Template

Backgrond

• In any environment, application downtime is always lurking on the edges.
• 2019 marks the year for 5G, which was eventually rolled out for commercial services to consumers.
• Most arguments about “What is the essence of …?” do more to reveal the prejudices of the participants than to uncover any objective truth about the topic of discussion. Attempts to define the term “object-oriented” precisely are no exception.
• However, it is non-trivial to turn the simple idea of location-aware (multipath) packet scheduling into a end-to-end practical system.
• Naively replicating hot objects to all cache nodes incurs high overhead for cache coherence. On the other hand, simply partitioning hot objects between the cache nodes would cause the load to be imbalanced between the cache nodes.

Introduction

Limitation:

• X have some essential limitations:
• X can either introduce stability concerns or suffer a large performance penalty.
• At a high level, we find that despite its high throughput, today’s 5G has several limitations such as large performance variations, vulnerability to obstructions, and frequent hand-offs even during low mobility.
• But none of them studies mmWave in commercial 5G context on smartphones.
• Despite the proliferation of numerous algorithms, however, there appears to be a lack of clarity and consensus across these solutions on several fronts.

Summary:

• We summarize our key findings as follows.
• We make the following contributions in this paper.
• We present timely measurement findings of mmWave 5G performance on smartphones with key insights.
• Our results provide an important “baseline” for studying how 5G performance evolves.

• Our study identifies key research directions on improving 5G users’ experience in a cross-layer manner.

• The remainder of this paper is organised as follows.

Related Works

• Researchers have demonstrated the feasibility of deploying mmWave in data centers [12, 44, 46], indoor [1, 2, 5, 11, 25, 37, 39, 41], and outdoor environments [28, 31–33, 43, 45], as well as have conducted many studies on beamforming and beam tracking [8, 27, 34].
• There exist a plethora of work on cellular network measurement.
• There is a large body of literature that considers mobile devices’ energy consumption based on bandwidth, signal strength, simultaneous use of inter- faces, and so on

Symbol

• Let C denote X.
• We use Xj as the contribution to loss rate of interface j.
• We use two types of COTS 5G-capable smartphones: Motorola Moto Z3 and Samsung Galaxy S10 5G (SM-G977U), henceforth referred to as MZ3 and SGS10, respectively.
• Thus, unless otherwise mentioned, all 5G results presented in the paper are from SGS10.
• Two different joins of the same pair of types must each be a subtype of the other.
• We then develop a custom HTTP(S) client that issues 8 parallel byte-range requests each fetching 1/8 of the file over 5G.
• The “w/ SS” boxes correspond to short (25-second) measurements from the very beginning of TCP connection establishment.
• We repeat the experiments for 10 times and observe qualitatively similar results, with one representative run illustrated in Figure 3(a).
• Unless stated, 95-percent confidence intervals, calculated using t-distribution, are omitted from graphs since they are typically very small, around 1%.
• Global mobile data is expected to increase 10-fold from 2014 to 2019.
• Whenever Green-Bag arranges a new segment, it chooses the link mode that consumes the least energy out of three link modes: dual-link mode, LTE-only mode, and WiFi-only mode, in a way that the remaining portion of a video file can be transferred in the link mode without incurring any playback interruption further according to the current bandwidth estimation.
• We term services that converge to a track with declared bandwidth closer to available bandwidth as more aggressive.
• The IEEE 802.11ad standard with its 2 GHz-wide channels provides data rates of up to 6.7 Gbps, a multi-fold increase over legacy WiFi throughput.

Instruction

• We first exemplify two most common obstructions: human body and hand.
• Using the typing rules in Figure 23-1, convince yourself that the terms above have the types given.
• Readers may want to refer to 5.2 to refresh their intuitions about Church encodings.
• For each server, we repeat the file download for 3 times at both locations, and measure the average throughput.
• We now walk through Algorithm 1, which combines chunk-based transfer (§4.1), subflow completion time balancing (§4.2), and adap- tive reinjection (§4.4).
• We describe each component in turn.
• Without loss of generality, we list the keys in monotonically decreasing order of popularity, i.e.,
• We define PASR (peak-to-average size ratio) to be the ratio between the 95𝑡h percentile chunk size and the average chunk size within a track.
• We model a video as a set of consecutive video segments or chunks
• We thus set up a multipath proxy running MPTCP v0.90 on a server with 64-core 2.6GHz CPU, 128GB memory, and 64-bit Ubuntu 14.04 installed.

Algorithm

• Specifically, we begin by probing two ratios adjacent to the current ratio, one slightly lower and one slightly higher, and proceed our search in the direction where we observe higher throughput.
• These process types are (in order of importance):

Conclusion

• They follow a similar trend except that when TCP slow start is considered, the throughput becomes lower.
• The above results highlight the need for cross-layer efforts that improve 5G performance under (even low) mobility.
• The somewhat surprising results make us realize that HTTP(S) download is very different from iPerf bandwidth probing.
• However, 5G throughput exhibits much higher variations than 4G despite the presence of LoS.
• The above results highlight the need for cross-layer efforts that improve 5G performance under (even low) mobility.
• While these values have worked well for our experiments, refining them to improve performance remains a subject for future work.
• Since this proposal deals with other layers than ours, both approaches are complementary and the advantages are cumulative.