In the second part of our series we look at intelligent fabric devices from Brocade, Cisco, EMC, IBM, MaXXan, and others.
By Ashish Nadkarni
In the first part of this article (see InfoStor, July 2005, p. 34), we covered what (and what not) to consider when evaluating intelligent switches. If you have made the business case to introduce these devices into your environment, it is now time to examine the various alternatives. This article provides an overview of some of the major players/products in this space. While it is not possible to provide detailed technical specifications on each device, we have highlighted some of the key features that differentiate the products.
Brocade: SilkWorm Multiprotocol Router
Brocade inherited its Multiprotocol Router technology from its acquisition of Rhapsody Networks. In principle, the router functions in a very similar fashion to Cisco’s SSM/ASM/CSM modules (see below), but with one difference: It’s a stand-alone switch. However, what makes Brocade’s router unique is that it has one ASIC per port (e.g., for the 16-port switch there are 16 ASICs). Brocade claims that this allows each port to be programmed with a unique identity, which means that you can have 16 different applications running on the switch concurrently.
The switch can be configured in either a symmetric (embedded) or an asymmetric fashion. To make it work in an embedded way, vendors need to port their applications. To date, vendors such as Alacritus (now part of Network Appliance), Incipient, Veritas, and a few others have ported their virtualization engines to Brocade’s switch. However, this has presented a marketing challenge for Brocade since it traditionally sells its switches via storage vendors such as EMC and Hitachi Data Systems. These vendors are unlikely to sell a switch that performs some or most of the functions found in their arrays.
This left Brocade with only two options: to market the switch as a multi-protocol router or as an asymmetric virtualization switch.
The first option allows the router to be added to an existing (Brocade or third-party) SAN to provide services such as FCIP, iSCSI, FC routing, and FC-NAT. The second option, which is yet to bear fruit, is to use the switch for data-path processing in conjunction with a service processor from a vendor such as EMC or IBM. This is facilitated by an API that vendors can program their virtualization engines to act as control path processors. This communication occurs over an out-of-band IP network, but there are plans to migrate to in-band IP over Fibre Channel.
Brocade is also working on an intelligent fabric blade for its SilkWorm 24000 director. Currently, the SilkWorm Multiprotocol Router model AP7420 works as a stand-alone switch, but this blade will allow organizations with slot capacity in their Brocade directors to add intelligence to the network without making major changes.
While Brocade’s Multiprotocol Router makes a great addition to existing Brocade environments, it is a totally different switch than the rest of Brocade’s switches because the code base and features are different.
Cisco Systems: MDS series modules
Cisco enjoys a unique position in the industry: It is dominant in the IP networking space and with its MDS switches can seamlessly bridge the gap between Fibre Channel and IP. No other vendor can achieve this today.
In the intelligent fabric space, Cisco offers five different types of modules:
- IP Storage Services Module, or IP SSM;
- Multiprotocol Services Module;
- Caching Services Module, or CSM;
- Advanced Services Module, or ASM; and
- Storage Services Module, or SSM.
All of these are line cards that can be installed in any MDS switch or director. Each of the modules is specific to the software ported by third-party vendors.
The IP SSM currently provides services such as FCIP and iSCSI. FCIP is used for remote SAN extension, while iSCSI is used to map native Fibre Channel targets to iSCSI targets. The IP SSM also provides Internet Storage Name Server (iSNS) and iSCSI Network Boot Protocol (iNBP) facilities. In addition, the IP SSM can be configured to provide EtherChannel with multi-path load balancing, FCIP compression, FCIP Write Acceleration, and FCIP Tape Acceleration. It is available in 4- and 8-port configurations.
The Multiprotocol Services Module provides all the above services in addition to providing native Fibre Channel and FICON support.
Cisco’s ASM line card brings intelligence into play. It works in a symmetric (embedded) mode where the virtualization engine actually resides and functions on the card. This means that both control and data paths are in-band and there is no external processor required to provide any virtualization services. This also means that a software vendor that wants to make use of Cisco’s blade has to port to it. The ASM card is currently available with Veritas’ Storage Foundation for Networks software, which provides most of the volume management features that are found in Veritas’ host-based Volume Manager software, such as Dynamic Multi-pathing (DMP), online storage re-layout, data redundancy, FlashSnap, and online snapshots.
Cisco’s CSM card is used to provide virtualization services in conjunction with IBM’s SAN Volume Controller (SVC). This module is similar to the ASM except that it provides a cache-based clustered architecture that is unique to virtualization products from various vendors.
The CSM offers features (when used with IBM’s SVC software) such as LUN virtualization (up to 1,024 virtual disks on 64 hosts, 4,096 physical LUNs on 64 controllers, 128 controller ports, and 2TB of virtual disk), synchronous Peer-to-Peer Remote Copy (PPRC), point-in-time copy (FlashCopy), data migration, transparent virtual disk to physical LUN mapping, online re-layout of physical storage, and a multi-pathing software subsystem device driver.
With the 32-port SSM, Cisco is combining the functionality of all of the other four modules onto a single module, or line card. The SSM provides the following features:
- Fibre Channel ports for native Fibre Channel communication;
- Fibre Channel Write Acceleration (FC-WA) and SCSI flow-statistics monitoring. This technology allows synchronous replication over greater distances by minimizing latency;
- Network-Accelerated “serverless” Backup. Using the SCSI-2 Extended Copy command, this technology enables backup applications to use the network for data movement, thereby offloading I/O and processing from media servers;
- Network-Assisted Storage Applications with the Cisco SANTap protocol. SANTap is a proprietary protocol that allows a storage appliance to get an I/O copy without impacting the integrity, availability, and performance of the primary I/O between servers and storage. The SSM intercepts I/O on the network and performs duplication for the purposes of secondary data processing functions such as data replication, continuous data protection, and data migration. Thus network-assisted storage applications can be deployed without appliances residing in the primary data path. Some of the products that are taking this route are Alacritus’ Chronospan (now a part of Network Appliance), Kashya’s KBX5000, Topio’s TDPS Fabric Edition, and Xiotech’s TimeScale Replication Appliance; and
- Network-Hosted Storage Applications with the Fabric Application Interface Standard (FAIS)-based Intelligent Storage Application Programming Interface (ISAPI). Cisco also provides an API that allows the SSM to be used for data-path transactions only while control information flows through an external processor that the software vendor is responsible for maintaining. In this scenario, the SSM and external processor communicate either via an external IP network or in-band via IP over Fibre Channel. Some of the products that will make use of this facility are EMC’s Invista, Incipient’s Network Storage Platform, and Veritas’ Storage Foundation for Networks.
So what are the benefits offered by Cisco? From a business perspective, if your IT organization is a Cisco shop for IP services, going with Cisco lets you consolidate all your networking needs using products from a single vendor. Although it can be argued that this is vendor lock-in, there are a number of benefits. For example, all Cisco switches use the same chassis so environmental requirements remain consistent. Also, the management interface for the MDS products is very similar to that of other Cisco products. This allows the IP networking team to manage the SAN devices.
EMC: Invista “Storage Router”
EMC is a relatively late entrant to the fabric virtualization market. The company announced the Invista platform (formerly code-named “Storage Router”) in May.
The Invista router is an out-of-band solution that interacts with intelligent switches from Cisco, Brocade, and McData (although McData has not yet launched its intelligent platform).
Invista consists of an out-of-band appliance called a Control Path Processor (CPP). It will be clustered and interact with switches via an API for control data or slow path data. Fast path data will continue to go through the switch.
For more information on EMC’s Invista, see “EMC ships virtualization platform,” InfoStor, June 2005, p. 1.
IBM: SAN Volume Controller
IBM’s SVC is essentially a software solution that runs on an x86 Linux platform. Cisco’s CSM card works in conjunction with IBM’s SVC software. IBM also sells the SVC running on an appliance that can be plugged into any supported SAN configuration. The SVC can also run on an LPAR in a mainframe environment. Unlike EMC’s Invista, IBM’s SVC is an in-band symmetric virtualization solution.
One of the key differentiators of the SVC versus other approaches is that it is a cache-based solution. Most other approaches, including EMC’s, are pass-through solutions that don’t provide local caching mechanisms like those found in storage arrays. This means that these solutions at best can provide the same performance as storage arrays. IBM claims that cache at the virtualization layer improves performance. It also means that you can get away with using less cache on the disk array.
Earlier this year, IBM announced the sale of its 1,000th SVC. For IBM, the SVC provides a key business benefit: It enables the company to penetrate into environments that do not have IBM storage arrays and use its professional services arm to migrate non-IBM environments to IBM environments.
MaXXan: MXV series
MaXXan has two intelligent switch models. The MXV250 is a 5U rack-mount platform with 16 to 64 Fibre Channel ports. It comes bundled with FalconStor’s IPStor software suite and MaXXan’s SANCruiser management software. It also provides multiple Gigabit Ethernet ports for IP connectivity and provides storage virtualization, replication, snapshots, and NAS connectivity. The MXV500 is a director-class switch that has 16 to 256 ports in one chassis and up to 512 ports in a single fabric with a second chassis. Each chassis supports scaling up to eight application systems for storage services. In addition to the features offered by the MXV250, the MXV500 offers FCIP, IP trunking, LUN masking, and LUN re-mapping.
MaXXan’s intelligent switches are interesting for a couple of reasons: First, they are one of the few products to offer NAS in the fabric. Second, they offer virtualization via FalconStor’s IPStor suite.
Veritas: Storage Foundation for Networks
Historically, Veritas has provided host-centric solutions-e.g., Volume Manager, file system, and clustering software. Its Volume Manager, for example, provides advanced services such as multi-pathing (DMP), snapshots (FlashSnap), and replication (Volume Replicator). With intelligence moving to the network, it was logical for Veritas to port its Volume Manager functionality to network-based devices.
Veritas sells most of its products directly (e.g., not bundled via hardware vendors). Switches, on the other hand, are mostly sold via hardware vendors. This is where Veritas’ biggest challenge lies: How can/will the company penetrate into the network while maintaining a hardware-neutral stand? Companies such as EMC will not sell a Cisco switch, for example, with Veritas’ Storage Foundation for Networks software installed on it. This means that Veritas has to convince companies that they should buy their switches directly from Cisco and then layer Veritas software on top. Expect Veritas to port its software to other vendors’ switches, too.
From a business perspective, one challenge you will face is how to make sure your environment is supported end-to-end without the storage or switch vendor throwing up the “unsupported” flag. And when bugs do occur, where do you go? If Veritas claims the storage is at fault, will the storage vendor buy that argument?
A number of products provide virtualization and a variety of storage services, but do not strictly fall into the fabric-based intelligence category. Examples include Hitachi Data Systems’ TagmaStore Universal Storage Platform (USP) and Sun’s StorEdge 6920 system. These vendors have chosen to provide virtualization inside their disk arrays. On one hand, these arrays allow you to add disks and use them as a regular storage device. On the other hand, they allow you to continue to use your existing investments by virtualizing other vendors’ storage arrays.
HDS’s TagmaStore USP provides virtualization via an embedded controller, which is similar to some of the other products we have covered. The controller has the ability to virtualize disks and present them as virtual disks with HDS labels or non-HDS labels, with up to 32 private virtual storage machines. HDS has also announced a NAS blade for the TagmaStore USP.
In 2004, Sun acquired Pirus Networks. The result of this acquisition is that Sun now offers storage virtualization in its arrays-similar to HDS. Sun’s StorEdge 6920 is the first product from Sun to offer this type of functionality. Sun claims “heterogeneous storage virtualization,” but there is limited (or no) support for third-party storage arrays from vendors such as EMC and IBM.
While it is imperative that you do in-depth research so that you can select the right product for your environment, it is also important that you take your time. The chart (left) includes a features and functions checklist for intelligent fabric solutions, which may be helpful in making initial product comparisons. To view the table in its entirety, with feature comparisons, go to www.glasshouse.com/fabric.
Ashish Nadkarni is a senior consultant with GlassHouse Technologies (www.glasshouse.com) in Framingham, MA. He can be reached at firstname.lastname@example.org.