Once this has been completed, direct contact to Tricord Technical Support (TTS), on-site spare parts and access to on-line information resources within Tricord are available.
No one knows the IS environment better than the IS staff itself. Once properly trained in the use of Tricord's maintainability features and services, fix it fast becomes as much of a truism as much as downtime is downtime.
As an operational partner with many enterprise level customers, Tricord has, over the years, gained much information and deep appreciation for the needs of those customers. Due to Tricord's unique market position, rarely is a Tricord customer just doing NetWare. Frequently, Tricord customers are early adopters, blazing the trail for the rest of the industry. While this is an exciting position for any company, it has significant costs. In helping these customers deal with leading-edge situations, Tricord has become part of the feedback loop among industry leading providers of advanced technology. Tricord has even become involved the development of operating systems and other enabling technologies that will shape the future Is enterprise environment. Frequent communications and involvement of Tricord system engineers and engineering technologists in our customers' system designs allows us a deeper view into the operational needs of our customers. This is why the PowerFrame Enterprise Server from Tricord matches so closely the IS needs of our customers.
Tricord is totally focused on providing enterprise servers designed to be intelligent building blocks of a flexible, scalable, and highly reliable/highly available enterprise computing system. Tricord's only corporate focus is meeting all the needs of the corporate Information Technology manager. With a field-proven architecture that can incorporate the latest in CPU, memory, disk and I/O technologies, a local support organization in place for 24 hour support, and a commitment to fix it FAST and a downtime is downtime philosophy, Tricord is your best choice for strategically sound network solutions.
Tricord has the technology, the architecture, the support programs and the commitment to maintain its leadership position in performance, availability, reliability and customer satisfaction.
END OF WHITE PAPER