Virtualisation of SAN drive arrays has typically been viewed as a Fibre-Channel-centric activity. With the rise of iSCSI as an alternative SAN networking scheme then virtualisation of IP storage - iSCSI drive arrays - is a logical application. This SANRAD white paper introduces and explains the features and benefits of doing it.
www.sanrad.com SANRAD White Paper: V-Switch 3000" - Storage Virtualization WP 001-04 SANRAD US Tel: +1-510-521-2424 International Tel: +972-3-767-4800 info@SANRAD.com Copyright SANRAD 2003 All rights reserved. The copyright and all intellectual property rights in this article belong to SANRAD. It is strictly forbidden to copy, duplicate or otherwise use this article or any part thereof in any way shape or form without the prior written consent of SANRAD. Untitled Document WP 001-04 Copyright SANRAD 2003 2Table of Contents INTRODUCTION..............................................................................................3 THE CHALLENGE FACING NETWORK ADMINISTRATORS .....................................3 BUSINESS DEMANDS NETWORKED STORAGE BUT HOW? ..............................5 FIBRE-CHANNEL BASED SANS- NO LONGER THE ONLY OPTION.........................6 ISCSI ENTRY LEVEL TO ENTERPRISE ..........................................................7 The Benefits of iSCSI...........................................................................................8 VIRTUALIZATION - BETTER STORAGE MANAGEMENT.........................................9 DIFFERENT METHODS OF VIRTUALIZATION ....................................................10 THE V-SWITCH 3000 - COMBINING THE BEST OF ALL METHODS .....................11 ISCSI TCP/IP HOSTS AND THE V-SWITCH 3000 ..........................................13 iSCSI Host Support ............................................................................................13 Application note: NAS Head Support.................................................................14 V-SWITCH 3000 STORAGE VIRTUALIZATION..................................................14 PERFORMANCE, HIGH AVAILABILITY AND LOAD BALANCING WITH A ROUTING SWITCH ........................................................................16 SCALABLE AND HIGH PERFORMANCE INTERNAL ARCHITECTURE......................20 The Turbo BUS and Scalability Port ..................................................................21 THE FINANCIAL BENEFIT ..............................................................................22 CONCLUSION ..............................................................................................24 Table of Figures FIGURE 1: SURVEY RESULTS CONDUCTED WITH 100 IT PROFESSIONALS BY COMPUTER WORLD IN 2000. ........................................................................6 FIGURE 2: COMPARISON OF FC, NAS AND STORAGE OVER IP (ISCSI) BY "PIPER JAFFREY" - OCT 2000..................................................................................8 FIGURE 3: SANRAD'S V-SWITCH 3000 TOPOLOGY .....................................................11 FIGURE 4: IP SAN TOPOLOGY USING SANRAD'S V-SWITCH 3000 AND THE SCALABILITY PORT ......................................................................................12 FIGURE 5: ISCSI HOST CONNECTION TO SANRAD'S V-SWITCH 3000..........................13 FIGURE 6: NAS HEAD CONNECTING TO SANRAD'S V-SWITCH 3000............................14 FIGURE 7. DOMAIN EXAMPLE.......................................................................................15 FIGURE 8. V-SWITCH 3000 PORTS..............................................................................16 FIGURE 9: EXAMPLE OF IP TAKE-OVER AND HIGH AVAILABILITY ....................................18 FIGURE 10: DIAGRAM OF A TYPICAL HIGH AVAILABILITY AND MULTI--PATH NETWORK CONFIGURATION USING THE V-SWITCH 3000 ...............................................19 FIGURE 11: TWO V-SWITCH 3000 CONNECTED THROUGH THE SCALABILITY PORT AND FORM A SINGLE SYSTEM WITH 6 GBIT ETHERNET PORTS AND 8 FC/SCSI PORTS........................................................................................21 List of Tables TABLE 1: STORAGE COST DIFFERENCES IN CENTS PER MB.........................................22 TABLE 2: COST DIFFERENCES AT DIFFERENT CAPACITY LEVELS...................................23 Untitled Document WP 001-04 Copyright SANRAD 2003 3 Introduction In the early 1990 s the "Internet" was complex, expensive and difficult to access. By 2001, with the use of new technologies and standards, the Internet has greater simplicity and is far more cost-effective than ever before. This has made the use and benefits of the Internet commonplace both at work and at home, with people using it as easily as they do the telephone. Supporting company networks, managing server applications, file systems and storage resources have been common challenges facing network and system administrators since the 1960 s. Administrators are looking for the same combination that made the Internet universal, to help them manage their ever growing networked storage. There is a demand for better storage management but technology, standards, simplicity and cost-efficiency have yet to be combined into a single answer. There are many storage management solutions available today but they lack one or more of the elements required for the majority of businesses. Despite this, the future looks bright. By mid 2002 technological advances will provide familiar, simple and cost-effective solutions. Based on existing and new standards, solutions will be made available that solve the storage management dilemma for all levels of business applications. This paper will review one of the leading solutions, the SANRAD V-Switch 3000, and compare it to earlier legacy products. We will review how the V-Switch 3000 provides technology, standards, simplicity, and cost-efficiency combined with scalability, flexibility, reliability and high performance. All these elements are included in a single platform addressing the need for better storage management. Simplicity is key and technology makes it possible. The Challenge facing Network Administrators Dealing with difficult to manage, diverse, and inaccessible storage resources. Most businesses have a combination of fibre-channel, SCSI, and/or NAS based storage. Storage systems are either directly attached to servers via SCSI or through fibre-channel. These systems participate natively or via a bridge in a fibre-channel based storage network, or are NAS systems attached directly to an Ethernet LAN. It s no surprise to find a couple or all of these configurations being used in small, medium, and large businesses. Storage configurations and management have been evolving for several decades. Administrators have made diverse storage decisions over the past years driven by business needs, budgets, and the technology of the time. This diversity of protocols, storage systems and physical connections has created a host of challenges for network and storage administrators. Downtime: Driven by the Internet and a global economy, 24 hr data availability is the norm. Fail-over, component redundancy, multiple-path connections, and data mirroring are all common requirements in the business-computing environment. Zero down-time is the goal but not always technically or fiscally achievable. Complex Administration: System complexity and diversity makes simple tasks difficult and time consuming. Which often leads to reduced network availability and degraded QoS. System and network administrators have limited personnel and fiscal resources, often leaving problems with quick fixes or left to solve another day. Untitled Document WP 001-04 Copyright SANRAD 2003 4 Inefficient Storage Utilization: It s a common occurrence that one application server is desperately in need of more storage while another server has three times more capacity than needed. Time and money are usually required to solve capacity problems when storage resources cannot be shared. Unauthorized Access: Many networked storage configurations leave storage resources thinly veiled from attackers. Weak access control inevitably leads to data integrity concerns, legal liability, and lost revenue. This large responsibility adds additional burden on network administrators. Poor Performance: Users are concerned with poor performance while network administrators are just trying to keep everything running with the limited available resources. It s a big challenge to find the time and resources needed to bring performance up while keeping complaints and QoS issues at a minimum. Lost Data: Accidents do happened- data can become corrupted or files lost. Data recovery tasks land on the network and system administrators. Users assume files are protected and easy to recover and rely heavily on thee network administrator to recover lost data. These challenges are even more acute when you consider that data is expanding annually at geometric rates while resources and budgets lag far behind. Achieving more with ever decreasing resources is not an impossible task, however it does require a shift in direction. Network and system administrations are eager to overcome their challenges and are more than willing to try a new solution if it will truly solve their problems. The problem for many IT professionals has not been the lack of storage management options. It s been a question of benefits vs. drawbacks Untitled Document WP 001-04 Copyright SANRAD 2003 5 Business Demands Networked Storage But How? Placing a network between the host file systems and stored blocks of data The common thread apparent in almost all small to large storage management solutions is the process of networking company storage. Whether it s a small NAS (network attached storage) solution with 50 gigabytes of capacity plugged into an Ethernet LAN, or a 24 TB enterprise class storage subsystem on a fibre-channel SAN - the reasons for going networked are basically the same. Detached storage, running on a network, is far easier and efficient to manage, and does a great deal to help overcome the many challenges facing network administrators. The question facing most IT professionals isn t whether to network - or not to network , the question is how to network? Network and system administrators are still dealing with their everyday challenges, and don t want the decision they made on how to network taking up more resources or creating new problems. A direct connection between a server and its storage, or DAS (direct attached storage), is simple. There is usually no command or stack translation, re-addressing of packets or frames, and concerns about connection latency involved with a direct attachment. However, business needs require that storage be networked, which introduces a network layer between servers and storage. Commands must now be standardized so they are more easily understood by all the elements within the network layer. WHAT SHOULD BE REVIEWED BEFORE DECIDING HOW TO NETWORK? 1. Storage network protocols and respective device driver elements encapsulate and send the SCSI storage commands and data across a network to the storage system. The storage system must understand the commands in order to return the appropriate response and data blocks to the host file system. Standards based, interoperability is important to help avoid difficult, expensive and unreliable installations. 2. The network type is very important because it affects performance, scalability, complexity, cost, legacy support, and reliability. Commands and data blocks will be sent across the network, between the hosts and storage elements. This is done via NICs (network interface cards) or HBAs (host bus adapters) and through switches, hubs and routers. This could involve a great deal of hardware and software and may become expensive and complex if you select the wrong network type. 3. Beyond the network type and protocol, vendor services and features are the last thing to review. There is a great deal being done to help network administrators meet their challenges. Centralized management, heterogeneous storage connectivity, secure access control, precise storage allocation, load balancing, - the list of features and benefits continue and can be realized, providing you select the right solution. Network type and protocol are generally tied together. FCP over Fibre-channel network and the new iSCSI over Ethernet network are the two combinations paving the way for increased networked storage. The following sections review both these options: Untitled Document WP 001-04 Copyright SANRAD 2003 6 Fibre-channel based SANs- No longer the only Option There are two predominant storage network communication protocols available for providing hosts with network connectivity to storage resources. The first is fibre-channel, which uses FC HBAs (host bus adapters) special device drivers, and requires special FC networking equipment. Fibre-channel is the dominant storage network type today and has been used in the majority of SAN installations. Fibre-channel storage networks have been adopted by businesses where the benefits out-weigh the drawbacks. But fibre-channel has not been widely adapted to solve the storage management dilemma facing all levels of business. Cost and complexity have out-weighted the benefits for many businesses, which have either adopted a simple NAS approach or are waiting for a solution where the benefits out-weight the drawbacks. BENEFITS OF FIBRE-CHANNEL " Well established and supported for specific business/storage applications. " Good overall performance and reliability DRAWBACKS OF FIBRE-CHANNEL " Expensive switch port and HBA costs. " Requires special training FC is not well understood by many network administrators " Limited to 10 km distance " Lack of standards are causing interoperability problems " Unique FC network type requires support separate from usual Ethernet network " Future FC performance roadmap is weak compared to new technologies Many of the current storage management solutions and SAN deployments are based on fibre-channel. This however, is because there were no other well-supported alternatives. FC based storage networks are being installed but at a very slow rate. The drawbacks are what have kept FC based storage networks to a minimum or completely out of the majority of businesses. Figure 1: Survey results conducted with 100 IT Professionals by Computer World in 2000. What's Holding Back (FC) SANs4%15%4%17%18%27%27%30%36%38%Don't knowNo deterrentsOther deterrentsInexperience IntegratorsLack of product interoperabilityDifficult Building a BCLack of standardsTechnology ImmaturityLack of staff / ResourcesHigh Implementation CostUntitled Document WP 001-04 Copyright SANRAD 2003 7 iSCSI Entry Level to Enterprise A New Option A new standard for networked storage is emerging. The IETF committee and industry giants worked together to create the new iSCSI standard storage protocol. iSCSI solves many of the problems causing the slow deployment of FC based storage networks. This new storage protocol resides on top of TCP / IP over Ethernet. Unlike fibre-channel, iSCSI does not require special equipment or cabling. It can use common NICs (network interface cards), Ethernet switches, hubs, and cabling. 1Gb Ethernet switches and NICs are common and in the same performance class as fibre-channel. An impressive 10x increase from existing 1Gb to 10Gb Ethernet will be available in the near future. Development of this performance level is well underway and supported heavily by the network and storage industry. FC performance scalability will not be able to deliver a 10x increase in a single iteration, but will require multiple iterations (2Gb, 4Gb, 10Gb), over time, to achieve the same performance as 10 Gb Ethernet. The business applications for iSCSI includes the majority of those currently provided by FC, but go far beyond FC, catering to the needs of a greater diversity of companies that require more efficient storage management. A QUICK NOTE ABOUT TCP/IP PROTOCOLS The Internet/network layer allows hosts to interject packets into any compatible network type and lets them travel independently to the destination IP address. The IP layer defines the packet and the protocol. It is a level 3 (network) and connectionless protocol. In a closed storage environment the ability for packets to find various routes is extremely beneficial in providing high levels of availability within a physically connected, multi-channel environment. Above IP is the TCP, which is a layer 4 protocol (transport). TCP allows peer entities on the storage and host to carry on a conversation. It is a connection-oriented protocol that allows a byte stream to proceed between a host and storage system. TCP insures that the byte stream is securely delivered in proper order and without error up to the iSCSI layer within storage or host. TCP also assists with flow control and packet resend requirements. A QUICK NOTE ABOUT ISCSI iSCSI transmits native SCSI commands and data over the TCP/IP protocol stack. iSCSI will transfer and store SCSI commands and data at any iSCSI enabled storage location with access to a LAN, MAN, WAN or the Internet. It will provide for the creation of high performance Gb Ethernet based SANs. Using iSCSI, standard SCSI commands and data are encapsulated as a serial string of bytes proceeded by an iSCSI header. This is passed to the TCP/IP layer that breaks it into packets suitable for transfer over the network, to the hosts and iSCSI enabled storage. If a read request for data has been sent, the data (blocks) are retrieved from the physical drives, encapsulated back into iSCSI and returned to the requesting host. The entire process is transparent to the user and file system. Untitled Document WP 001-04 Copyright SANRAD 2003 8 Most administrators are very comfortable and have the technical ability to deploy iSCSI storage networks because of their familiarity with Ethernet networks and TCP/IP. 1Gb Ethernet costs are dropping dramatically as Gb Ethernet replaces existing Ethernet network elements. The Benefits of iSCSI " iSCSI is based on IETF standards. (SANRAD is also a leading member of the IETF, which is responsible for the iSCSI standard) www.ietf.org/internet-drafts/draft-ietf-ips-iscsi-name-disc-02.txt " iSCSI deployment cost and storage system cost will be 1/2 to 1/10 that of FC " Can use existing Ethernet cabling and existing network elements " Uses common TCP/IP for global connectivity " Leverages the existing expertise of network administrators, integrators and support services " Is being quickly adopted by system, storage, and network vendors " iSCSI provides the same reliability features as FC. " iSCSI is in the same performance class as FC. " Future 10Gb Ethernet performance provides 10 times the performance of existing FC. " Leverage IP Security (e.g. IPSEC tunneling). " iSCSI can natively span over LAN , MAN and WAN the most obvious protocol for implementing disaster recovery solutions. Figure 2: Comparison of FC, NAS and storage over IP (iSCSI) by "Piper Jaffrey" - Oct 2000 Untitled Document WP 001-04 Copyright SANRAD 2003 9 Virtualization - Better Storage Management Attaching all your storage to a fast and reliable network provides the foundation for better storage management. There is a great deal being done to help network administrators meet their challenges. Centralized management, secure access control, precise storage allocation, increased performance, and easier data protection are just a few of the benefits that can be realized with better storage management. One popular technique for achieving these benefits and solving the storage management dilemma is virtualization. Virtualization gathers all physical storage resources into a single pool. These can include JODS, RAID systems, simple disk drives, tape libraries, optical jukeboxes, etc. From a central and simple interface, network administrators can administer common policies and services across the entire pool of storage. This is independent of the vendor brand, type, and protocol represented by each physically attached storage system. The network administrator can define new logical volumes from the pooled resources. Logical volumes can take almost any shape and size and are created independent of physical barriers such as enclosures, physical disks, protocols, and distance. The unique combined benefits of virtualized storage are clear- Less complex and easier to track Storage Pooling provides a single and simple storage interface to uniformly administer storage management tasks to all resources within the storage pool. It provides a simple point of control for all storage resource within small to enterprise class storage networks. Uses storage more efficiently Flexible storage configuration is accomplished by defining logical storage characteristics independent of physical limitations to better suite the needs of diverse business computing applications. Provides a trusted storage environment Host access control is enforced by the virtualization layer. Hosts no longer have access to all resources connected to the storage network, logical volumes are selectively allocated to a pre-defined host. Virtualization also helps to overcome the other challenges discussed earlier such a down-time, poor performance, and lost data. Combining virtualization with the right network type, protocol, and vendor technology gives system administrators what they need to finally overcome their longstanding challenges and truly do more with less. Untitled Document WP 001-04 Copyright SANRAD 2003 10 Different Methods of Virtualization There are several major approaches to achieving storage network management through virtualization. Today they are all based from fibre-channel. Host based approach: Relies heavily on software or agents installed on ever-networked hosts to provide adequate controls and administer the functionality of storage virtualization. Administrators must install software and agents on every host (assuming the host OS is supported by the virtualization software). This approach is very complex, not open to diverse host or storage environments and creates security issues. Storage based approach: The storage-based approach relies on the storage system to provide the virtualization functionality. This approach in not true storage network virtualization but is merely virtualization within the storage sub-systems supplied by one vendor. Virtualization functionality can only be delivered to same vendor/homogeneous enclosures within a FC network and they will preclude systems from other vendors, JBODs and other storage devices. Performance is usually assured but these systems tend to be homogeneous, very expensive to scale, difficult to manage centrally, and force the customer into vendor lock-in. Network Approach: Symmetric SAN Appliance: The virtualization control function resides on a dedicated appliance within the data-path. The appliance can be a switch/router. The appliance can also be a PC based host but use of a PC will not deliver the same performance as a switch/router. Network Approach: Asymmetric SAN Appliance: The virtualization control software resides on a dedicated meta data device out of the data-path. The access control list for volumes is stored within the meta data device. Each host must have host agent software installed in addition to receiving an access control list from the meta data server. This approach has many of the problems associated with a host-based approach (see above) in addition to the overhead created by the meta data device. Untitled Document WP 001-04 Copyright SANRAD 2003 11 The V-Switch 3000 - Combining the Best of all Methods Network Centric - iSCSI - FC - Virtualization - Switching - Routing - Protocol Conversion As stated at the beginning of this paper, Network administrators are looking for the same elements that made the Internet what it is today, in a storage management solution. The V-Switch 3000 provides standards, simplicity, and cost-efficiency accompanied by scalability, flexibility, reliability, security and high performance. All combined into a single platform to address the storage management challenges facing network administrators. Simplicity is key and technology makes it possible. " The V-Switch 3000 is a network centric, storage management and virtualization solution. It includes the V-Switch 3000 3000 and a browser based management GUI. " The V-Switch 3000 supports the new, standards based, iSCSI TCP/IP storage protocol. This provides iSCSI networked hosts with secure and trusted access to logical volumes. " The V-Switch 3000 operates in the data-path of a storage network, between the hosts and the storage, and is a high performance network component. . It provides a single, easy to use connectivity and management platform for IT professionals responsible for administering small to enterprise class storage resources. " The V-Switch 3000 storage management and virtualization features gather all physical storage resources into a single pool. Network administrators can define new logical volumes from these pooled resources independent of physical barriers such as enclosures, physical disks, protocols, and distance. " The V-Switch 3000 provides multi-protocol support for direct or network attached storage, as it can directly support fibre-channel, SCSI based, and iSCSI based storage systems all through the same switch. " The V-Switch 3000 is a convergence of functions, which include protocol bridging, routing, switching, security, load-balancing, high availability and volume management within a single easy to manage platform. Figure 3: SANRAD's V-Switch 3000 topology Disk systemDisk systemClientsDisk systemDisk systemDisk systemClientsiSCSI V SwitchDisk systemEthernetFile ServerDB ServerApp ServerDisk systemClientsiSCSI V SwitchDisk systemEthernetFile ServerDB ServerApp ServerUntitled Document WP 001-04 Copyright SANRAD 2003 12 V-Switch 3000 Host and Storage Connectivity Features " iSCSI Ethernet host network support (LAN, MAN, WAN and Internet) " Standards based iSCSI compatibility " Connectivity for all standard fibre-channel based SANs or individual FC storage systems by supporting Point, Loop and Fabric connections " Connectivity for all standard SCSI storage systems " Connectivity for all standard iSCSI storage systems Specifications of the V-Switch 3000 " Three host network ports " Up to four storage ports " Port traffic and user indicator LEDs " Easy to read configuration and test LCD with push button interface " CLI and GUI management interface " Ethernet/RS232 management ports " Removable CF card storing the configuration DB for a faster recovery. " SNMP and MIB support provided " Unique Scalability Port to connect multiple V-Switch 3000s that acts as a single director class storage switch. " Dual fault tolerant hot swappable power supplies (with a DC 48V option). " Redundant fans. EthernetClientsClientsVideo ServerApp ServerDB ServerDisk system 2Disk system 3Disk system 2Disk system 3Disk system 1Disk system 4Disk system 6Disk system 7Disk system 5Disk system 8Disk system 5Disk system 8Video ServerApp Server Figure 4: IP SAN topology using SANRAD's V-Switch 3000 and the scalability port Untitled Document WP 001-04 Copyright SANRAD 2003 13 ISCSI TCP/IP Hosts and the V-Switch 3000 The Switch is the Storage The V-Switch 3000 is as simple as TCP/IP. The V-Switch 3000 is network centric. It resides in the center of the network and is independent of host operating systems and storage system protocols. The switch does not require any agents or software to be loaded and maintained on the hosts and is not restricted to any specific storage sub-systems. It truly delivers an independent, network centric, storage management layer. iSCSI Host Support The V-Switch 3000 provides a storage virtualization layer. This layer represents multiple logical volumes. The network administrator uses the V-Switch 3000 to provide hosts with controlled access to these logical volumes. The V-Switch 3000 supports all iSCSI-enabled hosts attached to an Ethernet network. iSCSI transmits native SCSI commands and data over the TCP/IP protocol stack. Using iSCSI, standard SCSI commands and data are encapsulated as a serial string of bytes proceeded by an iSCSI header. This is passed to the TCP/IP layer that breaks it into packets suitable for transfer over the network to the V-Switch 3000. iSCSI is heavily supported: Microsoft, Linux, SUN (Solaris), HP, and IBM are only a few of the platforms supporting iSCSI. The packets are then sent over the network or the Internet and arrive at the V-Switch 3000. Here the packets are recombined into the original encapsulated SCSI commands and data. The V-Switch 3000 reads the iSCSI header to determine the logical location of the data blocks to be read or written. It than uses a mapping table to translate the logical location and commands to the actual physical location with corresponding commands. The new packet is than sent to one or more the SCSI, iSCSI or FC storage subsystems attached to the V-Switch 3000. The commands to write or read data are then executed at the storage system. If a read request for data has been sent, the data (blocks) are retrieved from the physical drives, encapsulated back into iSCSI and returned back over the network or Internet to the requesting host. The entire process is transparent to the user and file system. Figure 5: iSCSI Host connection to SANRAD's V-Switch 3000 The V-Switch 3000 has several host network ports and can simultaneously support tens to hundreds of iSCSI-enabled hosts. Untitled Document WP 001-04 Copyright SANRAD 2003 14 Application note: NAS Head Support Providing storage management to other NAS servers Other Ethernet NAS servers can also benefit from centrally managed storage. NAS systems are primarily file servers with their own file system and processing power within a single, easy to install enclosure, which attaches directly to an Ethernet network. Other NAS servers can benefit from more efficient storage management provided by the V-Switch 3000 if these NAS servers are physically independent of their storage. In this case the V-Switch 3000 can attach to the storage side of the NAS server and manage the storage requirements. NAS servers will be allocated storage as if the were another networked host. The storage allocated to the NAS server is part of the overall storage pool managed by the platform. Figure 6: NAS head connecting to SANRAD's V-Switch 3000 V-Switch 3000 Storage Virtualization Simple Administration, Less Things to Track, Efficient Storage Utilization Centralized Storage Management via Virtualization The V-Switch 3000 gathers all physical storage resources into a single pool. These can include JBODs, RAID systems, simple disk drives, tape libraries, optical jukeboxes, etc. From a central and simple interface, the network administrator can administer common policies and services across all the networked storage within the pool, independent of the vendor brand, type, and protocol represented by each physically attached storage system. Virtualization allows the network administrator to define new logical volumes from the pooled resources. Untitled Document WP 001-04 Copyright SANRAD 2003 15 Implementing Virtualization The V-Switch 3000 provides a browser based management consol. From this single interface, network and system administrators can use the V-Switch 3000 to administer all their storage virtualization and access control functions. When all the storage is attached to a single platform and all the hosts/servers are connected and communicating within the same platform you have the foundation needed to implement a single storage management layer. This single storage management layer is commonly known as the virtualization layer. At the start of configuring a virtualization layer the V-Switch 3000 conducts a discovery session that identifies and lists all the attached physical storage. This would be all SCSI, iSCSI and FC devices attached to the V-Switch 3000, either directly or through a common switch or hub, which could be the case for FC and iSCSI devices. All the storage system attributes are displayed in addition to addressing/ identification information such a WWN (world-wide names) and LUNs (logical unit numbers). The administrator can now see all the storage on one single interface. The V-Switch 3000 can now be used to transform this pool of physical storage into a pool of logical volumes. The physical system attributes such as RAID levels, system performance, and fault tolerance can obviously not be changed by the storage management platform. But volumes represented by each storage system can be redefined to better meet the needs of business applications. Once the pool of physical RAID systems, JBODs, and tape devices have been virtualized into new logical volumes, they can be allocated, assigned and delivered to the attached hosts. Figure 7. Domain example Untitled Document WP 001-04 Copyright SANRAD 2003 16 Performance, High Availability and Load Balancing with a Routing Switch Flexible High Performance and 24/365 Availability The V-Switch 3000 design provides a framework for high performance and reliability. It gives network administrators multiple host and storage ports. High availability and scaleable performance can be easily achieved with the V-Switch 3000. A single V-Switch 3000 has numerous ports on both the host and storage side. This provides network administration with good scalability within the switch and offers the flexibility to create a multitude of different configurations. Figure 8. V-Switch 3000 ports EACH INDIVIDUAL V-SWITCH 3000 WILL SUPPORT: " 3 iSCSI Gb Ethernet ports " 4 Ultra SCSI or FC ports " The Turbo BUS Scalability Port allows multiple V-Switch 3000s to connect together to form a single director class switch. Multiple connections can be created between a host and the V-Switch 3000. " Performance: The V-Switch 3000 can provide several IP addresses to a single host. The host can be configured to support additional connections to a V-Switch 3000 and therefore scale the overall throughput to satisfy performance requirements. " High Availability: Additional connections can be physically designed to provide multiple independent paths between the host and the V-Switch 3000. In the event a path should fail, other paths are still functional to insure continued storage access and data availability. " Load Balancing: Multiple connections between the host and the V-Switch 3000 provide automatic load balancing among all the available connections. The host system can be easily configured to always use the most available of it s connections to insure for the best possible performance. NETWORK PROCESSORSSPECIAL ASICsTurbo BUS3 iSCSI Gb Ethernet Switch to Switch HeartbeatMemoryRS232 and Ethernet service/mgt portsTurbo BUS Scalability PortNETWORK PROCESSORSSPECIAL ASICsTurbo BUS3 iSCSI Gb Ethernet Switch to Switch HeartbeatMemoryRS232 and Ethernet service/mgt portsTurbo BUS Scalability PortUntitled Document WP 001-04 Copyright SANRAD 2003 17 " Performance The V-Switch 3000 provides an algorithm for enabling equal or differential QoS among networked hosts. The V-Switch 3000s features a QoS based fairness algorithm. The switch can internally move host connections to more available ports and alter routing internally to less congested storage connections to provide equal or preferred quality of service and performance to different hosts accessing the same storage pool. The V-Switch 3000 has Routing and Switching Capability. " Performance: The V-Switch 3000 can read the address of an incoming packet or frame. It can re-address the packet and route it to the appropriate physical storage through one of its storage ports. The V-Switch 3000 can replicate a packet and send it through multiple storage ports in case mirroring is applied or spread single packets through a group of storage ports when striping is applied. This is done at the block level giving the V-Switch 3000 greater control over storage resources. Performance is greatly improved when spreading storage read and write request over server storage ports. " High Availability: Routing capability provides high-availability between the V-Switch 3000 and the attached storage. The V-Switch 3000 can very easily route communication to another physically connected port in the event the original connection between a port and storage is severed. " Load Balancing: The V-Switch 3000 has switching capability. Traffic between two ports within a switch can be redirected to another storage port or spread among several storage ports as long as those ports share a connection to the same physical storage and host network. " Remote Copy and Data Migration: Routing capability provide for remote copy and data migration when combining iSCSI and a WAN or the Internet. The V-Switch 3000 can re-address and re-route packets or frames between two or more widely distributed V-Switches that exist in two locations across a WAN or the Internet These functions are extremely beneficial, providing load balancing among different storage ports. Allowing commands and data to pass through multiple storage channels greatly increases the transfer rates between the V-Switch 3000 and storage systems. Combine this with mirroring and striping and it becomes clear why the V-Switch 3000 can provide a high number of transactions per second. Untitled Document WP 001-04 Copyright SANRAD 2003 18 The V-Switch 3000 IP take-over for Enhanced High-availability IP take-over is key to enabling highly available and fail-over paths with the V-Switch 3000. In the event the V-Switch 3000 is temporarily off-line or becomes overloaded, other V-Switch 3000s attached to the same storage and attached to the same host network can take over the IP addresses for the off-line switch. The hosts will proceed with communication through that new V-Switch 3000 until the original off-line V-Switch is brought back online or the connection paths are repaired. The IP addresses can now be restored to the returning switch and the original connections can be re-established. This provides continuous storage availability. Figure 9: Example of IP take-over and High Availability Untitled Document WP 001-04 Copyright SANRAD 2003 19 The V-Switch 3000 iSNS Service Performance and High Availability: The V-Switch 3000 will support iSNS (storage name service). iSNS provides additional IP addresses for specific volumes. This provides the iSCSI initiator within each host with additional locations for its storage. A host can query the iSNS within the V-Switch 3000, and once authenticated, can obtain additional addresses for the same volume. The host can create new connections between it and the new IP addresses. This feature provides a facility for high levels of availability cross network discovery process and increased performance via multiple connections. Figure 10: Diagram of a typical high availability and multi--path network configuration using the V-Switch 3000 Untitled Document WP 001-04 Copyright SANRAD 2003 20 Scalable and High Performance Internal Architecture Network Processors, ASICs, BUS Architectures All the functionality of the V-Switch 3000 is delivered at line-speeds. The V-Switch 3000 differs from current storage virtualization products because it was designed to be a high performance network component. Its design is based on the latest network processors, specially designed ASICs, and utilizes an ultra-wide-custom-made Turbo BUS architecture The V-Switch 3000 will have high transaction per minute performance The V-Switch will multiple the performance of PC-based solutions. The performance of the V-Switch 3000 dedicated hardware is obtained using Network Processors (which embed numerous execution units working in tandem), uniquely tailored ASIC that offload tasks from the CPU, and an ultra-wide-custom-made bus for memory access, I/O access and CPU access. Currently available general-purpose systems are unable to provide this level of performance. Network Processors and a Real Time OS The V-Switch 3000 uses the latest network processors and an ultra-wide-custom-made Turbo BUS architecture. This provides substantial performance for reading and writing to memory, managing communication through the 3 network ports (MACs), and for driving commands and data through the combination of 4 SCSI or FC storage ports. The processor processing capacity represents a quantum performance increases over X86 based systems. Additionally, the memory speed within the V-Switch 3000 is over 7 times that of standard server memory, read and write speeds. The V-Switch 3000 also uses a real-time operating system. It is not burdened by OS functions, which are usually present on desktop operating systems. It is dedicated to the tasks required and well integrated into the hardware architecture of the switch. ASICs One of the greatest latency problems associated with competing solutions is the inability to use additional hardware to perform the translation required between the layers of the protocol stacks. Performing this in software with standard X86 processors, already laboring over other tasks creates low performance and leads to saturation of the processing capacity. This can affect overall system performance and is sure to generate complaints from clients. The V-Switch 3000 will off-load tasks from the primary processor and dispatch routine tasks to be executed in the ASICs. This provides an extremely fast process for stack translation and eliminates un-necessary writes and reads to memory. Untitled Document WP 001-04 Copyright SANRAD 2003 21 The Turbo BUS and Scalability Port The V-Switch 3000 uses BUS architecture for supporting communication to the storage ports. The BUS speed can easily support 1Gbs full duplex FC communication to the 4 storage ports available on the switch. The BUS architecture also provides the ability to concatenate V-Switch 3000s. Multiple single switches can be connected via a BUS cable so they represent a single larger switch. For example: " Two V-Switch 3000s can represent an aggregate of 6 iSCSI Gb Ethernet ports with 8 FC or Ultra SCSI storage network ports (see Figure 11) " Four iSCSI Switches can represent up to 12 iSCSI ports with up to 16 storage network ports (see Figure 4). Figure 11: Two V-Switch 3000 connected through the scalability port and form a single system with 6 Gbit Ethernet ports and 8 FC/SCSI ports. NETWORK PROCESSORSSPECIAL ASICsTurbo BUSCombination of 8 FC or SCSI PortsMemoryRS232 and Ethernet service/mgt portsTurbo BUS Scalability PortSPECIAL ASICsTurbo BUSMemoryNETWORK PROCESSORSTurbo BUS Scalability Port6 iSCSI Gb EthernetSwitch to Switch HeartbeatNETWORK PROCESSORSSPECIAL ASICsTurbo BUSCombination of 8 FC or SCSI PortsMemoryRS232 and Ethernet service/mgt portsTurbo BUS Scalability PortSPECIAL ASICsTurbo BUSMemoryNETWORK PROCESSORSTurbo BUS Scalability Port6 iSCSI Gb EthernetSwitch to Switch HeartbeatUntitled Document WP 001-04 Copyright SANRAD 2003 22 The Financial Benefit The V-Switch 3000 is a significant technology. With greater technology comes simple installation and operation. The cost/ benefits of using the V-Switch 3000 are diverse. " Based on Ethernet and TCP/IP eliminates costs associated with training, additional staff, or outside professional services to create and manage a unique storage network. " Repurpose and provide an investment protection roadmap for existing FC and SCSI based storage systems " Decrease the requirement for additional Ethernet and FC switches. " Completely eliminate the need for protocol bridges and storage routers. " Use existing Ethernet network components. One of the major financial benefits associated with the use of the V-Switch 3000 is that network administrators will have the freedom to select simple cost effective storage devices. The complex management tasks that are performed by expensive and highly managed storage systems, such as EMC Symmetrix, IBM Shark and HDS Lightning, can now be performed by the V-Switch 3000. The benefits of high-end systems can now be achieved by combining simple, cost effective, reliable storage systems with the functions provided by the V-Switch 3000. Network administrators will realize significant cost savings and greater benefits by moving management functions out of expensive storage systems and into clusters of mid-range storage systems networked with the V-Switch 3000. Table 1: Storage Cost differences in cents per MB Gb CapacityAvrg. Cost per MB for Simple and Reliable Storage SystemTotal Capacity Cost for Simple and Reliable Storage SystemAvrg. Cost per MB of ISCSI V Swtich and Simple/Reliable StorageTotal Capacity Cost for ISCSI V Swtich and Simple/Reliable StorageAvrg. Cost per MB of Managed StorageTotal Capacity Cost for Managed Storage576 0.030 17,280 0.204 117,2800.250 144,0001152 0.030 34,560 0.117 134,5600.250 288,0001728 0.030 51,840 0.088 151,8400.250 432,0002304 0.030 69,120 0.073 169,1200.250 576,0004608 0.030 138,240 0.052 238,2400.250 1,152,0009216 0.030 276,480 0.041 376,4800.250 2,304,00013824 0.030 414,720 0.037 514,7200.250 3,456,00018432 0.030 552,960 0.039 712,9600.250 4,608,00023040 0.030 691,200 0.037 851,2000.250 5,760,000Cost Effective, Simple and Reliable Storage SystemsCost Effective Storage Systems Combined with and Being Managed by the iSCSI V SwitchExpensive and Managed Storage This table shows the cost differences in cents per MB between the three types of storage being discussed and the total cost for difference GB capacity levels. The three types of storage are: simple storage, simple storage combined with the V-Switch 3000, and expensive highly managed storage. Untitled Document WP 001-04 Copyright SANRAD 2003 23 Table 2: Cost differences at different capacity levels Storage Type Cost Comparison 17,280 51,840 69,120 138,240 276,480 414,720 552,960 691,200 2,304,000 3,456,000 4,608,000 5,760,000 34,560 117,280 851,200 712,960 514,720 376,480 238,240 169,120 134,560 151,840 576,000 432,000 288,000 144,000 1,152,000 0 1,000,000 2,000,000 3,000,000 4,000,000 5,000,000 6,000,000 7,000,000576 GB1152 GB1728 GB2304 GB4608 GB9216 GB13824 GB18432 GB23040 GBCapacityTotal Capacity Cost for Simple and Reliable Storage SystemTotal Capacity Cost for ISCSI V Swtich and Simple/Reliable StorageTotal Capacity Cost for Managed Storage The graph shows the cost differences at different capacity levels for simple storage, simple storage combined with the V-Switch 3000, and expensive highly managed storage. It clearly shows the cost savings when comparing highly managed storage with simple storage combined with V-Switch 3000. Untitled Document WP 001-04 Copyright SANRAD 2003 24 Conclusion The V-Switch 3000 is a network centric multi platform storage management solution independent of hosts and storage. It resides on the data-path and provides network administrators with a single logical view of all storage resources. The V-Switch 3000 is a convergence of functions that include protocol bridging, routing, switching, security, load balancing, snapshot, high availability and volume management within a single easy to manage platform. A standards based approach insures current and future interoperability, ease of installation, simple administration and maintenance. The V-Switch 3000 is a high performance network component delivering features at line speeds. A team of V-Switch 3000s can work together to provide easy to scale host networks and storage pools. The unique Scalability Port aggregates the connectivity and performance attributes of individual V-Switch 3000s to represent a single larger V-Switch 3000. Virtualization and volume access control provide network administrators with efficient storage utilization and assured security. The V-Switch 3000 solves the storage management dilemma for all levels of business applications. Technology, standards, simplicity and cost-efficiency for storage management are now available for network administrators. This combination will make the benefits of the V-Switch 3000 as simple to achieve as using the Internet. SANRAD US Tel: +1-510-521-2424 International Tel: +972-3-767-4800 info@SANRAD.com Copyright SANRAD 2003 All rights reserved. The copyright and all intellectual property rights in this article belong to SANRAD. It is strictly forbidden to copy, duplicate or otherwise use this article or any part thereof in any way shape or form without the prior written consent of SANRAD.