Sizing Information - Lenovo S3200 Quick Start Manual

Lenovo Storage S3200 with Fibre Channel
The Emulex 16Gb (Gen 6) FC HBA features up to 1.6M OPS per port, the highest
available on a 16Gbps FC adapter. This is achieved through Emulex's Dynamic Multi-
Core Architecture which dynamically scales HBA resources to any port that needs it,
essential when ports are used in active-standby mode. The Gen 6 HBAs can process up
to 71% faster SQL Server transaction and up to 50% faster data warehousing
transactions over 8GFC HBAs. They are ready for NVMe over Fibre Channel, which
accelerates network access to SSDs, supporting the upcoming NVMe over Fibre
Channel T-11 standard.
All Emulex adapters allow you to deploy faster and mange less with a unified
management tool (OneCommand® Manager) and common driver set.

Sizing Information

Server Sizing
Sizing a virtualized solution has numerous complexities for optimization that can be
taken into consideration for a completely optimal deployment. Numerous technologies
are on the market that can make this job significantly easier. VMWare® Distributed
Resources Scheduler (DRS) is the software product that enables VMWare to assist in
moving virtual machines to optimal locations in a cluster. With DRS, you can select to
allow migration to happen automatically or to merely monitor and suggest changes that
the administrator can allow or disallow. This feature is available in VMWare® Enterprise
and Enterprise Plus and simplifies both the initial sizing and understanding what
compute resources may be required going forward.
When a cluster is running DRS, server sizing requires a few simple yet complicated
questions:

What is the CPU capacity required for my applications?

How much RAM is needed for my applications?

Is there a requirement for a large amount of network or storage I/O?
If the current environment is not setup for virtualization yet, there are tools on both
Windows and Linux that can help with the sizing. Standard Linux tools such as sar, ps,
mpstat, top, and free can give a very strong idea on what may be required of a system's
virtualized equivalent when compared to the hardware that is inside a server. Windows®
has similar commands that can help with this transition such as PerfMon that will give
detailed information about how all systems are functioning. If already running VMWare®
vSphere on the cluster, esxtop is a good choice in gathering this information to help in
sizing.
Once this information has been gathered, it should be compiled together. Approximately
20% headroom for future growth or burst usage should be left as available. For RAM
10