Friday, November 28, 2014

Links of the day 28 - 11 - 2014

Today's links 28/11/2014: #Cloud Ressource Marketplace, Stream / Even processing, #bigdata benchmark, Facebook Fabric, Software Defined #IaaS

  • Deutsche Börse Cloud Exchange : market place for trading Compute, Memory and Storage. Interesting but the market place size looks really tiny, the available resources advertised could be added in just one day of daily operation of AWS. 
  • Flafka : Apache Flume Meets Apache Kafka for Event Processing
  • Terraform : common configuration to launch infrastructure — from physical and virtual servers to email
  • Facebook fabric: in depth analysis of Facebook data-center network fabric and its implication. 
  • Big Bench : Industry standard Bigdata analytic.


Thursday, November 27, 2014

Links of the day 27 - 11 - 2014

Today's links 27/11/2014: #statistical #programming, SMP #python, Stumbleupon #bigdata infra
  • Statistical Programming : Statistical Program Analysis and Synthesis, use github and other massive codebase to probabilisticaly fill in the blank in NEW code, Really cool [video]
    • JSNICE : statistical renaming, type inference and deobfuscation derived from the work above
  • SMP python : augment new or existing (Python) serial scripts for scalability across parallel hardware
  • Stumbleupon BigData infrastructure : lot of open source and moving pieces

Wednesday, November 26, 2014

Links of the day 26 - 11 - 2014

Today's links 26/11/2014: #linkedin Goblin, HPC with SR-IOV, #startup hotness

  • Comet: Realizing High-Performance Virtualized Clusters using SR-IOV Technology
  • Goblin : linkedin tool for simplify big data ingestion for Hadoop-based warehouses
  • Hot Startup: startup investment trends using crunchbase data.

Tuesday, November 25, 2014

Links of the day 25 - 11 - 2014

Today's links 25/11/2014: #AWS reinvent video , Opensource #CPU, messaging.


  • AWS Re:Invent video : 215 videos of 2014 Amazon event. 
  • Open Source Parallel CPU : feel deja vu with some hints of connection machine 
  • Kayos : fast, low cost, fault tolerant messaging and durable queueing system that offers predictable performance and can take advantage of high end dedicated hardware as well as unreliable, commodity infrastructure like EC2.... Wow , even if they achieve just half of that it would be great :)

Monday, November 24, 2014

Links of the day 24 - 11 - 2014

Today's links 24/11/2014: algorithmic design, storage startup and cc-numa
  • Algorithmic design manual : book on how to design algorithms, and analyze their efficacy and efficiency.
  • Primary Data demo : data virtualization solutions that improve enterprise efficiency and agility.
  • Bull Mesca BCS Systems : An expandable SMP node for memory-hungry applications using CC-NUMA. Up to 128 cores - 4 modules with IB - 16 NUMA nodes - 3 NUMA levels - RAM 2 TB

Friday, November 21, 2014

Links of the day 21 - 11 - 2014

Today's links 21/11/2014: Shared memory system, Debunking handbook, #IaaS price war

  • Large shared memory system : Numascale, Supermicro, and AMD have introduced the world's largest shared memory system to date. The system features 5184 CPU cores and 20.7 TBytes of shared memory 
  • debunking handbook : a guide to debunking misinformation, offers practical guidelines on the most effective ways of reducing the influence of myths.
  • IaaS price war: how amazon use low level entry IaaS price to attract customer and transition them to higher value more sticky services inorder to increase revenue and retention.

Thursday, November 20, 2014

Links of the day 20 - 11 - 2014

Today's links 20/11/2014: tracing, messaging, queue, cache is the new ram

  • Tracing summit : Tracing Summit 2014 held in DĂĽsseldorf, Germany, on October 13, 2014, video are also available here
  • Operating Apache Samza at Scale : how do leverage samza under kafka to scale your messaging infrastructure.
  • Cache is the new ram : with the rise of in memory database CPU cache becomes the next frontier ( but what's next ? cpu register is the new cache ?? ) 
  • Queues Don't Fix Overload : on why queue are great but not the silver bullet and what are their limitation.

Wednesday, November 19, 2014

Links of the day 19 - 11 - 2014

Today's links 19/11/2014: #resiliency, #cloud , distributed system, #stream processing


Tuesday, November 18, 2014

Links of the day 18 - 11 - 2014

Today's links 18/11/2014: #microservices, #cloud , oracle CPU, automated deployment, #RDMA, Intel Omni-path
  • Colossus : a lightweight framework for building high-performance applications in Scala that require non-blocking network I/O. In particular Colossus is focused on low-latency stateless microservices by tumlbr
  • M7: Next Generation Oracle Processor [ slides
  • Appolo : shared internal deployment service at Amazon and its public pendant AWS code deploy 
  • RDMA in the cloud : Vmware demonstrate RDMA for virtual machine in cloud env but as usual with vmware presentation the graph scale are poorly chosen... 
  • Omnipath : Intel is gunning at infiniband with its new 100 Gbps fabric.

Monday, November 17, 2014

Virtual Cores tech and partitioning hypervisor

A couple of weeks ago Soft Machines came out of stealth mode and introduced the Variable Instruction Set Computing (VISC). VISC attempts to avoid the difficulties of scaling multiple threads in hardware by providing a framework in which workloads that appear sequential to the operating system are then scheduled across a set of virtual cores in hardware.

VISC-3-640x448.png
What is interesting here is that most often virtualization has been leverage to deliver fine grained (over)subscription from your system. The disadvantage is that with such technology you have often a penalty to pay. It is on average between 5 to 10% but it can be less or worse depending of the workload and over-subscription scenario. One alternative for separating critical applications while guaranteeing performance and isolation exist: partitioning hypervisor such as jailhouse.

Jailhouse can create asymmetric multiprocessing (AMP) setups on Linux-based systems. What it effectively does is partition the physical hardware into cells and guarantee the isolation between them. Effectively it use some cores to run a Linux based hypervisor to manage and partition the rest of the hardware. Each OS or virtual machine effectively run only on its dedicated core and hardware. The typical workloads we expect to see in non-Linux cells are applications with highly demanding real-time, safety or security requirements. However the trade-off is that you are under subscribing your system in order to guarantee high performance and isolation.

Now enter VISC, with such technology you could in theory eliminate the under subscription problem while retaining the performance and isolation characteristics of partitioning hypervisor. What would be interesting to see if this cpu architecture will be able to leverage this type of hypervisor technology to gain momentum within the cloud. Workload such as network virtualization function or virtual network systems would greatly benefit from such approach as the overhead of classic virtualization is a significant limiting factor.

Links of the day 17 - 04 - 2014

Today's links 17/04/2014: distributed transaction, data center fabric, SSD friendly alternative to bloom filter.

  • Granola: Paper Summary: Granola, Low overhead distributed transaction coordination
  • Consistency and coordination : you do not always need coordination to be consistent while retaining ACID property [paper]
  • Quotient filter : SSD optimized alternative to bloom filter.
  • Facebook datacenter fabric : facebook we broke the network up into small identical units – server pods – and created uniform high-performance connectivity between all pods in the data center in order to maintain performance, scalability and reliability. 

Friday, November 14, 2014

Links of the day 14 - 11 - 2014

Today's links 14/11/2014: NUMA, distributed transaction, NVMe, Bimodal IT
  • Scale out Numa : an architecture, programming model, and communication protocol for low-latency, distributed in-memory processing by layering an RDMA-inspired programming model directly on top of a NUMA memory fabric via a stateless messaging protocol. 
  • Calvin : Paper Summary - Calvin, Distributed transactions for database systems
  • NVMe and Fabric : performance impact of fabric with NVMe 
  • Bimodal IT : Simon Wardley explain why Gartner's bimodal IT concept is an half baked old concept in new clothes


Thursday, November 13, 2014

Links of the day 13 - 11 - 2014

Today's links 13/11/2014: Mellanox ConnectX4, Immutable infrastructure and ARM server
  • ConnectX4 : EDR 100Gb/s InfiniBand and 100Gb/s Ethernet, 150M messages/second, impressive numbers from Mellanox. 
  • Fugue: immutable infrastructure delivering Automating the creation and operations of cloud infrastructure, Short-lived and simplified compute instances
  • Custom Cloud Arm Server : online lab design its own ARM based server for its cloud infrastructure.

Wednesday, November 12, 2014

Links of the day 12 - 11 - 2014

Today's links 12/11/2014: all about high performance distributed networking and communication


Tuesday, November 11, 2014

Links of the day 11 - 11 - 2014

Today's links 11/11/2014: Kafka and infrastructure, Product vision building, and software performance techniques
  • Infrastructure for Data Streams : how to persist incoming data stream with guaranteed data availability and redundancy using Kafka at Chartbeat
  • Idea stack : Idea Stack exercises and use case studies, illustrating how to build a vision for a product before developing it.
  • Below an image of the various software improvement techniques : 

Monday, November 10, 2014

Links of the day 10 - 11 - 2014

Today's links 10/11/2014: cluster CI , DHT, Multiqueues, blockchain ssh key pki
  • Cluster Runner: fast easy test feedback for your continuous integration system. Its always a challenge to retain speed and agility as your test and team growth. Cluster Runner helps solve that. 
  • DHT routing table maintenance: performance improvements of the DHT by the bitorrent crowd
  • MultiQueues : multiple sequential priority queues that outperform previous more complicated data structures
  • emcssh : blockchain based secure, decentralized management of PKI.

Thursday, November 06, 2014

On the emergence of hardware level API for dis-aggregated datacenter resources


The technologies enabling the modular dis-aggregated data-center concept are reaching maturation point as demonstrated by the latest technology showcase RSA from Intel or to a lesser extent FusionCube / FusionSphere from Huawei. The needs for such technologies arise from the fact that current cloud and data-center technology does not and cannot fulfill all the demands of cloud users for multiple reasons. On one hand, as the number of cores and amount of memory on servers continues to increase (over the next few years, we expect to see servers with hundreds of cores and terabytes of memory per server commonly used), leasing an entire server may be too large for many customer’s needs with resources wasted. On the other hand, with the emergence of a broad class of high-end applications for analytic, data-mining, etc., the available amounts of memory and compute power on a single server may be insufficient.

Moreover, leasing cloud infrastructure resources in a fixed combination of CPU, memory, etc. are only efficient when the customer load requirements are both known in advance and remain constant over time. As neither of these conditions are met for a majority of customers, the ability to dynamically mix-and-match different amounts of compute, memory, and I/O resources is the natural evolutionary step after the hyper-converged solutions.

The objective here is to address the gaps that allow us to go beyond the boundaries of the traditional server, effectively breaking the barrier of using a single physical server for resources. in other words, we will be able to provision compute, memory, and I/O resources across multiple hosts within the same rack, while being consumed dynamically by varying quantities at run-time instead of in fixed bundles. This will effectively enable a fluid transformation of current cloud infrastructures targeting fixed commodity sized physical nodes to a very large pool of resources that can be tapped into and consumed without classical server limitation.


Intel has been advertising its RSA stack for a while and it is finally becoming reality, however, the real interesting part is not the technology. Indeed, to a certain extent, a lot of technology already exists and enables us to implement resource pooling. We demonstrated that it was already feasible to deliver cloud memory pooling in the Hecatonchire Project as well as with other vendors such as TidalScale or ScaleMP, who already offer compute aggregation. However, the last two solutions are monolithic and lack the flexibility needed to be used with the cloud consumption model and as a result, they are confined to a niche market.

What can really kick the dis-aggregated model into top gear is that Intel has now teamed up with a couple vendors and has already, created a draft hardware API specification called Redfish. Such API can be leveraged by a higher level of the stack thus allowing more intelligent, flexible and predictable resource consumption of how, where, and when workloads (VMs, containers, standard processes/threads) get scheduled onto that hardware. In a certain way this then enables Mesos / Kubernetes to deliver enhanced scheduling for every hardware aspect.

This brings some interesting capabilities to existing cloud technologies, cores and memory which then can be dynamically reallocated across the workload and arguably, , it greatly reduces the need for load balancing via live migration. You would then dynamically re-allocate the resource underneath (core, memory) rather than the whole system, thus making such process more robust and less error prone.

On the container side it would solve a lot of security headache the community is now facing. Rather than going with the physical->virtual->container route, you could simply run physical->container with a fine grained per core allocation using RSA / Redfish. Effectively you would provide fine grained subscription from your system in order to get maximal separation and performance guarantees. One can use this for separating critical applications while guaranteeing performance and isolation and indeed something we can already do now with jailhouse, at the cost of under subscribing your system.

If Intel is successful in disseminating (or having the other vendors standardize around it’s Hardware API), it would allow the technology to leap forward, as it’s biggest enemy is the difficulty to port across management API from one fabric, compute, I/O, storage model to another.


Wednesday, November 05, 2014

Links of the day 05 - 11 -2014

Today's links 05/11/2014: #Docker and #LXD, scalability rule, Google #containers engine

  • LXD : Shuttleworth announce LXD, a secure container technology design to address the isolation and security concern of exiting solution.
  • Scalable commutativity rule : Whenever interface operations commute, they can be implemented in a way that scales
  • Container engine: interesting that Google offer multiple container per VM management (1:M) while all its concurrent only offer a 1:1 mapping. 

Tuesday, November 04, 2014

Links of the day 4 - 11 - 2014

Today's links 04/11/2014 : infra conf mgmt tools,distributed DB  sharding and replication, smart scheduling
  • ICMT eMAG : Infrastructure Configuration Management Tools eMag by infoq 
  • Dynomite : a sharding and replication layer for database inspired by Cassandra and Dynamo paper from the Netflix crowd. 
  • Smart Scheduler : ovirt smart scheduler for VM placement and migration in order to maximize occupation. The nice part is that it use a migration planner that plan out how to reach the optimal placement in multiple steps via migration, a little bit like the Hanoi tower problem but NP hard. 

Monday, November 03, 2014

Links of the day 03 - 11 - 2014

Today's links 03/11/2014: Intel Netburst, Fig and Crypto banking
  • Replay : well hidden feature of the Intel Netburst architecture. Interesting to see how much impact this non advertised feature had on the performance of the system. 
  • FIG : Fast, isolated development environments using Docker [screen cast
  • Crypto bank : they are trying to build the first cryptocurrency bank. A step up from bitcoin wallet.