Fact-checked by Grok 2 weeks ago

Elevator algorithm

The elevator algorithm, also known as the SCAN algorithm, is a disk scheduling technique in operating systems that optimizes the movement of the disk read/write head by servicing pending requests in a unidirectional sweep across the disk surface, reversing direction only upon reaching an endpoint, much like an elevator traversing floors in a building. In operation, the algorithm maintains a of disk requests sorted by track number; the head starts from its current position and moves inward or outward—depending on the initial direction—servicing all requests it passes until it hits the disk's boundary (track 0 or maximum), at which point it reverses and services the remaining requests in the opposite direction. This process repeats for subsequent requests, ensuring that every request is eventually addressed without indefinite postponement, thus avoiding . The algorithm excels in high-load scenarios by minimizing total head movement—often outperforming first-come, first-served (FCFS) and shortest-seek-time-first (SSTF) methods, as demonstrated in examples where it reduces seek distances from 640 cylinders (FCFS) to around 208-236 cylinders. It balances throughput and fairness better than purely greedy approaches like SSTF, which can lead to of distant requests, while providing lower response time variance than FCFS under heavy traffic. However, it may result in longer waits for requests near recently serviced areas, prompting variants like C-SCAN (circular SCAN), which returns the head to the starting point without servicing on the reverse trip for more uniform latency, or LOOK, which halts at the farthest request rather than the disk end to shave unnecessary travel. As a modular component of the , the elevator algorithm can be swapped or tuned independently, making it adaptable to different and workloads; however, in modern SSDs, where traditional seek-based optimizations are less relevant due to the absence of , other algorithms such as deadline-based scheduling (incorporating request deadlines) are used to enhance guarantees.

Introduction

Definition

The elevator algorithm, also known as the algorithm, is a disk scheduling technique used in operating systems to optimize the movement of the disk read/write head. It services pending read and write requests by moving the head in a unidirectional sweep across the disk tracks, servicing all requests in one direction before reversing at the disk's endpoint, similar to how an services floors in a building. Key concepts include the of disk requests sorted by number, the current position of the disk head, and its direction of travel (inward or outward). For example, if moving toward higher numbers, the head services all requests with tracks greater than or equal to its current position until reaching the maximum , then reverses to service the remaining requests. This approach is named for its analogy to dispatching, where the "" (disk head) continues in one direction, fulfilling requests encountered before changing course, promoting efficient in both mechanical and computational systems. In contrast to the first-come, first-served (FCFS) algorithm, which processes requests in arrival order and can result in excessive head movement due to erratic jumps, the elevator algorithm groups requests by direction to reduce total seek time and promote more predictable disk access patterns.

Purpose and Benefits

The elevator algorithm aims to minimize the total head movement and average seek time in disk I/O operations by systematically servicing requests in the current direction before reversing, thereby improving overall system performance under load. A primary benefit is enhanced efficiency in high-traffic scenarios, where it reduces unnecessary back-and-forth movements compared to greedy methods like shortest-seek-time-first (SSTF), which may starve distant requests. This leads to better throughput and fairness, ensuring no request is indefinitely postponed (avoiding ). Additionally, it provides lower variance in response times than FCFS, making it suitable for environments with varying workloads, and its simplicity allows easy implementation in operating system kernels. In modern contexts, while less critical for solid-state drives (SSDs) due to lack of mechanical seeks, the algorithm's principles influence deadline-aware scheduling extensions for latency-sensitive applications.

Historical Development

Origins

The elevator algorithm, or , for disk scheduling originated in the early 1970s as part of efforts to optimize operations on early hard disk drives, where seek times dominated . It was formalized and analyzed in academic literature amid the growth of minicomputers and mainframes, such as the , which used rotating requiring efficient request ordering to minimize head movement. A foundational analysis appeared in 1972 with the paper "Analysis of Scanning Policies for Reducing Disk Seek Times" by E. G. Coffman Jr., L. A. Klimko, and B. Ryan, published in the SIAM on . The authors modeled head motion under scanning policies, demonstrating that unidirectional sweeps () balanced seek time reduction with fairness, outperforming earlier methods like first-come, first-served (FCFS). Concurrently, T. J. Teorey and J. P. Pinkerton's "A Analysis of Disk Scheduling Policies" in Communications of the ACM compared with shortest-seek-time-first (SSTF) and others, highlighting its advantages in high-load scenarios through simulation. These works established as a standard technique, inspired by dispatching but adapted for linear track layouts on disks.

Evolution and Adoption

By the late 1970s and 1980s, the algorithm was integrated into commercial operating systems as disk capacities and speeds increased. Early UNIX variants, such as Version 7 (1979), employed basic disk scheduling, with SCAN-like policies emerging in research and proprietary systems to handle growing I/O demands in multitasking environments. The term "elevator" gained prominence in the , where I/O schedulers were abstracted as an "elevator" layer. The initial implementation, known as the "Linus Elevator," appeared in 2.4 (released January 2001), using a deadline-based variant of to merge and reorder requests for fairness and throughput. In 2.6 (2003), this evolved into multiple options, including the (CFQ) scheduler, which incorporated elevator principles with per-process queues to prevent . Adoption extended to other systems, such as kernels using similar rotational policies, and BSD variants. By the , with the rise of solid-state drives (SSDs), traditional seek-optimized algorithms like became less critical due to negligible access latencies, prompting shifts to deadline and multi-queue schedulers (e.g., 's mq-deadline in 2018 and BFQ for fairness). As of 2025, the elevator framework persists in for hybrid storage, influencing NVMe optimizations and systems.

Core Mechanics

Basic Operation

The elevator algorithm, also known as the algorithm, operates by simulating the motion of an in a building, where the disk head serves as the elevator and disk tracks represent floors. In a single-head scenario, the algorithm initializes the disk head at a specific track position, often determined by the current location when requests arrive, and assigns an initial direction—typically upward (toward higher track numbers) or downward (toward lower track numbers)—based on the distribution of pending requests. This setup ensures that the head begins servicing from its starting point without unnecessary . During the scanning phase, the head moves continuously in its current direction, stopping to service all requests encountered along the way in sorted order within that direction. It only pauses for requests that align with the travel path—for instance, if moving upward from track 50, it services any pending requests at tracks , 70, and 90 before proceeding further—until it reaches the end of the disk (track 0 or the maximum track, such as 199). This unidirectional sweep to the boundary minimizes seek time by grouping services efficiently while avoiding reversals mid-scan. New requests arriving during the scan are queued for the next pass in the appropriate direction to maintain fairness. Upon completing the scan by hitting a disk boundary, the algorithm triggers a reversal: the head changes direction and begins servicing the opposite set of queued requests on the return trip. For example, after scanning upward to the maximum track, it reverses downward, now stopping for previously unserved requests below the reversal point. This process repeats, alternating directions, until all requests are fulfilled, at which point the head idles or resets to a default position awaiting new inputs. The reversal logic relies on simple conditional checks following the boundary reach: the head then services requests in the new direction until the opposite boundary. The core decision-making can be outlined in pseudocode as follows, focusing on direction changes based on pending calls and boundary reaches:
Initialize: current_position = initial_track; direction = UP (or DOWN based on requests);
Queue requests sorted by track number;
disk_min = 0; disk_max = 199;  // Example disk boundaries

While requests remain:
    If direction == UP:
        // Service requests > current_position
        While there are requests > current_position:
            Move to next request > current_position;
            Service request;
            Update current_position;
        // Move to disk end if not already there
        If current_position < disk_max:
            Move to disk_max;
            current_position = disk_max;
        // Reverse if requests remain in opposite direction
        If there are requests < current_position:
            direction = DOWN;
        Else:
            // All done or wait for new
            break;
    Else (direction == DOWN):
        // Service requests < current_position
        While there are requests < current_position:
            Move to next request < current_position;
            Service request;
            Update current_position;
        // Move to disk end if not already there
        If current_position > disk_min:
            Move to disk_min;
            current_position = disk_min;
        // Reverse if requests remain in opposite direction
        If there are requests > current_position:
            direction = UP;
        Else:
            // All done or wait for new
            break;
This structure ensures systematic coverage without , as every request will eventually align with a direction.

Request Processing

In the elevator algorithm, disk I/O requests—specifying read or write operations to particular —are collected in a central maintained by the operating system . The is typically kept sorted by track number to facilitate efficient traversal during scans, allowing quick identification of requests in the current direction. New requests arrive asynchronously from processes and are added to the in , inserted at the appropriate position based on track number without interrupting the ongoing ; for example, a request for track 80 during an upward sweep from track 50 would be queued for servicing if 80 > 50, or deferred to the return trip otherwise. This dynamic insertion ensures fairness, as requests are not discarded but prioritized by the algorithm's directional logic rather than arrival order. Integration occurs seamlessly during each sweep: as the head moves, the controller scans the for the next eligible request matching the (e.g., tracks greater than current for upward), servicing them in ascending/descending order while ignoring opposite- ones until reversal. The indicator—a simple (up or down)—filters the queue access, ensuring the head responds only to compatible requests until a is reached. Edge cases, such as an empty in the current direction, are handled by proceeding directly to the disk before , preventing unnecessary idling. High-priority or deadline-sensitive requests (in extended variants) may be flagged in the for potential overrides, though standard SCAN treats all equally to avoid complexity. Overloaded under heavy I/O load are managed by the OS scheduler, which may throttle new requests temporarily to maintain system stability.

Variations

Standard Modifications

The LOOK variant improves upon the basic elevator algorithm by reversing the disk head's direction only upon reaching the farthest pending request in the current direction, rather than proceeding to the disk's (track 0 or maximum). This modification eliminates unnecessary empty travel beyond active requests, thereby reducing seek time compared to the standard approach. The C-SCAN variant adopts a unidirectional scanning pattern, where the disk head services requests in one direction from the innermost to outermost track (or vice versa), then jumps back to the starting track without servicing requests on the return path. This treats the disk tracks as a , ensuring more uniform waiting times for requests across the disk surface. The C-LOOK variant combines elements of C-SCAN and LOOK, servicing requests in one direction up to the farthest request, then jumping to the farthest request in the opposite direction without full traversal to the disk end. This further optimizes seek distances by avoiding empty sweeps, making it particularly efficient for workloads with clustered requests. Early studies on these seek-reducing modifications, including LOOK and C-SCAN variants, reported performance improvements such as 5-10% lower average response times under medium to heavy loads compared to less adaptive methods, with up to 40% reductions in specific traced workloads.

Modern Enhancements

In modern storage systems, particularly solid-state drives (SSDs), traditional seek-based algorithms like have diminished relevance due to the absence of mechanical heads and uniform access . Instead, enhancements focus on deadline-based scheduling, where requests are prioritized by assigned deadlines to guarantee bounds, or the Noop scheduler, which simply merges and reorders requests without complex scanning to SSD parallelism. Multi-queue schedulers, such as mq-deadline or BFQ in kernels (as of 2025), extend principles to handle concurrent I/O streams from NVMe SSDs by distributing requests across queues and applying variants like deadline enforcement per queue, improving throughput by 20-50% in multi-threaded workloads compared to single-queue . These adaptations incorporate request aging to prevent , similar to 's fairness, while optimizing for flash-specific constraints like .

Examples

Simple Illustration

Consider a disk with 200 tracks where the read/write head starts at track 50, moving toward track 0 (inward). Suppose pending read/write requests are at tracks 176, 79, 34, 60, 92, 11, 41, and 114. In this setup, the elevator algorithm (SCAN) services all requests in the current before reversing. Starting inward, the head moves to track 41, then 34, then 11. It continues to the disk's inner boundary at track 0 (even without a request there), servicing no additional requests en route. Upon reaching track 0, it reverses and moves outward, servicing the remaining requests at 60, 79, 92, 114, and finally 176. With no further requests, the process ends. The execution can be visualized as a seek sequence: 50 → 41 → 34 → 11 → 0 → 60 → 79 → 92 → 114 → 176, illustrating the scan-like motion across tracks without unnecessary reversals mid-sweep. This illustration results in nine seeks and a total head movement of 226 tracks (calculated as |50-41| + |41-34| + |34-11| + |11-0| + |0-60| + |60-79| + |79-92| + |92-114| + |114-176|), demonstrating efficient servicing in a basic single-disk scenario compared to alternatives like FCFS, which might exceed 300 tracks for the same requests.

Multi-Elevator Scenario

In multi-disk systems, the elevator algorithm can operate across multiple drives managed by a storage controller, where requests are assigned to individual disks to optimize overall . A representative setup involves a array or server with 3 disks, each handling up to thousands of depending on the model. The assignment prioritizes the disk closest to the data location, matching request type (read/write) and current load capacity to avoid bottlenecks and ensure balanced distribution. To illustrate coordination, consider a where requests arrive dynamically. Disk A, with head at track 30 and scanning inward, receives assignments for requests at tracks 50 and 70; it serves these sequentially without reversing prematurely, adhering to the SCAN principle of completing directional sweeps. Concurrently, Disk B, scanning outward from track 150, handles requests at tracks 140 and 120, minimizing seek times for aligned directions. If Disk B nears limits—say, exceeding 80% utilization—the controller dynamically reassigns queued requests to Disk C, which may be idle or better positioned, preventing I/O stalls. The storage controller applies elevator algorithm logic per disk by evaluating factors like estimated seek time and directional alignment to minimize conflicts where disks handle unbalanced workloads inefficiently. This includes rules to balance load during peak traffic and to zone assignments for high-demand data areas, ensuring no single disk handles disproportionate requests. Such coordinated operation yields improvements in system efficiency, including reduced average response times and better throughput distribution; for instance, variants like in zoned multi-disk setups have shown over 20% better response times than non-scheduled approaches in simulated environments with prefetching caches.

Analysis

Performance Metrics

The performance of the elevator algorithm (SCAN) in disk scheduling is evaluated using key metrics that quantify efficiency in servicing I/O requests. These include total head movement, which measures the cumulative distance the disk head travels to service requests; average seek time, representing the typical time for head positioning; throughput, indicating requests processed per unit time; and variance in response times, reflecting fairness in servicing. Power consumption, tied to seek operations, is also considered in energy-aware systems. Total head movement is the in track positions as the head services requests in directional order: \text{Total Head Movement} = \sum |p_{i} - p_{i+1}| where p_i are the positions serviced. In an example with requests at 98, 183, 37, 122, 14, 124, 65, 67 and head starting at 53 moving inward, SCAN achieves 208 cylinders of movement, compared to 640 for FCFS. seek time is the total head movement divided by the number of requests, providing an indicator of . Seek time is particularly sensitive to traffic patterns, with SCAN performing well during high loads by minimizing unnecessary back-and-forth movements. Throughput measures the number of requests serviced per second, often improved by SCAN's systematic approach, which reduces overhead at the . This metric highlights the algorithm's ability to handle directional request clusters efficiently. The number of direction reversals per cycle serves as a for , as excessive reversals increase and wear; SCAN limits reversals to endpoints, resulting in fewer interruptions than non-directional methods like SSTF. Power consumption is calculated based on seek distance and head speed, with SCAN reducing idle traversals and thus lowering usage in simulated workloads. Response time variance is low under moderate traffic due to the algorithm's fair servicing of requests in . These metrics are typically assessed through simulation-based methods, such as trace-driven simulations using disk request logs from real workloads or synthetic patterns, or via tools that record seek distances and latencies.

Advantages and Limitations

The elevator algorithm offers predictable service by systematically scanning the disk in one direction before reversing, which minimizes variance in response times and ensures fairness by preventing indefinite starvation of requests. This approach provides low computational overhead, as it relies on simple queue operations to maintain and process requests in directional order without complex optimizations. It performs effectively under uniform traffic conditions, where requests are distributed evenly across the disk surface, achieving high throughput and reasonable average waiting times compared to uncoordinated methods. Despite these strengths, the algorithm struggles in zoned or bursty traffic patterns, such as clustered requests in specific disk regions, leading to prolonged waits for opposite-direction calls as the head completes full sweeps. This can cause uneven load distribution, with outer zones potentially underserved during directional passes that prioritize inner areas. In low-traffic scenarios, modern analyses highlight its energy inefficiency, as the head incurs unnecessary power for traversing empty portions of the disk, exacerbating consumption in power-sensitive environments like mobile storage. Relative to first-come-first-served (FCFS), the elevator algorithm excels under high loads by reducing overall seek distances, but it underperforms predictive variants like shortest seek time first (SSTF) during bursty peaks, where localized requests benefit more from greedy selection.

References

  1. [1]
    Disk Scheduling and File System Design - UCLA Computer Science
    The Elevator algorithm is a hybrid algorithm that attemps to be fair and have good throughput. The head is like an elevator, the requests people and the sectors ...
  2. [2]
    [PDF] Disk Scheduling
    The disk-scheduling algorithm should be written as a separate module of the operating system, allowing it to be replaced with a different algorithm if necessary ...
  3. [3]
    [PDF] Disk Scheduling Algorithms Prepared By:
    LOOK Disk Scheduling Algorithm: LOOK is the advanced version of SCAN (elevator) disk scheduling algorithm which gives slightly better seek time than any ...<|control11|><|separator|>
  4. [4]
    [PDF] Module 13: Secondary-Storage Structure - Columbia CS
    Sometimes called the elevator algorithm. • Illustration shows total head movement of 208 cylinders. Operating System Concepts. 13.8. Silberschatz and Galvin c ...
  5. [5]
    [PDF] Chapter 10: Mass-Storage Systems - FSU Computer Science
    ▫ The disk-scheduling algorithm should be written as a separate module of ... Silberschatz, Galvin and Gagne ©2013. Operating System Concepts – 9th Edition.
  6. [6]
    [PDF] Disk Scheduling and SSDs
    ▫ Both noop and the elevator algorithm emphasize throughput at the cost of latency. ▫ Can result in starvation. ▫ Deadline scheduling: assign a deadline to each ...
  7. [7]
    The mechanics of lift routing - Designing Buildings Wiki
    Jun 22, 2018 · Known as 'collective control' or the 'elevator algorithm' (SCAN), the earliest approach to lift dispatching is still used today. ... strategy ...
  8. [8]
    CS322: Disk Scheduling - Gordon College
    This algorithm is also known as the elevator algorithm because it works the same way an elevator services requests in a building. When it is going up, it ...
  9. [9]
    Difference between FCFS and SCAN disk scheduling algorithms
    Jul 12, 2025 · The FCFS Scheduling Algorithm processes requests in the sequential order in which they arrive in the disk queue.
  10. [10]
    None
    ### Summary of Elevator Algorithms (SCAN-like and Basic)
  11. [11]
    Elevator control system operations & codes | Cibes Symmetry
    Aug 7, 2018 · This type of operation allows the user to stop mid-travel or even change direction prior to arriving at a floor. This is the only system allowed ...Missing: algorithm benefits
  12. [12]
    Elevator Controllers Explained Simply - MAS Industries Pvt Ltd
    1. Selective Collective Operation ... Common in high-rise commercial buildings, this controller serves multiple passengers heading to various floors. It collects ...
  13. [13]
    Elevator Control System Market Size & Share Global Analysis ...
    Selective Collective Operation optimizes elevator operation by grouping passengers with similar destinations, reducing travel time, and improving energy ...
  14. [14]
    Energy Efficient Elevator Technologies - ASME
    Sep 19, 2012 · They can reduce a building's overall energy usage by reducing the number of stops and even the total number of elevators required when used with ...
  15. [15]
    The Hidden Science of Elevators - Popular Mechanics
    May 25, 2016 · The Elevator Algorithm. The earliest and simplest reasonable approach to elevator dispatching is still surprisingly common. Known as “collective ...
  16. [16]
    Elevator history timeline - Otis Elevator
    Sep 20, 2023 · The company sells eight elevators in 1854 and 15 in 1855. In 1857, Otis installs its first passenger elevator in New York City's E.V. Haughwout ...
  17. [17]
    The Evolution of Elevators: Physical-Human Interface, Digital ...
    Otis's invention took a simple flat-leaf spring from a cart and applied it to the roof of an elevated hoist such that, in the event of the hoist rope's failure, ...
  18. [18]
    None
    ### Summary of Elevator Systems Adoption and Adaptations (1990s-2000s)
  19. [19]
    [PDF] A Review of Elevator Dispatching Systems - IAENG
    Jul 1, 2016 · This paper reviews a number of widely used elevator dispatching systems and their underlying algorithms including Estimated Time to Dispatch. ( ...Missing: history selective origins 1950s<|control11|><|separator|>
  20. [20]
    DISK SCHEDULING ALGORITHMS
    The seek time is increased causing the system to slow down. Disk Scheduling Algorithms are used to reduce the total seek time of any request.
  21. [21]
    [PDF] Chapter 12: Mass-Storage Systems
    Silberschatz, Galvin and Gagne ©2009. Operating System Concepts – 8th Edition. SCAN ... ▫ The disk-scheduling algorithm should be written as a separate module of.
  22. [22]
    None
    ### Summary of SCAN Algorithm from http://www.cs.umd.edu/class/spring2020/cmsc412/Slides/Set%208%20Disk%20Scheduling.pdf
  23. [23]
    SCAN (Elevator) Disk Scheduling Algorithms - GeeksforGeeks
    Jul 12, 2025 · In the SCAN Disk Scheduling Algorithm, the head starts from one end of the disk and moves towards the other end, servicing requests in between one by one and ...
  24. [24]
    [PDF] The History of Lift Traffic Control1
    The most common form of automatic control used for a single lift is collective control. This is a generic designation for those types of control where all ...Missing: selective | Show results with:selective
  25. [25]
    Elevator Control System - Electrical Knowhow
    In the topic " Basic Elevator Components - Part One" , I indicate that the basic elevator components are as follows: Car. Hoistway. Machine/drive system.
  26. [26]
    [PDF] Decision-Theoretic Group Elevator Scheduling
    Jun 9, 2003 · The execution of the schedule is performed by alternating the direction of movement of each car and servicing all hall calls assigned to it in ...Missing: queuing | Show results with:queuing
  27. [27]
    [PDF] Scheduling Algorithms for Modern Disk Drives
    The cyclical scan algorithm (C-LOOK), which always schedules requests in ascending logical order, achieves the highest performance among seek-reducing ...Missing: early | Show results with:early
  28. [28]
    [PDF] static zoning division elevator traffic simulation using agent-based ...
    An optimal dispatching control for elevator systems was proposed by Cassandras and Pepyne (1997) targeting up-peak traffic. They implemented a scheduling.
  29. [29]
    What are the latest innovations in elevator technology? - KONE U.S.
    Using artificial intelligence to optimize elevator performance, intelligent dispatch can predict which floors will be busiest at certain times and adjust ...Elevator Destination Control... · Intelligent Elevator... · Ai-Powered Predictive...Missing: 2020-2025 | Show results with:2020-2025<|separator|>
  30. [30]
    Top Elevator Modernization Trends for 2025
    Sep 24, 2024 · From improved control systems to connectivity with Internet of Things (IoT) applications, modern elevators are becoming more efficient, safer, ...
  31. [31]
    Recent Trends in Elevator Modernization
    Aug 26, 2025 · The way these systems work is they incorporate a network of sensors and data analytics to monitor elevator performance in real-time.
  32. [32]
    [PDF] An AI-Based Approach to Destination Control in Elevators
    To reduce uncertainty regarding destinations of passengers, two approaches are currently being pursued: (1) dynamic zoning and (2) des- tination control. In ...<|control11|><|separator|>
  33. [33]
    Optimization of Elevator Standby Scheduling Strategy in Smart ...
    Deep reinforcement learning with asynchronous actor–critic (A3C) networks achieved near-optimal dispatching under complex traffic conditions [9]. IoT-based ...
  34. [34]
    Destination Dispatch | Multi Elevator Installation | Otis USA
    You're in control ; Compass Infinity™. AI elevator dispatching. Predict and continually adapt to your building's traffic patterns. Explore Compass Infinity.
  35. [35]
    [PDF] THE DESTINATION CONTROL SYSTEM FOR OPTIMIZED PEOPLE ...
    KONE Destination uses destination floors and passenger numbers to improve elevator efficiency, leading to shorter travel times, fewer stops, and uncrowded cars.
  36. [36]
    Responsible technology for Elevators & Escalators | Schindler U.S.
    Utilizing AC and PM gearless motor technology, the efficiency of the elevator hoisting machine is increased, allowing for reduced energy consumption. Power ...Destination Technology · Power Factor 1 Regenerative... · Escalator TechnologyMissing: KONE scanning post- 2020
  37. [37]
    Green, all the way up! - KONE Corporation
    KONE's eco-efficient technological innovations come into play here as well. Modernizing an elevator pumps up energy savings up to 70%. The gains are also ...Missing: Schindler scanning post-
  38. [38]
    [PDF] Corporate Responsibility Report 2020 | Schindler Group
    Jun 17, 2021 · Elevator technology, too, can optimize a building's energy performance. ... The design of an energy-efficient building is a complex.Missing: KONE scanning
  39. [39]
    What algorithm is used by elevators to find the shortest path to travel ...
    Sep 22, 2016 · The other answer correctly gives the standard elevator algorithm, which is basically "keep going in the same direction as long as possible and ...Missing: definition | Show results with:definition
  40. [40]
    [PDF] A Comparison of Traditional Elevator Control Strategies - DiVA portal
    May 8, 2015 · 2.1.1 Collective Control. One of the earliest strategies for elevator control systems is the collective control strategy. Here each floor is ...
  41. [41]
    Implementation of dispatching algorithms for elevator systems using ...
    Aug 28, 2006 · The performance of an EGCS is measured by means of several metrics such as the average waiting time ... Four elevator dispatching algorithms are ...
  42. [42]
    Performance Benchmarking and Analysis of Various Elevator ...
    The following paper is a comparative study of various elevator dispatching algorithms, their working, and the results after simulating those algorithms.
  43. [43]
    Solving the Elevator Dispatching Problem Using Genetic Algorithm
    average waiting time must be calculated to determine the fitness value. The average waiting time (Tav) and fitness function (f) are defined as: Tav = 1. M. M.
  44. [44]
  45. [45]
    Modeling the aggregated power consumption of elevators
    Oct 1, 2019 · This paper proposes a bottom-up framework for modeling the aggregated power consumption of a fleet of elevators.
  46. [46]
    Elevator group optimization in a smart building - IEEE Xplore
    The overall energy consumption reduction is 20% in the generic scenario while the service time advantage is 13%. The stability of the algorithm is also shown ...<|control11|><|separator|>
  47. [47]
    Two elevators serving up-traffic - SpringerLink
    Nov 21, 1995 · For moderately light traffic the two elevators tend to travel close together, but for heavy traffic the headways tend to be nearly uniformly ...
  48. [48]
    Scheduling algorithms for modern disk drives
    As a result,. SCAN resists starvation more effectively. (i.e., has lower response time variance) than. SSTF. Several variations of the SCAN algorithm have been.
  49. [49]
    [PDF] Disks: Structure and Scheduling - Columbia CS
    Apr 8, 2013 · SCAN (Elevator) Disk Scheduling. Make up and down passes across all cylinders. Pros: efficient, simple. Cons: Unfair. Oldest requests (furthest ...
  50. [50]
    [PDF] PERSISTENCE: DISK SCHEDULING - cs.wisc.edu
    What advantage does caching in drive have for reads? What advantage does ... Disadvantages? Page 29. SCAN. SCAN or Elevator Algorithm: – Sweep back and ...Missing: limitations | Show results with:limitations
  51. [51]
    [PDF] A Performance Model of Zoned Disk Drives with I/O Request ...
    SSTF can be implemented using the SCAN algorithm [3] in which requests are serviced in order of the disk cylinder number in a particular direction.
  52. [52]
    None
    ### Summary of Energy Consumption of Disk Scheduling Algorithms (SCAN/Elevator) in Low-Traffic/Low-Load Scenarios