D. Ungar and F. Jackson. Tenuring policies for generation-based storage reclamation. In Conference proceedings on Object-oriented programming systems, languages and applications, pages 1--17. ACM Press, 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... javac jess mpegaudio mtrt jbb euler montecarlo raytracer 5 10 15 20 space saving (as of original heap space) Prolific Arrays Reduction in Space Occupied by Instances of Prolific Scalar Objects tification of large objects can be an absolute measure (e.g. more than 1024 bytes [46] or 256 bytes [27] or a relative one (i.e. identify the object type whose instances occupy substantial space [25] Many recent generation collection implementations use the LOS for storing large objects [23, 2] The cost of write barriers is also significant, especially for pointerintensive ....

D. M. Ungar and F. Jackson. Tenuring policies for generation based storage reclamation. ACM Transactions on Programming Languages and Systems, 14(1):1--27, 1992.

....Garbage Collection Evaluation Garbage collection is an excellent case in point of the tradeoffs between design for performance and design for evaluation. In recent years, trace driven simulation proven to be an easy and effective way to determine the performance of garbage collection algorithms [3, 5]. Zorn describes a tool for controlled comparison of the performance of different garbage collection algorithms called MARS (Memory Allocation and Reference Simulator) 4] MARS is a program that attaches to a commer cial Common Lisp system and is passed information about object allocations, ....

David Ungar and Frank Jackson. Tenuring policies for generation-based storage reclamation. In OOP- SLA'88 Conference Proceedings, pages 1-17. ACM, September 1988.

....having to collect the older ones, the collector must be capable of finding pointers into the young generations. This requires either the use of a write barrier similar to the one we found in the incremental functioning mode (see previous section) to keep track of such cross partition pointers [6, 41, 89, 119, 129], or indirect pointers from older to younger generations [81] The set of references pointing from older to younger generations is usually called remembered set. When a younger generation is collected, the pointers in the corresponding remembered set are part of the GC root. 2.2.4 System ....

David Ungar and Frank Jackson. Tenuring policies for generation-based storage reclamation. In Norman Meyrowitz, editor, Conference on Object Oriented Programming Systems, Languages and Applications (OOPSLA '88) Proceedings, pages 1--17, San Diego, California (USA), September 1988. ACM Press.

....supporting the idea that the older an object is the longer it is likely to live. Ungar and Jackson improved the efficiency of a two generation collector for Smalltalk by explicitly segregating some kinds of objects and by using a dynamically computed threshold for deciding when to promote objects [16, 17]. Barrett and Zorn extended this idea further by collapsing the two generations and using a movable threatening boundary to separate the spaces for short and long lived objects [2] In the absence of information about the lifespans of objects, a number of proposals have been put forward to ....

D. M. Ungar and F. Jackson. Tenuring policies for generation-based storage reclamation. ACM SIGPLAN Notices, 23(11):1--17, 1988.

....system. 50 large object area old space new space user stacks data area text area machine registers heap OS s user stack new OT kernel stack old OT Figure 4.6: Memory layout 51 We have adopted large object area (see Fig. 4. 6) and demographic feedback mediated tenuring [ Ungar and Jackson 1988; Ungar and Jackson 1992 ] as our tenuring policy. A large object area keeps the data of all large objects such as bitmaps and strings. In Ungar Jackson scheme, the headers of these large objects are stored in the youngest generation and then scavenged, but these large objects are never tenured ....

D. Ungar and F. Jackson: "Tenuring Policies for Generation-Based Storage Reclamation," in OOPSLA [ OOPSLA 1988 ] , pp. 1--17.

.... build up data structures quickly, and then use those data structures for long periods (often nearly the whole running time of the program) These patterns are well known, from anecdotal experience by many people (e.g. Ros67, Han90] from research on garbage collection (e.g. Whi80, WM89, UJ88, Hay91, Hay93, BZ95, Wil95] 32 and from a recent study of C and C programs [WJNB95] 32 It may be thought that garbage collected systems are sufficiently different from those using conventional storage management that these results are not relevant. It appears, however, that these ....

David Ungar and Frank Jackson. Tenuring policies for generation-based storage reclamation. In Norman Meyrowitz, editor, Conference on Object Oriented Programming Systems, Languages and Applications (OOPSLA '88) Proceedings, pages 1--17, San Diego, California, September 1988. ACM Press.

....live data is S S r . Calculation using remembered set reachability treats all objects in the uncollected remainder U as if live, since it does not examine U , and thus potentially overestimates the set S. In any region based collector, there is potential for such excess retention or nepotism [Ungar and Jackson, 1988] on each garbage collection. How much excess retention is there Clearly, a collection scheme could have dramatically high amounts of excess retention if it were the case that (1) pointers tended to cross region boundaries predominantly into the collected region, thus there were many pointers in ....

....20 , however [Zorn, 1990a] of the youngest generation, and thus most pauses are short. Numerous practical implementations and studies of predictive and adaptive management of generations have reported good performance [Caudill and Wirfs Brock, 1986; Courts, 1988; Shaw, 1988; Sobalvarro, 1988; Ungar and Jackson, 1988; Wilson, 1989; Wilson and Moher, 1989; Appel, 1989b; Wilson et al. 1991; Hudson et al. 1991; Stefanovic, 1993b; Stefanovic and Moss, 1994; Diwan et al. 1995; Barrett and Zorn, 1995] The performance of a collector is affected by a number of factors: the number of generations in the system; ....

David M. Ungar and Frank Jackson. Tenuring policies for generation-based storage reclamation. SIGPLAN Notices, 23(11):1--17, 1988.

....according to their choice of objects for collection. In all these collectors, we fix the size of the collected region rather than allowing it to vary during program execution, to simplify our analysis. Previous research shows that dynamically sizing the collected region can improve performance [29, 36, 34, 1, 5], but this investigation is beyond the scope of our paper. A youngest only (YO) collector always chooses some youngest (rightmost) subsequence of the sequence of heap objects (Figure 2) In our implementation, the YO collector fills the entire heap and then repeatedly collects the youngest portion ....

.... Therefore, without making specific comparisons, which are difficult when evaluation metrics as well as underlying languages are widely different, we recognize that our study draws on previous experience with generational garbage collection implementations [19, 27, 20, 24, 28, 35] their policies [29, 30, 31, 34, 1, 13], their write barrier mechanisms [33, 15, 14] and their evaluation with respect to object allocation and lifetime behavior [3, 26, 11] Achieving performance improvements with generational collection critically depends on setting or adapting the configuration parameters right incorrectly chosen ....

[Article contains additional citation context not shown here]

Ungar, D. M., and Jackson, F. Tenuring policies for generationbased storage reclamation. SIGPLAN Notices 23, 11 (1988), 1--17.

....the algorithm can be efficient on a stock multiprocessor. As Appel et al. describe, this algorithm can be used with standard techniques for improving copying collectors, such as separating generations of objects [LH83, Ung84] separating objects that contain only non pointers from normal objects [CWB86, UJ88, App90], and providing a separate space for allocating new objects[Ung84] The changes described in this paper do not require objects to have any particular format. The algorithm is adaptable to nearly any environment, though its efficiency and real time latency will depend on the machine architecture, ....

David Ungar and Frank Jackson. Tenuring policies for generationbased storage reclamation. In Proceedings of OOPSLA `88, pages 1--17, November 1988. printed as SIGPLAN Notices, 23(11).

....unnoticed for human users. GC over old objects, typically kept in a larger area, now and then produce longer stops (in an interactive environment this would happen a few times a day, during lunch and coffee breaks ) Tenuring policies control when objects are moved to older areas (generations) [UJ88]. The GC scheduling is done in the last minute and tenuring can generate GC cycles in the next older generation and so on. In the worst case there is a snowball effect of GC cycles generated in all generations. A recent improvement, the train algorithm [HM92,SG95] has presented a technique to ....

D. Ungar, F. Jackson. Tenuring Policies for Generation-Based Storage Reclamation. Proceedings of OOPSLA'88, ACM SIGPLAN Notices 23(11).

....generation based collectors include: opportunistic collectors [Wilson and Moher, 1989] ephemeral collectors used in Symbolics machines [Moon, 1984] and the Tektronix Smalltalk collector. All three commercial Smalltalk systems, DigiTalk, Tektronix and ParcPlace, adopted generation scavengers [Ungar and Jackson, 1988]. The New Jersey SML compiler [Wilson, 1992a] also includes a generation collector. Demers et al. [1990] have investigated a generation scheme combined with a mark scan garbage collector for use with Scheme, Mesa and C intermixed in one virtual memory. Before Demers et al. [1990] many believed that ....

....the most complex single processor collection schemes, they suffer poor performance if many cells live just long enough to be promoted before dying, the so called premature tenuring problem. Ungar and Jackson propose an adaptive tenuring scheme based on extensive measurements of real Smalltalk runs [Ungar, 1988; 1992] This scheme varies the tenuring threshold depending on dynamically measured cell lifetimes. It also proposes a refinement that has been included in the ParcPlace [1991] collector. With languages like Smalltalk, interactive response is at a premium and many large cells, mainly bit maps and ....

Ungar DM and Jackson F (1988) Tenuring policies for generation-based storage reclamation, OOPSLA'88, ACM SIGPLAN Notices 23(11)1-17.

....system model. The algorithms we studied are described in Section 4; Section 5 gives our experiment configurations, and Section 6 gives experimental results. We conclude in Section 7. 2 Related Work Most of the work in the area of storage reclamation has been focused on volatile memory systems [1, 2, 3, 7, 9, 13, 16, 19, 20, 21, 24]. Comparatively little attention has been paid for implementing automatic storage reclamation for large persistent object storage systems, and none that we are aware of considers the client server model. The most directly relevant work is by Butler [4] who showed that copying collectors are ....

....partition oriented copying algorithm. It provides the reclustering and compacting benefits of Baker s incremental technique without its space overhead and its delay before reclaimed storage is available. This scheme uses one of the basic ideas behind the generational garbage collection algorithms [20, 21, 24]: instead of copying the entire database, limit the scope of collection to a small region of the storage that is expected to provide the maximum free space for the collection effort. However, unlike generational garbage collectors, which select the region by age of the object stored, the partition ....

Ungar, D. and Jackson, F. "Tenuring Policies for Generation-Based Storage Reclamation," OOPSLA 1988 Conference Proceedings, San Diego, California, 1988, pp.1-17

....have various lifetimes. Some stable objects live for a long time; other temporary objects live only for a short time. Objects have a tendency that many of them are short lived, but objects live for a long time once they have survived for more than some period of time [Lieberman and Hewitt, 1983; Ungar and Jackson, 1988]. This means that the possibility of becoming garbage for a newly created object is higher than the possibility for an old one. Baker et. al, 1991] measured life times of files in Sprite distributed file system, and the measurements showed that most files have short lifetimes. Lieberman and ....

....short lifetimes. Lieberman and Hewitt, 1983] suggested a garbage collection algorithm based on the lifetimes of objects. The basic idea was that the frequency of garbage collection was selected according to the age of objects. For younger objects, garbage collection was performed more frequently. [Ungar and Jackson, 1988] used tenuring policies for garbage collection algorithm to improve performance. Once an object has survived for more than some threshold, the object gets tenure, and the garbage collection algorithm is performed very infrequently on tenured objects. In timeouts, the TTL of an object is the period ....

David Ungar and Frank Jackson, The Tenuring Policies for Generation-Based Storage Reclamation, OOPSLA, September 1988, pages 1-17.

....generally be very large and therefore garbage collection will take a long time and cause a lot of paging activity. On the other hand, not collecting the old space when there is much garbage will also slow down normal processing, because of increased paging due to fragmentation of virtual memory [Ungar88]. It is not strictly necessary that old space (or at least not all of it) be a paging virtual memory segment. If old space is a multilevel address space, where objects are swapped in and out of virtual memory, then the problem of the old space fragmenting a paged virtual memory disappears. Garbage ....

D. Ungar and F. Jackson, "Tenuring Policies for Generation-Based Storage Reclamation, " Object-Oriented Programming Systems, Languages and Applications, pp. 1-17, San Diego, California, 1988.

....there is little or no excess retention in either direction. For other programs, there is considerable excess retention, often higher when the collected region is among older objects. We can say that youngest region collectors are in practice more resilient to excess retention [Ungar, 1984; Ungar and Jackson, 1988], but other collector schemes can nevertheless copy less data overall, and thus achieve better absolute performance. 3.2.2 Size of remembered sets Another repercussion of choosing alternatives to generational collection is the potential increase in the number of pointers tracked, and the time to ....

....sizes can cause performance to degrade severely. We have confirmed these matters in our observations of NONGEN, 2GYF, and 3GYF on some of the traces. Therefore, we can say that choosing a good regime of generations is not an easy task, and it is not yet fully understood despite numerous studies [Ungar and Jackson, 1988; Zorn, 1989; Wilson and Moher, 1989b; Appel, 1989; Barrett and Zorn, 1995] However, we can also say that it is a matter of tuning the performance within the class of true youngest first collection schemes. Our goal in this study is not to examine how to tune a particular scheme, but instead to ....

Ungar, D. M. and Jackson, F. (1988). Tenuring policies for generationbased storage reclamation. SIGPLAN Notices, 23(11):1--17.

....generation collectors are one of the most promising collection techniques, they suffer poor performance if many cells live a fairly long time, the so called premature tenuring problem. Ungar and Jackson propose an adaptive tenuring scheme based on extensive measurements of real Smalltalk runs [Ungar, 1988; 1992] This scheme varies the tenuring threshold depending on dynamically measured cell lifetimes. It also proposes a refinement that has been included in the ParcPlace [1991] collector. In systems like Smalltalk, interactive response is at a premium but the system contains many large cells that ....

....collectors include: opportunistic collectors [Wilson and Moher, 1989] ephemeral collectors and the Tektronix Smalltalk collector. In terms of usage, all three commercial U.S. Smalltalk systems (DigiTalk, Tektronix and ParcPlace systems) have adopted generational automatic storage reclamation [Ungar and Jackson, 1988]. The SML NJ compiler [Wilson, 1992] also uses a generational collector. Deimer et al. [1990] have investigated a generational scheme combined with a conservative mark and sweep garbage collector designed for use with Scheme, Mesa and C intermixed in one virtual memory. Wilson, Lam and Moher [1990] ....

Ungar D and Jackson F (1988) Tenuring policies for generation-based storage reclamation, OOPSLA'88, SIGPLAN Notices 23(11), ACM, 1-17.

....GC implementation, this model could be validated. We have done this for a Smalltalk application and tested it with several GC implementations; there is, however, no room here to discuss the specifics of our model As an alternative, published data from real experiments such as those mentioned in [UJ88, UJ92] could be used to determine a PO . 3.1 Uniform distribution We have first tested our profiler setup by using a uniform distribution; in this case, the GetRndData function is specifically simple. Fig. 2 shows the spectrum of remaining life time for S max = 4 and Tmax = 30 s. Of course, the ....

David Ungar and Frank Jackson. Tenuring policies for generationbased storage reclamation. In Proc. OOPSLA'88, pages 1--17, September 1988.

....collector. So far, no one has found a perfect solution. Each scheme has its own weakness; for example reference counting fails to reclaim circular structures; Baker s algorithm [Bak77] and its descendents, including [LH83, AEL88] require a potentially costly read barrier; Generation Scavenging [Ung84, Ung86, UJ88] may fill up memory with tenured garbage, opportunistic strategies [Wil89] may fail to receive adequate opportunities, multiple generation stop and copy algorithms [CW86] may experience long waterfall pauses when an old generation requires reclamation. Now, at Berkeley I (Ungar) learned from ....

....but two new ones were lurking just around the corner. Looking back, I believe that some Lisp folks tried to warn me at OOPSLA 87. Later, when Frank and I implemented the first commercial Generation Scavenger in a compiled system, we ran trace driven simulations of four actual long term sessions [UJ88]. Although most pauses were short, there were a significant number of long ones, and worse, the long ones were clustered. We had met the indigestion devil, the pig in a python. As a result, we started thinking of good times and bad times: simple Generation Scavenging is fine for the good times, ....

D. Ungar, F. Jackson, "Tenuring Policies for Generation-Based Storage Reclamation," OOPSLA'88, 1-17.

No context found.

D. Ungar and F. Jackson. Tenuring policies for generation-based storage reclamation. In Conference proceedings on Object-oriented programming systems, languages and applications, pages 1--17. ACM Press, 1988.

No context found.

D. Ungar and F. Jackson. Tenuring policies for generation-based storage reclamation. In N. K. Meyrowitz, editor, Conference on Object-Oriented Programming Systems, Languages, and Applications (OOPSLA'88), September 25-30, 1988.

No context found.

D. Ungar and F. Jackson. Tenuring policies for generation-based storage reclamation. In Conference proceedings on Object-oriented programming systems, languages and applications, pages 1--17. ACM Press, 1988.

No context found.

David M. Ungar and Frank Jackson. Tenuring policies for generation-based storage reclamation. ACM SIGPLAN Notices, 23(11):1--17, 1988.

No context found.

D. Ungar and F. Jackson, `Tenuring policies for generation-based storage reclamation', in Norman Meyrowitz (ed.), Proceedings of the OOPSLA '88 Conference, San Diego, CA, 1988. ACM/SIGPLAN Notices, 23, (11), 1--16 (1988).

No context found.

D. M. Ungar and F. Jackson, "Tenuring policies for generation-based storage reclamation," Proc. OOPSLA '88 (ACM SIGPlan Notices, 23:11), Nov. 1988, pp. 1-16.

No context found.

D. Ungar, Tenuring Policies for Generation-Based Storage Reclamation, OOPSLA Proceedings, pp. 1-17, Sept. 25-30, 1988.

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