Developing Distributed Systems Quiz

Many distributed systems are unnecessarily complex due to mistakes requiring later corrections.
Common false assumptions often contribute to this complexity.
Network reliability is not always guaranteed.
Network security is sometimes an issue.
Networks aren't always homogeneous (consistent).
Network topology may change.
Latency is often not zero.
Bandwidth is often not infinite.
Transport costs may not be negligible.
There might not be a single administrator.

High-performance distributed computing began with parallel computing.
Multiprocessor/multicore systems vs. Multicomputers:
- Shared memory systems use an interconnect to share memory among processors.
- Private memory systems have individual memory for each processor using an interconnect.

Multiprocessors are easier to program than multicomputers.
A shared-memory model can be implemented on multicomputers.
Virtual-memory techniques allow mapping pages from multiple processors to a virtual address space.
If a process needs a page from another processor, the OS can fetch it.
Distributed shared memory systems encountered performance limitations.
They are now widely abandoned due to better methods.

Cluster computing involves interconnected high-end systems on a LAN.
Systems are typically homogeneous (same OS, similar hardware).
A single managing node coordinates the cluster.
Compute nodes perform tasks under management from a master node.
Components of applications are run on the cluster nodes, and libraries are often used in parallel.
Supercomputers are often examples of clusters.

Grid computing extends cluster computing to a wider area.
Grids use many nodes from various locations—making them heterogeneous.
Grid computing spans several organizations.
Virtual organizations enable collaboration by allowing authorized users access to resources.

Grid computing has four layers: Fabric, Connectivity, Resource and Collective.
Fabric: interfaces to the physical resources.
Connectivity: communication and transaction protocols, authenticating resources.
Resource: supports single resource management, processes and data.
Collective: manages access to multiple resources.
Application layer: grid applications run in an organization.

Cloud computing is a layered architecture.
Software, Web services, multimedia, business apps at the highest layer.
Application frameworks (like Java, .Net, Python) are in the next layer, sitting on top of the layer below.
Storage (databases) are on the next layer, alongside platforms.
Computation (virtual machines), storage, the next layer.
At the bottom, fundamental hardware resources: CPU, memory, disk, bandwidth.
Datacenters hold the physical hardware.

Hardware: Includes processors, routers, power, and cooling, not usually seen by customers directly.
Infrastructure: Uses virtualization, allocates and manages virtual storage devices, and virtual servers.
Platform: Provides higher-level abstractions for storage; for instance, Amazon S3 provides an API, organizing files into buckets.
Application: Contains actual applications (like office suites).

Distribution: Devices are networked, distributed, and accessible in a transparent manner.
Interaction: Interaction between users and devices is highly unobtrusive.
Context awareness: The system is aware of a user's context to optimize interaction.
Autonomy: Devices operate independently.
Intelligence: The system handles many actions.

Sensor data is stored and reported to an operator in one possible configuration.
Sensors can process and store data—then send data to the operator.
Querying sensors for information is another extreme configuration.

Many sensor networks need to operate on strict energy budgets.
Duty cycling is used to save energy.
Sensor nodes only use energy to transmit data and respond to queries periodically.