How to Design Storage System for a Multinational Company?


Data centers are turning out to be progressively intricate as organizations depend all the more vigorously on IT for basic operations, data development is huge, and spending plans are tight. Steady accessibility is a prerequisite for most firms, constraining windows for upkeep and bringing down resilience for blackouts. Late developments in Ethernet networking present the likelihood of network unification using a solitary innovation for both SAN and LAN. Progresses in Ethernet innovation and the development of protocols, for example, iSCSI and FCoE offer deterministic execution, low latency, and steady data accessibility on Ethernet.

An upgraded lossless form of Ethernet is at the heart of empowering FCoE, which offers current clients of FC SANs a way to move to network joining while keeping up zoning hones, ability sets, and administration devices that are as of now well known. At server side, the possibility of shared excess ports for both front-end LAN movement and in addition back-end SAN networking guarantees cost reserve funds and in addition an expanded capacity to virtualize server I/O. The appropriation way of FCoE is showing in two stages, persuaded by the accessibility of supporting innovation from network, server, and storage sellers.

As conventional backup arrangements just can’t comprehend modern threats. A few techniques wastefully store data over and again crosswise over sites and servers, altogether extending all out storage under administration and expanding costs. What’s more, customary tape based recovery can take days or even weeks, particularly if the data is put away offsite. Recovery from plate is normally quicker, however regularly requires different steps, which can be repetitive. Customary NAS backup arrangements normally depend on level-0 and level-1 backups; repeating week after week full backups and day by day incremental backups.

Nonetheless, the full backups regularly broaden past the accessible backup window, which chances leaving data unprotected. What’s more, sometimes, day by day incremental backups might take the length of full backups; particularly if just a bit of an expansive document has changed, is moved following total record. The utilization of tape is additionally testing because of temperamental media and equipment, delays in offsite shipments, and the danger of losing decoded tapes that contain delicate data. From a recovery viewpoint, customary strategies frequently include a dull, multi-step process, requiring reclamation of the last great full and consequent additions to achieve the wanted recovery point.

Also, recovering tapes from offsite storage can bring days or more with no assurance the data is recoverable. Likewise, the basic routine of putting away depictions and essential data on the same NAS system can prompt data loss if the nearby NAS system falls flat, the principle volume is defiled or is devastated and data has not been imitated offsite. Keeping in mind offsite replication is frequently wanted, a lot of data and constrained WAN transfer speed can make it troublesome. Thus, NAS solidification and utilization are frequently constrained by the backup window, not storage limit. Furthermore, customary backup confinements can prompt data loss, declining client efficiency, compelled NAS execution, and overburdened IT staff.

Component Overview:

As we are about to design a SAN environment for end to end SAP systems for an organization let’s have a look at how SAN is the better option for us than other solutions. VMware ESX Server can be utilized as a part of conjunction with a SAN, a particular high‐speed network that associates PC systems to elite storage subsystems. Utilizing ESX Server together with a SAN gives additional storage to combination, enhances dependability, and assists with calamity recovery.

In its least complex frame, a SAN comprises of multiple servers joined to a storage array and besides utilizing one or many SAN switches. Every server may have various applications that require devoted storage for applications preparing. Third necessity for SAN will be SAN switches that interface different modules of the SAN. Specifically, they may associate has to storage arrays. SAN switches additionally permit managers, in the case of a way disappointment from host server to switch or from storage array to switch.

The SAN fabric is the genuine network bit of the SAN. Whenever multiple SAN switches are associated, a fabric is made. The FC convention is utilized to impart over the whole network. A SAN can comprise of numerous interconnected fabrics. Indeed, even a basic SAN frequently comprises of two fabrics for repetition. That additionally incorporate the Connections which essentially includes on two things: Storage Processors & Host Bus Adapters. Storage systems are associated with the SAN fabric through ports in the fabric as well Host servers. A host associates with a fabric port through a HBA. Here in the image below is the basic idea of a SAN components:

storage system

Considerations for SAN Components:

While picking SAN segments we generally need to study the extent of our association, we likewise need to figure out if we have to utilize the Fiber Channel Arbitrated Loop (FC-AL). On the off chance that we need that we might require extra SAN parts, for example, FC-AL upheld centers. The bigger a SAN fabric is, the more appealing executives get to be, considering the expense per usable port and simplicity of administration and upkeep. Switches are basic and adaptable parts. They can be interconnected for high accessibility and oversaw by an expert SAN administration programming, similar to Tivoli Storage Network Manager (TSNM).

Zoning is the place where parts of SAN are concealed from Host Bus Adapters (HBAs), the targets to be seen by HBAs are also selected here. It is done for security usually behind the reason for disconnecting private systems or data.

Logical Unit Number (LUN) masking is regularly utilized for consent administration. LUN masking is likewise alluded to as particular storage presentation, access control, and apportioning, contingent upon the seller. LUN masking is performed at the SP or server level; it makes a LUN undetectable when an objective is checked. The executive arranges the disk array so every server or many servers can see just certain LUNs. Masking abilities for every disk array are merchant particular, similar to the instruments for overseeing LUN masking.

In this proposed plan we will be taking after these methodologies so that our association profits by utilizing SAN as much as it can:

Heterogeneous storage focuses on the same SAN should be in partitioned data zones. Any upheld host can have entry to either or both, yet the storage systems themselves not to have the same zone.

Heterogeneous HBAs might exist together on the same SAN, however should be set in particular data zones. HBAs from various producers must be zoned.

Homogeneous fabric

Know that HBAs use merchant particular drivers and firmware. Affirm that the firmware and miniaturized scale codes dispatched on the HBAs are upheld.

HBAs inside of a host ought to be homogeneous. HBAs from various makers, for example, Emulex and JNI, not being blended on the same host.

Heterogeneous working systems on similar SAN should be in partitioned data zones. Distinctive working systems having identical SAN network might bring about clashes.

Architecture Plan & Design Phase


We have investigated the interoperability necessities. As per the given situation, now we will look whether all SAN storing gadget is legitimately upheld in the given situation. EMC Symmetrix ought to likewise be incorporated into the thought.

Machine type Quantity O/S HBA
7026 H80s 2 AIX 4.3.3 IBM Gigabit FC Adapter
PC server 1 NT 4.0 Qlogic QLA2200F Adapter
HP 9000 V 1 HP UX 11.0 HP A5158A Adapter
SUN Enterprise 4500 1 Solaris 7 Qlogic QLA2200F Adapter



Three separate zones have been characterized for the two pSeries servers, the two pSeries servers are gathered together in light of the fact that they have similar HBAs and alike working systems. The three storage devices are allocated to discrete data zones since they are heterogeneous. This permits the AIX systems to grasp the majority of the storage devices, yet keeps the storage systems from interfacing. The same three data zones would need to be made for alternate servers on the grounds that the HBAs on these servers or system is heterogeneous. A sum of twelve zones needed for this situation. See the image below:

storage system

Consideration for Cables:

Sorts of optical fiber cables are regularly recognized by the distance across of the center and cladding, measured in microns. An optical fiber cable having a center width of 50 microns. These fiber cables come in two unmistakable sorts, which are connected to their sizes: Single-Mode Fiber (SMF) for larger spaces and Multi-Mode Fiber (MMF) for short separations like 2 km.



Mode Laser type Distance
50 Multi-mode Short wave <= 500 m
9 Single-mode Long wave <= 10 km
62.5 Multi-mode Short wave <= 175 m


BM pSeries Gigabit Fiber Channel PCI Adapters bolster 50 and 62.5 micron multi-mode associations. For 9 micron single-mode associations with these connectors you should utilize a suitable SAN gadget either switch or center point as a connection between the two modes and cable sorts.


All Fiber Channel cable associations are made to the front board of the switches utilizing short wavelength (SWL) fiber optic and long wavelength (LWL) fiber optic with double standard connector (SC) plugs. The IBM SAN Fiber Channel switches give a 10/100BaseT Ethernet Port that can be utilized for associating switch administration reassures. This element permits access to the switch’s inner SNMP specialists, furthermore permits remote Telnet and Web access, empowering remote checking and testing. A serial port is given on IBM SAN Fiber Channel switch Model 2109-S08 for recovery from loss of secret key data, and troubleshooting purposes by recouping manufacturing plant settings and for the underlying design of the IP address for the switch.

Suggested Design:

According to our scenario we are at the point when outlining a SAN for numerous applications and servers, we should estimate the performance, dependability, and limit characteristics of the SAN. Every application requests assets and access to storage gave by the SAN. The SAN’s switches and storage arrays must give opportune and dependable access to all contending applications. The SAN must bolster quick app reaction times reliably despite the prerequisites made by applications shift over peak periods for both I/O every second and bandwidth. An appropriately composed SAN must give adequate assets to process all I/O asks for from all applications. Outlining an ideal SAN environment is neither basic nor fast.


So we designed it in a way that every RAID bunch gives a particular level of I/O performance, limit, and excess. LUNs are doled out to RAID groups in view of these necessities. The storage arrays flow the RAID groups over every single inward channel and access ways. This outcomes in load balancing of all I/O solicitations to meet performance prerequisites of I/O operations every second and reaction times. If a specific RAID bunch can’t give the required I/O performance, limit, and reaction times, you should characterize an extra RAID bunch for the following arrangement of LUNs.

Adequate RAID‐group assets were given to every arrangement of LUNs. SAN plan on peak‐period movement and consideration was based on how I/O would behave in every peak period. It was also considered that extra storage array asset limit is required to oblige prompt peaks. In our situation peak period might happen amid early afternoon handling, portrayed by a few peaking I/O sessions requiring twice or even multiple times than normal for the whole peak period. Without extra assets, I/O requests that surpass the limit of a storage cluster result in deferred reaction times.

In spite of the fact that ESX Server frameworks advantage from compose store, the reserve could be soaked with adequately extraordinary I/O. Immersion decreases the reserve’s adequacy. Excess SAN equipment segments including HBAs, SAN switches, and storage cluster access ports, are required. At times, various storage arrays are a piece of a fault‐tolerant SAN plan. I/O ways from the server to the storage array must be repetitive and progressively switchable in the case of a port, gadget, link, or way disappointment. Reflecting assigns a second non‐addressable LUN that catches all compose operations to the essential LUN. Reflecting gives adaptation to non-critical failure at the LUN level.

If a site fails for some cause, we need to immediately recover the failed applications and data from a remote site. The SAN provides right to use of data from an alternate server to start the data recovery process. The SAN handles the site data synchronization. We placed ESX Server there for this very reason as using them can make disaster recovery much easier, we simple restore VM Image and resume the work where it was stopped.

storage system

SAN Element SAN Component
Software OS:


IBM pSeries servers: AIX
Version 4.3.3/updated or AIX Version 5.1


HACMP: Version 4.3/4.4.0 with APARS IY12057, IY11565, IY12022, and IY11563
RS/6000 SP servers: PSSP Version
3.1.1 or higher  and AIX Version 4.3.3

SAN interlocks HBA FC# 6228: Gigabit Fiber Channel Adapter for 64-bit
PCI Bus with HBA Firmware Version 3.82A0 or higher
Server IBM pSeries servers 660 7026-6H1

RS/6000 SP servers: AIX Version 4.3.3

Topology One host server per FC port on the switch, End-to-end systems.


Cascaded Switches 2109 with 2×4 Arrays


FC Hub Attach: IBM Fiber Channel Storage Hub



Storage Disk systems Enterprise Storage Server Enterprise Storage Server (2105-Ex0)


NAS , Tape systems


Please enter your comment!
Please enter your name here