Some application and workloads lend themselves particularly well to scale-up rather than scale-out deployments. Factors contributing to a decision to scale up rather than scale out include processing and memory requirements, the number of end users, scalability cost and administrative cost.
Scaling up allows you to add CPU, Memory, I/O and Storage to your servers, and it does not require you re-write your application to harness the new horsepower.
For large databases, scale-up architectures can provide higher levels of scalability than large numbers of scale-out distributed databases, and scale-up servers are often easier and less expensive to manage.
Lets take a look at some Enterprise applications that are better suited for 4 and 8 socket scale-up deployments:
SAP HANA: SAP HANA as an integrated solution combining software and hardware, frequently referred to as the SAP HANA appliance. It consists of the SAP HANA database and adds components needed to work with, administer, and operate the database. SAP HANA, almost by definition is a scale-up deployment.
SAP Applications. SAP is typically deployed in a 3-tier architecture, and the database layer cannot be scaled merely by adding servers. Properly scaling to handle larger database loads requires deploying them on hardware that is suited to house both the SAP applications and their associated databases. Scale-up can also be used to consolidate multiple instances of SAP scattered around the organization onto fewer, virtualized, scale-up servers
Microsoft SQL Server. Microsoft designed SQL Server to scale up rather than to scale out in a single-instance scenario. While it supports many high-availability features such as clustering, a single SQL Server database is not designed to be deployed across multiple server platforms as a single virtual instance. This means that scaling is best achieved by upgrading the underlying hardware on which the database runs rather than by adding instances on separate servers.
Oracle Database. While Oracle Database software products support both scale-out and scale-up computing, deploying Oracle Databases on scale-up servers brings a number of benefits, including simplified server management, reduced server licensing costs, and reduced operating costs related to IT staff time and management. When scaling-up, Oracle Real Application Cluster (RAC) is not required. Another benefit is that machine resources (especially CPU) are instantly available for sharing.
Microsoft Exchange. Historically, many customers added servers to grow capacity. This approach led to server sprawl and to time-consuming, inefficient Microsoft Exchange systems management. Today, multirole deployments combine Mailbox, Hub Transport, and Client Access Server (CAS) roles onto a single server in a particular service delivery location. With this architecture, effective scaling with reduced overhead is achieved through scale-up — running the multirole server on more powerful hardware — rather than through scale-out, adding more low-end hardware to handle more mailbox servers. By combining these servers onto a single scale-up server, organizations can take advantage of the high-performance hardware resources on modern servers while simplifying deployment and management.
Enterprise Custom Applications. Custom applications, many of which are legacy applications running on aging centralized hardware, drive many mission-critical workloads in enterprises today. And as demands on these applications grow, the underlying hardware is not always able to keep pace. There are many advantages associated with rehosting custom-developed applications on scale-up x86 servers. First is the opportunity to include workloads on standardized IT infrastructure based on x86 server hardware, running Microsoft Windows or Linux – in place of operating systems that have reached their end of life (EOL). Rehosting on scal-up x86 server hardware also
brings greater IT simplicity, reduces operational costs associated with management and downtime, and reduces server footprint in the datacenter.
Other workloads that benefit from scale-up, deployments that provide single system high performance, large memory and individual server scalability.
• In-Memory Analytics
• IBM DB2
• Business Intelligence
• Virtualized Consolidation
• Business Intelligence
• RISC Migration
• Decision Support
• Large-Scale, Virtualized Applications
• Enterprise-Critical Middleware
• Large memory, consolidation workloads
As workload performance increases and memory and scalability requirement increase, the need for scale-up workloads and big applications increase. This drives the need for individual system scalability and resiliency. Handling high performing, scale-up, mission critical workloads is where IBM X6 servers shine.
An analyst report titled “ Can IBM Revitalize 8P with x86”, also talks about applications driving the need for scale-up servers and why the x3950 X6 is a superior choice for these applications. Look under "Analyst reports"
Another analyst report titled "State of the Art Single System Availability - X6 Style", talks about the importance of availability and RAS for Enterprise Mission Critical workloads. Look under "Analyst reports"
In the next blog, I will talk about the business value of 4 and 8 socket scale-up deployments.
IBM has three powerful 4 and 8 socket scale-up servers to meet client needs:
4 to 8 processor scalability
Up to 120 cores of processing
Up to 12TB of memory
Up to 22 PCIe slots
2 to 4 processor scalability
Up to 60 cores of processing
Up to 6TB of memory
Up to 11 PCIe slots
2 to 4 processor scalability
Up to 48cores of processing
Up to 1.5TB of memory
Up to 8 PCIe slots
Randall Lundin is a World Wide Product and Marketing Manager for IBM System x. He manages the Enterprise 4 & 8 socket rack systems. Follow Randall on Twitter at @RWLUNDIN or view Randall's LinkedIn profile.