1 / 9

System Scalability

System Scalability. 1. General Observations

donar
Télécharger la présentation

System Scalability

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. System Scalability

  2. 1. General Observations • The choice of platform for an application should consider the ability to grow the application with more users and more data. Scalability is the ability to grow a system over a wide capacity range to meet the processing needs of the business. Scalability includes not only processor power but also processor memory, I/O bandwidth, the ability to attach large numbers of disks, the ability of the software to exploit large systems, and the ability to manage a large system. • Simply adding additional resources, such as processor engines or additional systems, does not always improve total system performance. Adding extra components causes the number of interactions between them to grow exponentially. This is known as the large system effect. In fact adding additional resources can actually reduce system capacity rather than increase it.

  3. There are two broad approaches to handling application growth: • Vertical growth is the ability to upgrade the installed server processor to a larger one within the same range, for example by adding further processor engines to an SMP. In vertical growth, the SMP ratios of the processor range are important. • · Horizontal growth is the ability to install additional processors alongside the current ones. Ideally we want all of the processors to work together in a cluster, which will provide the benefits of • "Clustering Architectures".

  4. Figure 1. Parallel Sysplex Scalability For An IMS-TM/DB2 Workload

  5. 2. Sysplex Scalability Linear scalability of clusters is important, as it governs how efficiently clusters can run (just as SMP ratios govern how efficiently SMPs can run). This affects how many servers are needed and how large the cluster can grow. S/390 Parallel Sysplex achieves near linear scalability as additional servers are added. Figure 1 shows this for configurations going from two nodes to eight nodes for an IMS-TM/DB2 workload For both the first (R61) and second (RX3) generations of IBM's 9672 Parallel Enterprise Servers the scalability of the Parallel Sysplex (3.9 times) is very close to the theoretical maximum (4 times). It is also superior to the SMP scalability of the 9021 range (two-way Model 821 to eight-way Model 982). The potential throughput capabilities of the IBM Parallel Sysplex can be seen in Figure 2.

  6. Figure 2: Parallel Sysplex Throughput

  7. For a system of 32 9672-RX3 processors a capacity of approximately 2,260 IMS-TM/DB2 transactions/second (135,000 transactions/minute) is possible. For the new IBM 9672 Generation 3 servers we estimate that 32 9672-RX4 processors in a Parallel Sysplex would have a capacity of approximately 4,240 IMS-TM/DB2 transactions/second (250,000 transactions/minute). This is believed to be an order of magnitude greater than achievable with any other platform. For comparision: The TPC council publishes transaction performance data for many Unix configurations (see http://www.tpc.org). As of January 2000, the top performer was an AIX 6000 S80 server with a rating of 135 815 tpmC, closely followed by a Sun Enterprise 6500 Cluster with 135461 tpmC. No S/390 numbers are listed by the TPC Council, but an S/390 RX6cluster is assumed to exceed 1 000 000 tpmC. For more information on these Parallel Sysplex scalability numbers see System/390 MVS Parallel Sysplex Performance, SG24-4356-01

  8. 3. TransactIon Performance • International Technology Group (ITG), in a documentation "E-Transactions in Financial Services", quotes in 1Q2000 the following figures: • The largest OS/390 banking system in the Unites States currently handles transaction volumes 78 times greater than the largest system on any other platform. • · In brokerage and insurance, OS/390 systems are 35 and 29 times lareger respectively.

  9. 4. Web Server Performance The most popular Web sites handle in excess of 100 million page views a day. Market forecasts make it clear that those types of numbers will become commonplace. Revenue for most online retailers is growing by more than 200% annually, with total online sales topping $13 billion this year, according to Boston Consulting Group Inc. Over Thanksgiving weekend, traffic on CDNow Inc.'s music site tripled, while orders on Amazon.com's book and music store were up 400% over last year.

More Related