iSCSI make it possible to attach SCSI block storage Host system may connect to multiple SCSI buses. . caubracderfama.ml . Hufferd, J.L.: “ iSCSI: The Universal Storage Connection”, Addison-Wesley. ISCSI Universal Storage John Hufferd - Ebook download as PDF File .pdf), Text ISCSI: the universal storage connection / John L. iSCSI (Computer network. iSAN – A Storage Area Network made up of iSCSI connections. PDU - Protocol The iSCSI protocol enables universal access to storage devices and Storage.
|Language:||English, Indonesian, Portuguese|
|ePub File Size:||19.58 MB|
|PDF File Size:||14.28 MB|
|Distribution:||Free* [*Registration Required]|
iSCSI: The Universal Storage Connection The Copenhagen Connection Basics and Application of Fibre Channel SAN, NAS, iSCSI,InfiniBand and FCoE. The iSCSI protocol enables universal access to storage devices and session may contain one or more TCP connections and provides recovery in the event. iSCSI. The Universal Storage Connection by John L. Hufferd Addison-Wesley, ISBN: X. John Hufferd is a known expert on the iSCSI.
Micky Rodeh and Joseph Raviv, for their support of the whole project from the outset, their agreement to fund it, and the energy they spent convincing everyone that iSCSI is good business.
It contains discussions of the marketplace where appropriate and of some technology competitors, such as Fibre Channel.
How ever, mostly there will be positioning of the various technologies to em phasize their appropriate strengths.
Therefore, one needs to view all iSCSI capabilities and determine its applicability to the area in which the reader is interested. Many parts of the book are general enough that a thorough knowledge of SCSI is not needed. There are, however, more detailed parts of the book where SCSI knowledge would be helpful. I wrote this book to provide both the manager and the technician with a useful understanding of the technology. Product marketing and strategy professionals should also find the information useful and meaningful.
The book attempts to service divergent readers by ensuring that, when a chapter contains in-depth technical content, the reader will be so advised by a "To the Reader" commentary, which is flagged with the iSmiley face.
If appropriate, readers may skip forward to a topic or the Chapter Summary also marked with the iSmiley face for the highlights. Organization The book begins with a general background of the market and an answer to why iSCSI is of interest. A taxonomy of the various markets is given, along with an explanation of how iSCSI fits into them. This is followed by a short history of iSCSI so that the reader can get a sense of what propelled its development.
Next the book heads into the technology itself, with an overview that includes iSCSI layering. Various other key concepts are introduced here to ensure that the reader knows not only the importance of data integrity to storage technology, but also that new hardware is being introduced specifically to address bandwidth and latency issues.
A few pages are spent explaining the iSCSI naming conventions, because of their major significance to the use of the technology. Following the discussion of iSCSI naming conventions, the book takes the reader through the login process and the identification and option negotiation process. These processes are key in the establishment of a communication path between the host system and the storage controller. The process of sequencing the commands and data, as well as controlling the flow of commands and data, is reviewed.
This is perhaps the most tedious chapter, even though not deep in technical content. The various forms of task and error management are explained in a very technical discussion. The detail and technical depth build from that point to the end of the book.
Finally the reader is taken through the various companion technologies that iSCSI uses to complete its suite of capabilities. The main part of the book concludes with an explanation of what hardware vendors are doing to permit direct memory placement of iSCSI messages without additional main processor involvement. Of course, there is an overall summary that considers expectations for the future.
The message units are presented in alphabetical order for ease of reference. Appendix B contains a compact listing of the various negotiation keywords and values. It explains the logical model and its consequences. Finally, Appendix G contains the various reference sources, along with their Web page locators in most cases.
Speaking of references, bracketed citations, such as [SAM2], are fully referenced in this appendix. In iSCSI, serial numbers are always incremented and compared in what is called a bit serial number arithmetic. This is a way to determine, when serial numbers wrap around, if those numbers are greater or less than other serial numbers. Chapter 1. ATA drives are found mostly on desktop systems and laptops.
These systems may have several such SCSI buses attached to them. It can be considered a general interconnection technique that permits devices of many different types to interoperate with computer systems. See Figure Figure Small computer system interface SCSI. It defines how the SCSI device can be addressed, commanded to perform some operation, and give or take data to or from the host computing system. The operational commands are defined by a data structure called a command description block CDB.
For example, a read command would have a CDB that contained an "opcode" defined by the protocol to mean, "read. Generally, presenting the address on the hardware lines of the SCSI bus performs the addressing. This address technique calls out a particular SCSI device, which may then be subdivided into one or more logical units LUs.
An LU is an abstract concept that can represent various real objects such as tapes, printers, and scanners. Each LU is given an address. This is a simple number called the logical unit number LUN.
Host processors can have many SCSI buses. The next thing to consider is what happens when many computers are in the same location. If there are numerous disks LUs for each system, this configuration creates a very large grouping of storage units. Many installations group their servers and storage separately and put appropriate trained personnel in each area. These people are usually skilled in handling issues with either the computer system or the storage.
One of the most prevalent issues for the storage specialist is supplying the proper amount of storage to the appropriate systems. As systems are actually used, the amount of storage originally planned for them can varyeither too much or too little.
Taking storage from one system's SCSI bus and moving it to another system's SCSI bus can be a major disruptive problem often requiring booting of the various systems. Users want a pool of storage, which can be assigned in a nondisruptive manner to the servers as need requires. Another issue with the SCSI bus is that it has distance limitations varying from 1.
The bus type has to be matched with the requirements of the host and the SCSI storage devices often called storage controllers , which seriously limits the amount of pooling a SCSI bus can provide. Further, many SCSI bus storage devices can have no more than one bus connected to them, and unless high-end storage devices are used, one generally has at most two SCSI bus connections per storage device.
In that case the storage devices have at most two different host systems that might share the various LUs within the SCSI devices. Two hosts sharing one storage control unit. Often the critical host systems want a primary and a secondary connection to the storage devices so that they have an alternate path in case of connection or bus failure. This results in additional problems for systems that want alternate paths to the storage and, at the same time, share the storage controllers with other hosts which might be part of a failover-capable cluster.
Often an installation requires a cluster made up of more than two hosts, and it uses a process called file sharing via a shared file system e. Pooled storage via SCSI connections. The term "logical connection" is used because Fibre Channel FC components can be interconnected via hubs and switches. These interconnections make up a network and thus have many of the characteristics found in any network.
This is being rectified, but the administrator of an IP network cannot now, and probably never will be able to, use the same network management tools on an FC network that are used on an IP network.
This requires duplicate training cost for the FC network administrator and the IP network administrator. These costs are in addition to the costs associated with the actual storage management duties of the storage administrator. I have had many storage customers request that storage be set up on an IP network, for which they have trained personnel. This request comes from the fact that FC networking has not been taught in colleges and universities.
This is a very expensive burden that must be borne by the customer of FC equipment. The more storage shipped that is FC connected, the more ruthless the demand for trained personnel. Without universities providing trained graduates, companies will keep hiring people away from each other. Have versus want. Some people minimize this point and then go further and state that storage has different needs from other products located on a general IP network.
This is true; however, those needs are in addition to the management of the actual network fabric. Fibre Channel needed to invent general FC network fabric management as well as storage management. It is the fabric management that people have been wishing were the same for both storage and the general IP network. Universities have not been training students because of a combination of factors: 1.
Fibre Channel does not yet replace any other curriculum item. Storage interconnect is seen as a specialty area. Few instructors have expertise in storage and storage interconnects. Many university servers are not FC connected. The processors used by professors are not likely to be FC connected. That the main university servers are not Fibre Channel connected is a problem currently being addressed.
However, the professors' local systems, which have significant budget issues, will probably be the last to be updated. Fibre Channel. Readers not interested in history should at least read the Measurements section and then pick up with the Chapter Summary. The chapter will focus on the early days of iSCSI development. How did this technology stand up to Fibre Channel? Was there a future for the technology in areas where Fibre Channel would not go?
Was there a definable path to get to the technology promises? The researchers determined that if all they wanted to do was to operate on local networks. The bottom line was that the TCP overhead was not as great as feared.
There were more questions to answer. The researchers built prototypes on an NT base and on a Linux base. Several things needed to be measured and understood. The researchers debated the appropriate way to interface storage to common networks. IBM Watson Research. This was not a sufficient condition to set new directions within IBM development. As a result the work was divided up and decisions were made about what tests and proofs of concept to run.
By early the measurements had led to the creation of an IBM internal white paper. It also became very clear that. They said that a TOE on a chip was not possible or reasonable. During the second half of To get a handle on these questions. A number of measurements were taken in to get a basis for answering this question.
To ensure that they were measuring apples to apples. Where in the protocol stack is the processor time most impacted? These were critical questions. The client side of the operation is not shown.
The results can be seen in Figure That is. The measurements also determined that reducing the buffer-to-buffer moves would achieve some important performance improvements. One can extract two important pieces of information from the measurements.
Have the same Ethernet frame size. It is not.
Transfer the same amount of data. They also clearly showed. With backing from this group. They had an initial external draft ready by February Australia March These efforts were not accepted by the IETF ips workgroup. Subsequently a similar draft from Nishan Systems Corporation. At that meeting the draft was presented and a new workgroup was started.
The process moved the draft though several iterations until it was agreed that all outstanding issues had been resolved. This group was called IP Storage ips workgroup.
The goal was to enlarge the working team. Subsequent to forming the IETF ips workgroup. It should be noted that parallel efforts were under way within Adaptec and Nishan Systems. Most of the current FC chip vendors are shipping. Membership in the IETF ips workgroup has grown to over people representing over different companies.
IBM has stated their intention to make iSCSI part of their main line fabric and not just a connection to its i product. Intel is shipping its HBA. The performance analysis showed that. IBM's measurements were able to show that.
Many vendors are now shipping iSCSI products. Chapter 4. To the Reader The following text is at a fairly high level. A file system. An application programming interface API. Application-to-LU command flow. The SCSI processing flow is as follows: To understand how this is done.
The SCSI class driver invokes the appropriate hardware device driver. Sends instructions to the appropriate subordinate LU. Receives the data. The hardware device driver interacts with the HBA via a vendor-specific interface. Within the SCSI device. The section on iSCSI protocol layers further on discusses this process. We also need to understand that. The SCSI class driver gets its instructions from the application.
This solved many problems. The other arguments. This device will have an IP address. Other protocols. A PDU is designed to carry. This transport guarantees the error-free delivery of every byte of data. IP will determine the detail routing address. The iSCSI device driver. Assume that the initiator discovers the location of an iSCSI device. As will be seen later.
Let's say the above in a different way. These additional features will be described later. Protocol data units PDUs are the basic form of message unit exchange between hosts and storage controllers.
See the encapsulations shown in Figure The iSCSI layers. These protocols can follow the Ethernet standard or one of the OC optical connection standards. Figure illustrates the layering structure. IP looks to a set of link-level wire protocols to transport its packets across the network. As you can see. We will focus on the Ethernet links here. Sometimes this is done indirectly through a file system. Note that TCP looks at IP as its transport for sending packets of information to its remote counterpart.
To do that. It is the job of. As was shown in Figure It then passes the total Ethernet frame to the remote data link layer.
iSCSI: The Universal Storage Connection
When the SCSI layer needs data for the command e. In order to send data to a storage device LU. At the target. TCP and IP will each check and strip off their headers. As a solution each target vendor has created what the industry calls an initiator wedge driver to balance the workload over the multiple FC links. This means that storage vendors need to include their own code in the host system. Multiple connections between hosts and storage controllers. This can also be done for iSCSI.
To enable this. This is somewhat ironic. In order to prevent this confusion in the iSCSI space. In general this concept is simple and straightforward. The only problem with this approach is the lack of a SCSI-defined method for sending commands and data across multiple links. In Fibre Channel this is handled by the host system adding more sessions between the initiator and the target. There are several things.
This can cause a lot of operating system software conflicts. It allows commands and data to be transported across the different links connections and to arrive at the ultimate SCSI layer target in the same order as if they had been transported over a single connection. The more types of storage controllers the customer has. The iSCSI protocol defines how the commands and data can be spread over all of the session links yet to be delivered in order to the target SCSI device.
Complexity is somewhat eased. The other two extensions are digests. Notice that it is made up of several segments. It has a byte basic header segment BHS. These commands extended copy and compare have a "master CDB" that is contained in the BHS and may have multiple extensions each with source and target LU addresses along with descriptions of that part of the extended operation. Since third-party commands are rare compared to normal SCSI commands.
The copy manager is a SCSI entity that performs copy functions remote from the host when appropriately commanded via thirdparty commands from the host. As one can see in Figure General structure of the PDU. To understand the construction of the general PDU. These extensions describe the totality of the operation to a function called the copy manager.
One of those is the data segment. This BHS will be studied in more detail a little further on. Now we will make things a bit more complicated. Castagnioli et al. Basic header segment BHS layout. We can see that it has some flags discussed later. We have examined the relatively straightforward main path. This will be covered in more detail in Chapter 8.
For request PDUs. As mentioned in Chapters 1 and 2.
Figure shows the basic header in some detail. It may also contain. The reason for this CRC will be discussed later. Many vendors building them in order to increase performance and reduce cost have decided to embed their iSCSI processing with their TOE. A number of optional additional header segments AHSs. An optional data digest a CRC value. An optional header digest. An optional data segment.
In normal situations the reassembly buffer is not very large. This is especially true if the distances between endpoints are great say from Los Angeles to New York.
HBA vendors. This permits a very-low-cost HBA. This will enable the vendor to build a highly functional HBA but still limit the amount of additional RAM required on it. It is only when error processing causes retransmission that there needs to be a considerable amount of RAM to hold all the fragments that come in before the missing PDU header.
When the header arrives. The weakness of this is that it offers no protection for the packets as they pass through routers. Thereby it leaves the bit CRC able to accompany the data end to end. Since the packet gets a new Ethernet CRC as it is sent on to the target. The architects of iSCSI have determined that they cannot ensure the integrity of the data in installations that have less than perfect routers and switches. The usual approach to handling the issue of undetected error loss is to use better routers and switches.
The Ethernet links themselves are protected by a bit cyclic redundancy check CRC calculation that travels with the packet across the Ethernet links. This should not be viewed as a frequent problem. Before we panic. When I say a "weak detector of errors. In this way the probability of an undetected error is greatly reduced. This will be discussed further in Chapter It is also expected that for most laptops and desktops that have a software iSCSI implementation.
On the other hand. The CRC value is called a digest. With IPsec. Perhaps they have top-of-the-line equipment that does not need the extra protection. Naming in iSCSI eases the administrative burden somewhat.
The iqn names are formed by a different set of policies. The characters following the iqn. The most significant 24 bits of the EUI identity are the company id value. Examples of each are iqn. As a goal. Each EUI identity is unique in the world. The EUI value generated from these two numbers is hex acdeabcd.
The manufacturer chooses the least significant bit extension identifier. Both have a central naming authority that can ensure their uniqueness.
EUI identities are also used in the formation of FC worldwide names. Naming and Addressing In order to explain some of the following. Its format in an iSCSI name is eui. Because some companies have suborganizations that do not know what other parts of the organization are doing.
To avoid this duplication. The following are some examples also see Figure To be safe. The date should be when the root name was assigned to the companythat is. The point is to ensure that a later sale of the company or the name will not cause a conflict. The naming authority would pick an additional string of characters that make the initiator node name unique in the world.
The resulting two locations might both have an iqn name that looks like the one above. Then each suborganization can selfassign its iSCSI node names. Remember that the iqn name format is the reverse of the domain name format. The ajax group might. Subsequent administrators in the new company may not have a clue about what names were previously allocated and thus could create a naming conflict with subsequent allocations. In the case of a download or merger. To avoid this. If an IPv6 name is supplied.
The reverse name format ensures that the name is not confused with a real FQDN name. The iSCSI name is an iqn or eui name.
It is intended to look different and so is not resolvable to an IP address yet is unique in the world. Because the iqn is the reverse of the FQDN. While we are on the subject. This address can be specified as a URL uniform resource locator as follows: If an IPv4 name is supplied. Examples of fully qualified URLs. Anything in square brackets [ ] is optional. The domain name is either an IPv4 or an IPv6 address. As with normal mail. There are two types of node name: It is also possible to have a fully qualified domain name a host name instead of the resolved address.
This is usually the case at iSCSI gateways. Because of this unique identification. Naming and Addressing Summary The iqn name form can be made unique in the world. In spite of all that. The eui form of a node name is written in hex notation: The default iSCSI port number is currently With that integration. Ethernet CRC trailers.
We also discussed the concept of an iSCSI session. Connections can be logically separate on the same physical link or can be separate connections on different physical links. There are appropriate levels of encapsulationthe Ethernet frame. As for naming and addressing. Chapter 5. We will address session establishment in several ways: A general introduction to the login process in this chapter A much more detailed description of the login process also in this chapter Detailed descriptions of the login request and response PDUs.
After that. If you need a more in-depth understanding of iSCSI. I suggest you read the introduction to the login process. To the Reader This chapter provides a general introduction of the login process followed by a much more detailed discussion.
If you only want a high-level view. That response either accepts or rejects the login. As each connection is established. Note that the leading first connection is the one that establishes the session-wide values. There are two types of login sessions: The fullfeature phase is the normal mode. The connection may not perform any other function until the login is complete and the fullfeature phase is entered.
There will be one or more login requests. This process of sending and receiving login requests and responses is handled as if it were a single task. The discovery session will be addressed later. The login is composed of requests and responses.
Permit authentication of the iSCSI endpoints. Negotiate the parameters to be used by the iSCSI endpoints. The initial login must include in a field of the PDU called the DataSegment the login parameters in text request format. What follows is an example of login keywords that must be sent on the first command of every connection's login. The login process of the discovery session follows the same rules as for a normal login.
Login and login responses. As mentioned. These addresses and ports should be used in connecting to the nodes. The login process will go through the exchange of iSCSI operational parameters as well as security authentication processes. On the completion of the login. This format consists of a series of textual keywords followed by values. If such a device is located at the target IP address. The opposite side is supposed to pick the first option or value that it supports.
Then the initiator can begin sending SCSI commands to the target node. If desired. Session establishment is primarily identification of the remote site along with negotiations to establish which set of functions options will operate between the initiator and the target.
It can also be determined if the target can support the initiator shipping data in support of a SCSI write command. At that point either the connection is broken or the fullfeature phase of the session is established.
Even if the target does support the "unsolicited" arrival of data. Login has several phases discussed below. For our purposes at this time. To accomplish this. The login PDU will send. The techniques needed to determine what devices the host is permitted to contact. Even the order of the data's arrival in order only or out of order accepted can be negotiated between the initiator and the target. This process continues back and forth until both sides agree that they are through with the negotiation.
Technically the login does not need ITTs since only one login command can be outstanding at any one time on any specific connection. Login request PDU. This is the connection ID. The CID is used by the logout explicit and implicit function to identify a connection to terminate. This field accompanies all command PDUs so that responses can identify the command to which they are responding. Notice that the figure contains a field called CID.
Login PDUs The target. These request and response PDUs can be issued repeatedly. Portals carry an IP address and. These fields. The collection of portals on an initiator or a target is called a "portal group. All connections are between what iSCSI calls "portals. See the sections Keywords and the Login Process and Discovery Session in this chapter for additional descriptions of portals. Also within the login command are two sequence number fields which use sequence number arithmetic.
Normally the ExpStatSN is not specified except when a reconnection is issued. The target portal must be in the same portal group i. The target session identifying handle TSIH is set in the last target response during the leading login. The login command is an "immediate command.. See example in Figure The immediate. When the login for this subsequent connection is issued. When a subsequent connection login is started.
This is implied with the leading login. The version number fields versionMax and versionMin in the login PDU are used to ensure that the initiator and target can agree on the protocol level they support. When the initiator or target wishes to transit between its current phase and another phase. It is also used on any subsequent session connection to tie the connection to the base session. This topic will be revisited in Appendix A. Phase transition through login and login responses. The initiator session ID ISID is set by the initiator on the leading login to make the session unique within the initiator system.
The NSG next stage field will specify what phase is desired next. NSG is only valid when the T bit is set. The CSG current stage field will specify the phase that the session is currently in. Notice in Figure that the login response contains fields similar to those found in the login request. An example is the case when another connection is needed to replace a failing one so the new login must be executed immediately.
It also contains the StatSN status sequence number. Login response PDU. These fields are the T bit. The DataSegment contains login text requests. On the first login request of the leading login. For compatibility with normal command handling. On a leading login. The issues and approaches dealing with sequence numbers will be addressed in Chapters 7 and 8. One of the important things the target specifies in the last response to the leading login is to pick a TSIH and return it to the initiator.
During the login process both the initiator and the target will continue to reflect back to each other. Refer to the IETF specification for the current codes settings. Login Response Return Codes Description Login proceeding OK Requested target node moved temporarily to address provided Requested target node moved permanently to address provided Miscellaneous iSCSI initiator errors Initiator not authenticated Initiator not allowed access to target Requested target node not at this IP address Requested target node removed.
This "To-From" verification is the way all sessions begin and is part of the security phase of login. Based on this socket call. If the target name is appropriate. It is also possible for the user ID to have a one-to-many or a many-to-one or even a many-to-many relationship with any specific iSCSI initiator node name. The target first checks to be sure that the iSCSI target name. With the establishment of a working connection.
Depending on the iSCSI authentication mode chosen. This connection is established via a normal socket call. IKE will perform the appropriate certificate or key exchange and then return control to IPsec. IPsec determines the degree of security the initiator requires and the degree of security the target requires.
Let's first examine the single-connection session. The important. If the installation can avoid user IDs. This is because the administrator will not have to deal with managing the relationship between the ID and the iSCSI initiator names. These characters are encoded in UTF-8 Unicode. The key part of the pair must be represented exactly as shown in the iSCSI specification. Upper case and lower case are required in the key, as shown in Appendix B it is case sensitive , and there are no blank white space characters or nulls.
The numeric numbers can be represented by either decimals or hexadecimals. Hexadecimal numbers are indicated by a leading 0x set of characters. An example is 0xFAc2B. Very large bit strings can be represented in base encodings indicated by a leading 0b set of characters. See Appendix D.
Values should not exceed characters, unless expressly specified by the keyword writeup. This character limitation applies to the internal value, not the string used to encode the value in the text field. For example, the hex value expressed externally as 0x2ab2f will take up 4 bytes of internal representation, not For example, Xcom.
In this way, the other side will pick the first value in the list that it wants to support. The specification includes flag bit settings and the like.
However, only the general ideas will be presented here. For more details see Appendix A; the values themselves are shown in Appendix B. The quoted strings are comments that will not flow on the wire. The login normally starts in the security negotiation phase. It is expected that the target will decide whether to move into login operational negotiation phase or fullfeature phase or stay in the security negotiation phase. If the initiator does not want to negotiate security, it can initially set its CSG to login operational negotiation phase LONP , and its next stage to either the same, or full-feature phase FFP.
However, if the target is unwilling to operate without security, it may just end the session by returning a login reject response with an Authentication Error , and drop the connection. Perhaps a better approach for an initiator that wants to operate without security, but is willing to negotiate security if required by the target, is to set the CSG to security negotiation phase, but then just not send any security parameters. It will then be up to the target to offer its preferred set of security parameters and have the initiator make the appropriate selection.
The general rule for phase moves is that the initiator can request a transition from one phase to another whenever it is ready. However, a target can respond with a transition only after it is offered one by the initiator. Also, the target, via SRP, authenticates only the initiator. And at the end of parameter negotiations: At this point the login is complete and the session enters its full-feature phase, in which commands and data will be sent from the initiator to the target for execution by the appropriate LU.
You can see from the example that the login phase can be very chatty. In spite of this verboseness, the base concept is very simple. Since the sessions are very long lived, the overhead in this process is hardly noticeable. Now we will describe how these nonleading logins are processed in order to establish a secondary connection in a session. See SendTargets in the Discovery Session section to come. The new connection may originate either from a different physical network connection on the initiator or from the same physical connection.
In this way there can be parallel connections from the host to the storage controller. If all the connections in this set use different physical network connections, the total bandwidth of a session will equal the sum of the individual connections. A new connection within an existing session is started just like the original connectionthat is, via a socket call, which in turn causes the establishment of the IPsec coverage, and so forth. The initiator then sends a login PDU to the target.
All the other PDU fields and processing required as parts of the "leading connection" authentication are repeated for the secondary connection. Any values that were negotiated on the leading connection will apply unless reset by key negotiations on the secondary connection. All the keywords associated with the security phase are legal only during the login security phase. It is not a requirement that subsequent logins be on different physical network connections, either on the initiator or the target.
Though it seems strange, there are situations that make this reasonable.
Follow the Author
In this case the target connections have more bandwidth than initiator connections, so it makes sense to share a target's physical connection by letting the initiator log in several of its physical connections to the same target physical connection.
Another situation has to do with the initiator transmitting commands while simultaneously assembling a large data PDU. In this case the physical connection needs multiple logical connections so it is not slowed down by long data assembly processes from the initiator's main memory to the HBA. This is because other logical connections can be sending short commands during the long assembly process.
Others will have multiple physical network connections mounted on the same HBAs and will support secondary connections across those physical network connections. Still others will support secondary connections both on their multi physical connection HBAs, and across multiples of those HBAs. And for each physical connection there could be multiple logical connections. At this point it is appropriate to repeat the name given to the connection point the IP address and the TCP listening port on the target.
Some installations may decide that it is not important to secure discovery sessions. It should be understood that several different techniques can be deployed as part of discovery and the discovery session is only one of them. The other types of discovery are only companion processes that can be optionally implemented within iSCSI devices. A more extensive review of the alternatives is offered in Chapter The customer may not actually use these security functions on a discovery session if the vendor offers a way for the installation to explicitly disable them.
ISCSI Universal Storage John Hufferd
The iSCSI specification is silent on this issue. Discovery Session In this section we will cover the iSCSI discovery session and the information that can be obtained from it. I cover it extensively here because it is a basic part of the iSCSI protocol. That decision depends on the installation.
Like any iSCSI session. After entering the full-feature phase on the discovery session or on a normal session. The installation should bear in mind that security is always a set of fences. The discovery session is established like any other iSCSI session.
All may not be sent on normal sessions. In contrast. If an all parameter value is used in the text request. The implied target for the null request is the current target the one sustaining the normal session. More information on the discovery process is provided in Chapter It is of course possible to have a multiple-connection session to the same IP address.
When the SendTargets text request is issued on a normal session. A portal group is a collection of portals that can be used in a multipleconnection session. The process of establishing a discovery session was also defined. Also discussed was how to handle the exchange of parameters: A TPGT is a number attached to a target portal group that distinguishes the group from other target portal groups. The concepts of portal. Chapter 6. This is the same process and negotiating routine that is part of the login processeven the syntax is the same.
If you are not interested in the details of this process. It also has a technique for negotiating values between the two sides. See Chapter 5. Each new Text Request PDU sequence will have an ITT assigned to it that is unique across the session and which the target will return in its responses. These are: The exchange process can continue across several text request and response PDUs in a sequence.
In this way requests and responses are clearly coordinated. An example of this is the SendTargets command. The initiator initiates all requests. When the F bit is set. This means that the pairs may continue on a subsequent PDU. The SendTargets command is an example of a text request that may have a long text response. Upon receiving a TTT. The initiator will determine when the text communication is at an end.
Since there can be only one text request outstanding on a connection at any time.. Also 0x is the string symbol that leads a hex string. Each side is required to send an appropriate text request or response to the other side. Things may then continue as normal. The normal technique for completing a text interaction is that both sides have the F bit set. The following shows how that should be handled. The target is more limited in its ability to end a communication.
Whenever it seems the text response may solicit additional related comebacks.. In summary. The flow of messages is controlled by use of the F bit. They have various fields and character sets defined in Appendix D. They are encoded in UTF-8 Unicode and are case sensitive. Login Response. Text Request. They must not permit their key names to exceed 63 bytes. By default they have a maximum length of an individual value not its. No Hex-constant example: Each instance must of course be unique and separated by at least one null hex The constant None is used to indicate a missing function.
Rules for Keyword Value Negotiation 1. In literal list negotiation.. Values can be longer than bytes. Not proposing a key for negotiation is not equivalent to proposing the current or default value. A default remains a default only if not offered by either party. They can span text request or response boundaries i. If they are split between PDUs. The accepting party answers with the first value from the list it supports.
A value should be proposed if unsure of the other side's default. See Appendix B for details on this value. Each side keeps state during the negotiations. During that time. All negotiations start out stateless. Responses are required in all other Boolean cases.
This is based on the results function specific to the key and becomes the negotiation result. This makes the automatic outcome of the negotiation No and no response is required. This makes the automatic outcome of the negotiation Yes and no response is required. Selection of a value not admissible under the selection rules is considered a protocol error see the section Rules for Negotiation Failure to come.
All Boolean keywords have a result function. Nonnumeric strings are primarily used for declarations. If an acceptor is not supporting or not allowed to usewith a specific proposerany of the proposed options.
For Boolean negotiations keys taking the value Yes or No the accepting party responds with the required key and the chosen value.
See rule 12 and Appendix B for the keywords' appropriate Boolean result functions. The last value transmitted becomes the negotiation result. When the Boolean function is OR and the value received is Yes. For non-Boolean single-value negotiations: For numerical values the accepting party responds with the required key and the value it selects. Based on rule The initiator signals its intention to end the negotiation by setting the F bit final flag to 1.
If a specific key is not relevant to the current negotiation. For list value negotiation. When the initiator sends a text request that has the final flag set to 1: If the target has only one response. The constants None. Operational parameter negotiation may involve several request response exchanges. Rules for Negotiation Flow 1. The target sets the F bit to 1 on its last response.
If the target has more than one response. The initiatorif it has no more to negotiatemust keep sending the text requests even if empty with the F bit set to 1 until it gets the text response from the target. The acceptor. The value chosen by the acceptor becomes the result of the negotiation. When dependent parameters are sent within the same command. An initiator that has set the F bit to 1 in a request: Can be answered by the target with an F bit setting of 0.
No parameter should be negotiated more than once during any sequence without an intervening reset. Text request sequences are independent of each other. Targets must not submit parameters requiring an additional initiator text request when the target is responding with the F bit set to 1.
The values Reject. This works since the target is expecting its TTT back. See Appendix A. Whenever the target responds with the F bit set to 0. Whenever parameters' actions or acceptance are dependent on the actions of other parameters. Can be answered by the target with an F bit set to 1 indicating that the target's response sequence is completed. The operational parameters of the session or the connection will continue to be the values agreed upon during an earlier successful negotiation or will be the defaults.
During login. The initiator. If the initiator detects the failure. The login phase and its connection must be terminated. If multiple negotiations of the same parameter. Rules for Negotiation Failure 1. If the target detects the failure. Any partial results of this unsuccessful negotiation must be undone. We also covered Rules for negotiating the various values a parameter might use.
We explained how the text requests and responses could be used as a commandresponse sequence. Individual keywords are described in Appendix B. Terminate the text command negotiation and reset. Rules governing the handling of multiple response sequences for example.
Negotiation failure rules: Terminate the login.This results in additional problems for systems that want alternate paths to the storage and. In this case the target connections have more bandwidth than initiator connections, so it makes sense to share a target's physical connection by letting the initiator log in several of its physical connections to the same target physical connection.
The researchers determined that if all they wanted to do was to operate on local networks. Non enterprise storage controllers usually have only one or two SCSI bus connections.
The clear trend in the industry is to tie applications to databases. Tape libraries. The existing wiring infrastructure that most companies have is Category 5 Cat. A few pages are spent explaining the iSCSI naming conventions, because of their major significance to the use of the technology. This is the primary location for FC devices.
To understand the construction of the general PDU.
- POLITICAL THEORY EDDY ASIRVATHAM PDF
- THEVARAM IN EBOOK
- THE FOUND BOAT ALICE MUNRO PDF
- STAR WARS THE FORCE AWAKENS BOOK
- TOUCHING THE VOID JOE SIMPSON PDF
- MATHEMATICS FOR ECONOMICS BOOK
- EDITABLE PDF FORM
- PDF ER FREE ONLINE
- KAPSBERGER PDF
- 2015 KALNIRNAY CALENDAR (PDF) IN HINDI LANGUAGE
- SPECIFICATION BY EXAMPLE EBOOK
- APLIKASI PDF UNTUK HP BLACKBERRY
- SOMETHING WONDERFUL JUDITH MCNAUGHT PDF