Richard P. King, Nagui Halim, Hector Garcia-Molina,and Christos A. Polyzois. Management of a remote backup copy for disaster recovery. ACM Transactions on Database Systems, 16(2):338--368, June 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....backup copy of the database. Media failure on the log is, as we will see in later sections of this paper, often overcome by maintaining a global consistent checkpoint to which the global database state can be recovered. In addition to these failures, recovery from a disaster is often considered [24]. While we do not consider disaster recovery in this paper, it can be dealt with by means of remote backups. Our main goal is to provide a survey of the state of the art with respect to checkpointing of distributed databases. We provide a categorization of distributed checkpointing including a ....

Richard P. King, Nagui Halim, Hector Garcia-Molina,and Christos A. Polyzois. Management of a remote backup copy for disaster recovery. ACM Transactions on Database Systems, 16(2):338--368, June 1991.

.... their data, and the heavy cost of a breach in service due to failures, there is a surge in the demand for systems that provide continuous service with absolutely no down time [8] An approach to providing continuous failure free operation is to maintain a backup (also referred to as hot standby) [5, 4, 6, 3, 10, 2]. In remote backup systems, two copies of the database are maintained. Transaction processing takes place at the primary and the log records generated are propagated to the backup which uses them to reconstruct a recent state of the database. In the event of a primary failure, the backup takes ....

....to replay these log records (similar to the way the backup executes the log record of the primary during normal processing) P i is ready to takeover transaction processing from the backup. 6 Previous Work Many algorithms for maintaining remote backup system have been developed in the literature [6, 3, 10], Quite a few of these can be used for implementing primary backup systems in multiprocessor environments of the kind discussed in this paper. However, most of these approaches have concentrated on exploring different techniques for log propagation, efficiently applying logs to the backup system, ....

[Article contains additional citation context not shown here]

H. Garcia-Molina, H. Halim, R. P. King, and C.A. Polyzois. Management of a remote backup copy for disaster recovery. ACM Transactions on Database Systems, 16(2):338--368, June 1991.

.... in service due to failures, there is a surge in the demand for systems that provide continuous service with absolutely no down time [Lyo90] An approach to providing continuous failure free operation is to maintain a remote backup system (also referred to as the hot standby) GMP94, GMP90, GMPH90, GMHKP91, MTO93, GM90] In remote backup systems, two copies of the database primary and the backup, are maintained. The backup copy is possibly located at a geographically remote computer to support continuous operation even in the presence of artificial and natural disasters (e.g. earthquakes, power ....

....reached their CDS. We refer to this requirement as the consistency requirement of the backup. The following section shows how the consistency requirement can be ensured. A similar problem also exists in 1 safe designs when multiple log streams are used. Two algorithms dependency reconstruction[GMHKP91] and the epoch[GMPH90] algorithm have been developed to ensure the updates are installed correctly at the backup for 1 safe processing. These algorithms could be potentially extended to address the consistency requirement of the o2 safe approach. However, due to their limitations, neither of the ....

[Article contains additional citation context not shown here]

H. Garcia-Molina, H. Halim, R. P. King, and C.A. Polyzois. Management of a remote backup copy for disaster recovery. ACM Transactions on Database Systems, 16(2):338--368, June 1991.

.... heavy cost of a breach in service due to failures, there is a surge in the demand for systems that provide continuous service with absolutely no down time [11] An approach to providing continuous failure free operation is to maintain a remote backup system (also referred to as the hot standby) [5, 4, 6, 3, 14, 2]. In remote backup systems, two copies of the database primary and the backup, are maintained. The backup copy is possibly located at a geographically remote computer to support continuous operation even in the presence of artificial and natural disasters (e.g. earthquakes, power outages, ....

....In such environments, partial failures occurs if one or more of the database partitions at the primary backup fail, but the system as a whole survives. Existing research either does not deal with partial failures as in [5, 6, 2] or identifies the problem but did not offer solutions (as in [3]) Exceptions to this include Tandem s RDF described in [11] in which a partial primary failure results in either the failed database partition to become unavailable causing transactions that access the failed partition to block, or a complete backup takeover which is costly and causes the system ....

[Article contains additional citation context not shown here]

H. Garcia-Molina, H. Halim, R. P. King, and C.A. Polyzois. Management of a remote backup copy for disaster recovery. ACM Transactions on Database Systems, 16(2):338--368, June 1991.

....[3] are early examples of the ROWO strategy. The basic idea is to allow the secondary copies to lag behind the primary in update, in exchange for faster reading time. Checkpointing schemes such as the Tandem Remote Duplicate Database Facility and backup schemes such as the Remote Backup Procedure [26] are 3 also examples of ROWO. In addition, all of the so called caching schemes are examples of ROWO, including quasi copies [2] These alternatives are not completely mutually exclusive and combinations are possible. For example, the Missing Writes protocol [16] uses ROWAA when all the copies ....

R.P. King, N. Halim, H. Garcia-Molina, and C. Polyzois. Management of a remote backup copy for disaster recovery. Transactions on Database Systems, 16(2):338--368, June 1991.

....with the secondary storage devices remotely located. This significantly complicates the problems associated with consistency and communication. A discussion of these problems, including the advantages and disadvantages of remote replication can be found in King, Halim, Garcia Molina and Polyzois [KHGMP91] and Mohan and Treiber [MT93] 2.6.2 Logging The database system must ensure that committed transaction s effects remain and aborted and active transaction s effects do not persist in the stable database. This normally means that when a transaction commits, all of the changes made to data items ....

R. P. King, N. Halim, H. Garcia-Molina, and C. A. Polyzois. Management of a remote backup copy for disaster recovery. ACM Transactions on Database Systems, pages 338--368, June 1991.

....that a participating site has failed. Although IYV with delegation of commitment employs two coordinators, it is different from disaster recovery protocols where a back up site records the effects of committed transactions and takes over only when it detects the failure of the primary site (e.g. [3]) 3.1 IYV with a Commit Coordinator Protocol (IYV WCC) As mentioned above, in this version of IYV, each unreliable coordinator is paired with a more reliable commit coordinator that is responsible for the commit processing of the transactions initiated at the unreliable site. All such pairings ....

H. Garcia-Molina, N. Halim, R. P. King and C. A. Polyzois, "Management of a Remote Backup Copy for Disaster Recovery," ACM TODS, Vol. 16, No. 2, pp. 338--368, 1991.

....it does not appear to work correctly when logical logging is used. Our crash recovery algorithms make no assumptions about lock granularity, commit numbers, or style of logging. In addition, they support partial failures and have parallel redo and undo phases when multiple DMs recover. King et al. [14] describe disaster recovery algorithms that send redo log records and transaction read sets from a primary database to a remote backup database. For each transaction, the primary database uses a two phase commit protocol to decide whether to commit or abort the transaction. If the transaction ....

R. P. King, N. Halim, H. Garcia-Molina, and C. A. Polyzois. Management of a remote backup copy for disaster recovery. ACM Trans. Database Syst., 16(2):338--368, June 1991.

....same order at the backup site as at the primary site. It is not 2 safe but only 1 safe. If the primary site fails and some of its messages regarding committed transactions are lost, the backup site is not aware of these committed transactions and they are not executed. RBP Remote Backup Procedure [19] The database system is assumed to consist of a primary site and a backup site. Each site is composed of many stores, among which the data are partitioned. At the primary site twophase commit is used, and each transaction is assigned a ticket upon its completion. The log is propagated to the ....

R.P. King, N. Halim, H. Garcia-Molina, C.A. Polyzois, "Management of a remote backup copy for disaster recovery", ACM-TODS, 16(2), June 1991, pp. 338--368.

....l For reliability even in extreme situations a two level replication scheme was developed. The first level of replication is between machines that are located in physical vicinity of each other. The second level is used between systems that are located at a large distance from each other. [KING91] provided much input to this work. The major new work here is to integrate the algorithms in [KING91] into a new two phase commit algorithm in an efficient manner. The proposed handling of replication and consistency within a system and between systems is very flexible and can sometimes even ....

....first level of replication is between machines that are located in physical vicinity of each other. The second level is used between systems that are located at a large distance from each other. KING91] provided much input to this work. The major new work here is to integrate the algorithms in [KING91] into a new two phase commit algorithm in an efficient manner. The proposed handling of replication and consistency within a system and between systems is very flexible and can sometimes even differ on an attribute level. l Another example which is covered in this thesis is the algorithm used for ....

[Article contains additional citation context not shown here]

R.P. King, N. Halim, H. Garcia-Molina, C.A. Polyzios, Management of a Remote Backup Copy for Disaster Recovery, ACM Trans. on Database Systems vol. 16, no. 2, Jun 1991, p. 338368.

....algorithms on the basis of a simulation. Section 6 summarizes the main ideas presented in the paper. 2 Related Work 2 Safe and traditional 1 safe algorithms are discussed in [GR93, Lyo88] A traditional 1 safe algorithm, called dependency reconstruction algorithm (DR) is described in [PGM92, KHGMP91] It can be used for bundle and stream model architectures. In the bundle model, all log records belonging to a transaction are bundled together and sent to the hot standby site. The stream model is similar to the one described in Section 3.1, where, each primary node sends the log records to its ....

....Minimum divergence means that no transactions other than those missing or depending on missing transactions should be aborted. 2 Safe and 1 safe hybrid schemes are discussed in [GMP90a, MTO93] The advantages and disadvantages of the 1 safe and 2 safe algorithms are discussed in [PGM92, KHGMP91] in detail. 1 Safe products are provided by: Tandem [Tan87] IBM [ibm95, MTO93] Oracle [ora95] Sybase [syb95] Informix [inf94] and other vendors. 3 Context 3.1 System Architecture P1 P2 Uniform site configuration Pn Symmetric site configuration Primary site Pn P1 P2 Hot standby site Site 1 ....

[Article contains additional citation context not shown here]

R.P. King, N. Halim, H. Garcia-Molina, and C.A. Polyzois. Management of Remote Backup Copy for Disaster Recovery. ACM Transactions on Database Systems, June 1991.

....that in [Eic86] log records of a transaction are marked after the transaction has committed, so that only log records of committed transactions would affect the BDB. It should also be mentioned that a logdriven approach is often used to manage remote backups for disaster recovery purposes (e.g. KGMHP88, Tan87] 10 Conclusions The increasing size of contemporary databases, and the availability of stable memory and very large physical memories are bound to impact the requirements from, and the design of recovery components. In particular, for checkpointing and restart processing, the ....

R. P. King, H. Garcia-Molina, N. Halim, and C. Polyzois. Management of a remote backup copy for disaster recovery. Technical Report CS-TR-198-88, Princeton University, Computer Science Department, 1988.

....of the backup snapshot. Traditional strategies often require human intervention to mount magnetic tape, run the backup, and store backup tapes in a coherent way. Recently, systems designed to better automate this process have been introduced by IBM [5, 17] and UniTree Corporation [18] King et al. [20] suggest that a remote system storing backed up data could provide quick recovery in the event of a crisis. We discuss differences between the InfoMonitor and backup systems in Section 2.2. Other work has focused on making the primary medium itself more reliable. Patterson et al. argue for ....

Richard P. King, Nagui Halim, Hector Garcia-Molina, and Christos A. Polyzois. Management of a remote backup copy for disaster recovery. TODS, 16(2):338--68, 1991.

No context found.

King, R. P., Halim, N., Garcia-Molina, H. Polyzois, C. A. Management of a remote backup copy for disaster recovery. ACM Trans. on Database Systems 16, 2 (June 1991) 338-368

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC