Notes

Reliability is the ability of a system to remain operational over time, i.e., to offer the service it was designed for. Cloud Hardware and software fails. In this note, we will try to find methods to analyze and predict when components fail, and how we can prevent this problem. Defining the vocabulary Reliability and Factors of Influence Reliability is the probability that a system will perform its intended function without failure over a specified period of time. There are many factors that influence this value: ...

We need to find an efficient and effective manner to allocate the resources around. This is what the resource management layer does. Introduction to the problem What is Cluster Resource Management? Most of the time, the user specifies an amount of resources, and then the cluster decides how much to allocate (but approaches like (Delimitrou & Kozyrakis 2014), do it differently). There are mainly two parts in cluster resource management: Allocation: deciding how many resources an application (techniques for this is presented in Cluster Management Policies. Assignment: from which physical machine you can effectively put the application. Types of management architectures We mainly divide the management architectures in three ways: ...

pWith confidential computing we want to guarantee confidentiality and integrity of a user’s computation running on a remote (cloud) system, including: The program Its inputs and outputs Intermediate state, control flow, etc. Even if do not trust the cloud provider! Usually it is easy to guarantee that kind of privacy if you are storing or communicating using encryption methods (see Asymmetric Cryptography, Block Ciphers), but it’s difficult to do so if the program is running. ...

Containers In this note, we introduce the famous docker containers. We also explore how #Linux Containers are implemented, and some parts of how #Docker works. What is a Container We have explored Virtual Machines in some past section. Containers do not virtualize everything, but just the environment where the application is run. This includes: Libraries Binaries We can see it as a lightweight VM, even if they do not offer the full level of isolation of traditional virtual machines. ...

CDNs are intermediary servers that replicate read intensive data to provide better performance when user requests them. A close relative of CDNs is edge computing (e.g. gaming stations) where lots of computation is done directly close to the user. Types of CDNs Mainly three types of CDNs: Highly distributed ones. -> Akamai Database based ones. Ad-hoc CDNs. Advantages and disadvantages The main reason we use CDNs is to lower the value of latency: we are in fact bringing the data closer to the user. We have much less data in length to be transmitted. Yet we have some disadvantages too: ...

CIAA principles of security We have already outlined these principles in Sicurezza delle reti and talked about the concepts of authentication and integrity. Here we try to deepen these concepts and delve a little bit more on the attack vectors. This note mainly focuses on the principles summarized by the acronyms CIA and AAA. Confidentiality This is one concerns about the secrecy of the sent message. We do not want others to be able to access and read what we are doing. ...

Mixture of Experts There is a gate that opens a subset of the experts, and the output is the weighted sum of the outputs of the experts. The weights are computed by a gating network. One problem is load balancing, non uniform assignment. And there is a lot of communication overhead when you place them in different devices. LoRA: Low-Rank Adaptation We only finetune a part of the network, called lora adapters, not the whole thing. There are two matrices here, a matrix A and B, they are some sort of an Autoencoders, done for every Q nd V matrices in the LLM attention layer. The nice thing is that there are not many inference costs if adapters are merged post training: ...

We have resources, but need to know how to assign these to the jobs that need them. This note presents some of the most common resource management policies for cloud clusters. Introduction to cluster management How can we allocate the resources in a cluster in an efficient manner? How can we allocate resources fairly? Two step allocations There are two main kinds of allocation: first you need to allocate resources to a process, then allocate the process physically in the cluster. We have seen an example of a working infrastructure in Cluster Resource Management. ...

This allows us to extend the memory hierarchy (see Memoria) that we have today. The problem is that we have heterogeneous access patterns specifications and hardware. One of the main trends is disaggregation: we want to be able to scale different resources independently. Introduction to CXL (Compute Express Link) This is a new part of the memory hierarchy. NVM is a kind of non volatile memory that is used as a storage device that is close to the device (others are network attached or slower than network part anyways. It is persistent and has low latency. It is used in the memory hierarchy to extend the memory capacity. ...

Queueing theory is the theory behind what happens when you have lots of jobs, scarce resources, and subsequently long queues and delays. It is literally the “theory of queues”: what makes queues appear and how to make them go away. This is basically what happens in clusters, where you have a limited number of workers that need to execute a number of jobs. We need some little maths to model the stochastic process of request arrivals. ...