Efficient Resource Allocation for Multi-Tenant Monitoring of Edge Infrastructures

Abstract

By relying on small sized and massively distributed infrastructures, the Edge computing paradigm aims at supporting the low latency and high bandwidth requirements of the next generation services that will leverage IoT devices (e.g., video cameras, sensors). To favor the advent of this paradigm, management services, similar to the ones that made the success of Cloud computing platforms, should be proposed. However, they should be designed in order to cope with the limited capabilities of the resources that are located at the edge. In that sense, they should mitigate as much as possible their footprint. Among the different management services that need to be revisited, we investigate in this paper the monitoring one. Monitoring functions tend to become compute-, storage-and network-intensive, in particular because they will be used by a large part of applications that rely on real-time data. To reduce as much as possible the footprint of the whole monitoring service, we propose to mutualize identical processing functions among different tenants while ensuring their quality-of-service (QoS) expectations. We formalize our approach as a constraint satisfaction problem and show through micro-benchmarks its relevance to mitigate compute and network footprints.

Monitoring Probe Deployment Patterns for Cloud-Native Applications: Definition and Empirical Assessment

“…We started from the papers that propose monitoring and probe deployment approaches for multi-tenant and technology-heterogeneous cloud environments [31], [28], [32], [29], [30], [37], [46], the most used cloud monitoring tools [40], [39], [43], [44], [42], and our experience, to identify and distill a set of relevant features for probes deployment. We discuss below the semantics of the considered features.…”

“…-Holder Type: it represents the holder type [28] * Target: the holder is the target of the monitoring activity, that is, the holder hosts both the target and the monitoring probes * External Unit: the holder is an external object which monitors the target from the outside (e.g., a sidecar container [28], [29], [43], [39], [40], [44], [42]) -Probe Multiplicity: it defines the number of probes that can be executed within the unit [30], [29] * Single-probe: only one probe can be executed * Multi-probe: one or more probes can be executed -Holder Sharing: it defines if the holder can be shared among multiple users [31], [32], [46], [37] * Reserved Holder: the holder is reserved to a single user * Shared Holder: the holder can be shared among users…”

An Overview of Service Placement Problem in Fog and Edge Computing

“…Due to specific hardware, privacy requirements or given policy, some components can be restricted to be deployed on specific areas (zone) or devices. Locality constraints can be based on: a set of Fog nodes [154,164]; a specific geo-spatial location using GPS for instance [124], impose a co-localization of components [3], etc.…”

Cloudlet Computing: Recent Advances, Taxonomy, and Challenges

“…For instance, important data, such as personal information, corrupt data, blurry videos, and sensors producing data in bulks, can be filtered and processed locally over cloudlets, which will forward only the required data to the enterprise cloud [57]. The Akamai Cloudlet [51] and Nokia MEC [58] are beneficiaries of first-line security Cloud outage is a major concern for users. The most common operational failures include interruption of service to enterprises, frozen applications, and violating committed QoS to users.…”

“…A distributed execution application framework is presented in [51] in which a compute-intensive code is offloaded to a cloudlet or cloud at run time or a mobile application is synchronized with another remotely executed application over the cloud. This technique encourages the live migration of a mobile application to cater to compute node mobility.…”