2 - USP-V High Level Overview

The USP V is what is termed an External RAID Array. This “external” implies that it exists physically outside of the server, i.e. the disks are not physically located within on the end of a serial SCSI cable inside of a server. The “RAID Array” implies that it is an advanced and so-called intelligent disk subsystem and not just a bunch of disks (JBOD) hung together with no associated intelligence. And of course RAID implies that it supports many levels of RAID protection to both protect the data it stores as well as affect the performance thereof.

Because it is “external” it can have multiple servers attach to it concurrently. The USP V supports modern servers and operating systems attached to it including Mainframe systems. In fact it has its roots in Mainframe and has only more recently been extended to Open Systems hosts. However, this manual deals only with the Open Systems side of the array.

The basic concept of the USP V is to provide servers and applications with persistent storage. The USP V itself can hold more than 1,000 disks (1,152 to be exact) in a fully configured 5 frame system. These disks are hardware RAID protected and carved up into logical devices (LDEVs). These LDEVs can be any user configurable size within reason. Current limits are from as little as 40 or so MB up to 4TB for a single LDEV, but considerably larger using array based functions to concatenate multiple LDEVs that appear to hosts as a single volume. These LDEVs are then presented out to servers as SCSI Logical Units, often referred to as LUs or LUNs. To the attached servers, these LUNs look just like internal disks and can be used as such. For example these LUNs can be mounted as standard Unix and Windows mount points and have user data stored to them exactly like normal internal disk drives. Servers connect to the USP V via Fibre Channel Ports often referred to simply as front end ports. Servers can connect directly to the front end ports on the USP V or via SAN switches.

The USP V has a large global DRAM cache and is also often referred to as “cache centric”.
The global cache is the real heart of the USP V and is where the good stuff happens. At the
time of writing, the USP V can be configured with 512GB of mirrored cache. This allows to
USP V to operate at near the speeds of the DRAM cache rather than the disks on the back
end. The USP V is extremely fast.

Behind the cache are disk adapters, referred to commonly as Back End Directors or BEDs. These BEDs control, among other things, the physical disks. They also provide hardware RAID protection. Behind the BEDs are the disks. May be “Drives” is a more appropriate term since the recent introduction of SSD Enterprise Flash Drives. This is the ultimate destination of data written to the USP.

Layered on top of all of that are many additional services provided by the USP V. These can be broken down in to the following three broad categories –

* Copy Services – Disk-to-disk backups, and Copy-On-Write snapshots....
* Replication Services – Synchronous and asynchronous distance replication
products...
* Value-Add services – Storage Virtualisation, Dynamic Provisioning, Thin
Provisioning...
In a nutshell, the USP V is a top of the range enterprise class external RAID array that is considered by many to be best of breed.