Keeton, K., Patterson, D. A., and Hellerstein, J. M. (1998). A Case for Intelligent Disks (IDISKs). SIGMOD Record, 27(3).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....systems using conventional workloads. Several orthogonal issues in the design of large storage systems have also been studied in the literature. Techniques for characterizing I O workloads have been studied in [12, 14] Techniques for making storage devices more intelligent have been studied in [7, 13]. Techniques for load balancing and handling workload skews in disk arrays using chained declustering have been studied in [11] Randomized data placement techniques have been compared to traditional striping techniques in [20] We note that our results for wide striping and narrow striping hold ....

K. Keeton, D A. Patterson, and J. Hellerstein. The Case for Intelligent Disks (IDisks). In Proceedings of the 24th Conference on Very LArge Databases (VLDB), August 1998.

....and disks in NASD support writes and attribute read requests. This is the price that SPIRAL pays in order to achieve client side transparency and to minimize changes to the NAD OS. Some other approaches proposed to offload computation to smart storage devices include active disks and IDISK [1, 21, 14]. The derived virtual device (DVD) model [18] proposed in the Netstation project provides a mechanism for safe shared device access in an untrusted environment by creating DVDs and managing them through a network virtual device manager. The proposed third party transfer scheme using DVDs is ....

K. Keeton, D. A. Patterson, and J. M. Hellerstein. The case for intelligent disks (IDISKs). SIGMOD Record, 27(3):42-- 51, September 1998.

....flexibility. For example, the River system [2] exploits this flexibility to provide robust stream based access to large data sets. For individual accesses, dynamic set interfaces [28] can provide similar flexibility. The approaches described here may fit under such interfaces. Active disk systems [1, 15, 24] enhance storage systems with the ability to perform some application level functions. Data filtering is one of the more valuable uses of active disk capabilities, and the project database function is a column based access pattern. Thus, such a use of active disk interfaces would allow the ....

Kimberly Keeton, David A. Patterson, and Joseph M. Hellerstein. A case for intelligent disks (IDISKs). SIGMOD Record, 27(3):42--52, September 1998.

....to clients [16] The network server load is reduced by splitting the file system software between the file manager and the networkattached drives. Also, several studies propose active disk systems that run application level code on disk drives in order to reduce data tra#c and improve parallelism [1, 20, 27]. Related experiments show significant speedup with data mining and image processing applications. However, all these studies are fundamentally di#erent from ours, because they advocate running file system or application code directly on the disk drive. The performance and services of file server ....

Keeton, K., Patterson, D. A., and Hellerstein, J. M. A Case for Intelligent Disks (IDISKS). Sigmod Record 27, 3 (Sept. 1998), 42--52.

....grows, designers are forced to develop techniques to leverage the negative effects of this gap. One attractive solution for this problem is to distribute intelligence to storage structures such as memories and secondary storage devices (e.g. Active Memories [4, 19, 28, 32] and Active Disks [2, 17, 23, 30]) The common goal of these techniques is to offioad parts of the processing to the off chip structures such that the amount of communication between the storage and processing elements is minimized. As the VLSI technology advances, the gap between the performance of the execution cores and the ....

Keeton, K., D. A. Patterson, and J. M. Hellerstein, A case for intelligent disks' (IDISKs). SIGMOD Record, 1998.27(3): p. 42-52.

....as is the case for these database engines. Further, in a cluster with hardware capabilities like a storage area network or NASD, it would be useful to incorporate intelligence at the disk so that these operations can be carried out to reduce transfer traffic as pointed out by others [11, 13, 15]. 8 Summary of Results and Concluding Remarks This is the first study to embark on a detailed characterization and to present a range of performance statistics for the execution of TPC H queries on a medium sized Linux cluster of SMP nodes (a popular configuration in today s commercial market) ....

K. Keeton, D. A. Patterson, and J. M. Hellerstein. A Case for Intelligent Disks (IDISKs). SIGMOD Record, 27(3):42--52, September 1998.

....advantage of S4 over previous versioning systems is that it has been partitioned from the operating system. While this creates another level of indirection, it adds to the survivability of the storage. S4 s device embedded storage management is another instance of many recent smart disk systems [1, 7, 15, 27, 36]. All of these exploit the increasing computation power of such devices. Some also put these devices on networks and exploit an object based interface. There is now an ANSI X3T10 (SCSI) working group looking to create a new standard for object based storage devices. The S4 interface is similar to ....

Kimberly Keeton, David A. Patterson, and Joseph M. Hellerstein. A case for intelligent disks (IDISKs). SIGMOD Record, 27(3):42-52, September 1998.

....amount of extra cost over the disk cost due to the low costs of embedded processors and memory chips. Although this new type of smart disk based architecture seems very attractive, it poses many challenges. Previous work in this area focuses on the architectural and operating system related issues [25, 34, 44]. Acharya et al. 1] on the other hand, focus on the implementation of We use the term smart disks to refer to a class of architectures that put substantial computational power on disks, such as Active Disks [1, 34, 42] and IDISKs [25] the individual database operations. Smart disks seems to ....

.... architectural and operating system related issues [25, 34, 44] Acharya et al. 1] on the other hand, focus on the implementation of We use the term smart disks to refer to a class of architectures that put substantial computational power on disks, such as Active Disks [1, 34, 42] and IDISKs [25]. the individual database operations. Smart disks seems to be an attractive alternative especially for database applications. Therefore, investigating the individual database operations on this architecture is very important. But, to gain more insight on the possible improvements by the smart ....

[Article contains additional citation context not shown here]

K. Keeton, D. A. Patterson, and J. M. Hellerstein. The case for intelligent disks (IDISKs). In SIGMOD Record, 27(3), 1998.

....the disk cost due to the low costs of embedded processors and memory chips. Although this new type of smart disk based architecture seems very We use the term smart disks to refer to a class of architectures that put substantial computational power on disks, such as Active Disks [1] and IDISKs [25]. attractive, it poses many challenges. Previous work in this area focuses on the architectural and operating system related issues [25, 34, 44] Acharya et al. 1] on the other hand, focus on the implementation of the individual database operations. Although investigating the individual ....

....very We use the term smart disks to refer to a class of architectures that put substantial computational power on disks, such as Active Disks [1] and IDISKs [25] attractive, it poses many challenges. Previous work in this area focuses on the architectural and operating system related issues [25, 34, 44]. Acharya et al. 1] on the other hand, focus on the implementation of the individual database operations. Although investigating the individual database operations on this architecture is very important, to gain more insight on the possible improvements by the smart disk architecture, we have to ....

[Article contains additional citation context not shown here]

K. Keeton, D. A. Patterson, and J. M. Hellerstein. The case for intelligent disks (IDISKs). In SIGMOD Record, 27(3), 1998.

....policies. Thus, Proboscises allow the disks of a cluster to be used in the same way as disks on a SAN or network attached disks [24, 14, 28] but our approach is easier to customize, as the control software runs on regular workstations. In this respect, the Proboscises resemble programmable disks [1, 18] but without requiring special purpose hardware. Thus, the flexibility of the Proboscis framework allows the Proboscises to be used as building blocks for more advanced shared storage systems for clusters. The rest of this paper is organized as follows. Section 2 describes the Proboscis ....

K. Keeton, D. Patterson, and J. Hellerstein. A case for intelligent disks (IDISKs). SIGMOD Record, 27(3):42--53, August 1998.

....amount of extra cost over the disk cost due to the low costs of embedded processors and memory chips. Although this new type of smart disk based architecture seems very attractive, it poses many challenges. Previous work in this area focuses on the architectural and operating system related issues [25, 34, 44]. Acharya et al. 1] on the other hand, focus on the implementation of the individual database operations. Smart disks seems to be an attractive alternative especially for database applications. Therefore, investigating the individual database operations on this architecture is very important. ....

.... is traversed and only the relevant tuples are brought to the processing unit, the improvements achieved by the smart disk based system We use the term smart disks to refer to a class of architectures that put substantial computational power on disks, such as Active Disks [1, 34, 42] and IDISKs [25]. may be lower because only the relevant data will be transfered from the storage unit to the processing element. In this paper, we present a detailed quantitative evaluation of a smart disk based architecture. To achieve this, we compare the performances of a smart disk system, two types of ....

[Article contains additional citation context not shown here]

K. Keeton, D. A. Patterson, and J. M. Hellerstein. The case for intelligent disks (IDISKs). In SIGMOD Record, 27(3), 1998.

....from more expressive storage interfaces. For example, some have proposed to dynamically place data near the disk head [12, 54] or to piggyback write backs on rotational delays [4] Perhaps the clearest examples are recent proposals for object based storage [19, 18, 35] and those for Active disks [1, 41, 25]. Object based storage defines a new storage interface that is much like a file system with a flat namespace; this does make the interface somewhat more expressive, but not nearly enough in our view. Active disks would allow application code to be downloaded and executed inside disk fLrmware; this ....

....employ for their internal scheduling decisions, which are based on overall positioning overheads (seek time plus rotational latency) 49, 24] However, this may require that freeblock scheduling decisions be made by disk firmware. Fommately, the increasing processing capabilities of disk drives [1, 18, 25, 41] make advanced on drive storage management feasible [18, 54] 5.3 Applications of Free Bandwidth Freeblock scheduling is a new tool, and we expect that system designers will find many unanticipated uses for it. This section describes some of the applications we see for its use. 13 Scanning ....

[Article contains additional citation context not shown here]

Kimberly Keeton, David A. Patterson, and Joseph M. Hellerstein. A case for intelligent disks (IDISKs). SIGMOD Record, 27 (3): 42-52, September 1998.

....in three dimensions. The scalability issue becomes even more crucial for higher dimensions and for high performance systems that use hundreds or even thousands The implementation is optimized with a dynamic programming technique that evaluates all queries with an M grid. of parallel disks [22, 1, 30]. Description and Comparison to schemes with (limited) analytical guarantees Subsequent to the rst publication [6] of the GRS scheme described in this paper, a number of schemes with analytical guarantees have been proposed. Atallah and Prabhakar [3] gave a COLORING scheme whenever M is a power ....

K. Keeton, D.A. Patterson, and J.M. Hellerstein. A case for intelligent disks (idisks). SIGMOD Record, 27(3), 1998.

....be consumed. In fact, please note that products for servers supporting large numbers of disks are currently constrained by backplane bus limitations, which is many times smaller the combined raw disk bandwidth [www98,www97] This fact has been recognized by many storage system researchers [RGF98,KP98] and is a key problem on which they focus. In this framework, and using the Conventional Play method, the TS is used for storage augmentation purposes, simply to store the large number of colder videos and elevate them to SS when a request for them arrives. Since typically the TS bandwidth is ....

K. Keeton and D. Patterson. A case for intelligent disks (idisks). Sigmod Rec., 27(3),

.... devices, such as smart disks 1 with embedded processors, ASIC, and the capability to process and filter data close to where it resides have the potential to address 1 We use the term smart disks to refer to a class of architectures that put substantial computational power on disks, such as [1] [17] and [27] this issue by reducing data movement, and thereby, reducing bandwidth requirements. This is based on the premise that in the near future 100 GBytes disks with embedded processors with hundreds of MIPS and a few hundred Mbytes of RAM will be available [13] Therefore, it is attractive ....

....consider solutions where computations can be moved closer to data to take advantage of these features. However, such approaches introduce many new challenges and problems that must be addressed. Previous work on this area addresses the architecture design and operating system design related issues [27,32,17]. Acharya et al. 1] on the other hand, focus on the implementation of individual database operations such as selection and join operators and of some scientific kernels on smart disk architectures. However, in order to gain further insight to the capabilities of such architectures, we need to ....

[Article contains additional citation context not shown here]

K. Keeton, D. A. Patterson, and J. M. Hellerstein. The case for intelligent disks (IDISKs). In SIGMOD Record, 27(3), 1998.

....advantage of S4 over such systems as these is that it has been partitioned from the operating system. While this creates another layer of abstraction, it adds to the survivability of the storage. S4 s device embedded storage management is another instance of many recent smart disk systems [1, 3, 11, 17, 22]. All of these exploit the increasing computational power of such devices. Some also put these devices on networks and exploit an object based interface. There is now an ANSI X3T10 (SCSI) working group attempting to create a new standard for object based storage 18 devices. The S4 interface is ....

Kimberly Keeton, David A. Patterson, and Joseph M. Hellerstein. A case for intelligent disks (IDISKs). SIGMOD Record, 27(3):42-52, September 1998.

.... (i.e. in parallel disk seek operations) Meanwhile, emerging technologies in storage area network (e.g. Fibre Channel Arbitrated Loop, switch based I O bus, and Gigabit Ethernet) have also enabled one to build a massively parallel storage system that contains hundreds or even thousands of disks [1, 19]. As the number of disks increases, the ecacy of the adopted declustering scheme becomes even crucial. Many applications that adopt declustering schemes have to deal with multidimensional data. These applications include, for example, GIS [22] remote sensing databases [8, 9] parallel search ....

K. Keeton, D.A. Patterson, and J.M. Hellerstein. A case for intelligent disks (idisks). SIGMOD Record, 27(3), 1998.

....into two categories: application processing in Active Disks and system level processing in Active Disks. Riedel et al. 52] investigated the performance of Active Disks for data mining and multimedia algorithms and developed an analytical performance model. The ISTORE project [15] uses the IDISK [40] architecture as a building block to create a meta appliance; a storage infrastructure that can be tailored for specific applications. Acharya et al. 2] introduced a stream based programming model for disklets and their interaction with host resident peers. Restructured versions of a wide range ....

K. Keeton, D. A. Patterson, and J. M. Hellerstein. A case for intelligent disks (IDISKS). ACM SIGMOD Record, 27(3):42--52, Sept. 1998.

....to collect this type of information by circulating a message locally to gather statistics, because each disk has intelligence. There are several other academic research projects to utilize a disk resident processor and memory for executing application programs: the IDISK project at UC Berkeley [9] and the Active Disk projects at Carnegie Mellon [10] and UC Santa Barbara Maryland [11] They focus on the functions and mechanisms for making a combination of the disk resident processor and a host execute storage centric user applications, such as decision support systems, relational database ....

Kimberly Keeton, David A. Patterson, and Joseph M. Hellerstein. A Case for Intelligent Disks (IDISKs). SIGMOD Record, 27(3):42--52, Sep. 1998.

No context found.

Keeton, K., Patterson, D. A., and Hellerstein, J. M. (1998). A Case for Intelligent Disks (IDISKs). SIGMOD Record, 27(3).

No context found.

K. Keeton, D. Patterson, and J. Hellerstein. A case for intelligent disks (IDISKs). SIGMOD Record, 27(3), 1998.

No context found.

Kimberly Keeton, David A. Patterson, and Joseph M. Hellerstein. A case for intelligent disks (IDISKs). ACM SIGMOD Record, 27(3):42--52, September 1998.

No context found.

Kimberly Keeton, David A. Patterson, and Joseph M. Hellerstein. A case for intelligent disks (IDISKs). SIGMOD Record, 27(3):42--52, September 1998.

No context found.

K. Keeton, D. A. Patterson, and J. M. Hellerstein. A case for intelligent disks (IDISKs). SIGMOD Record, 27(3):42-52, September 1998.

No context found.

K. Keeton, D. A. Patterson, and J. M. Hellerstein. A case for intelligent disks (IDISKS). ACM SIGMOD Record, 27(3):42--52, Sept. 1998.

First 50 documents

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