Last week we learned about the [increased tail-latency sensitivity of microservices based applications with high RPC fan-outs. Seer uses estimates of queue depths to mitigate latency spikes on the order of 10-100ms, in conjunction with a cluster manager. Today’s paper choice, RPCValet, operates at latencies 3 orders of magnitude lower, targeting reduction in tail latency for services that themselves have service times on the order of a small number of µs (e.g., the average service time for memcached is approximately 2µs).
The net result of rapid advancements in the networking world is that inter-tier communications latency will approach the fundamental lower bound of speed-of-light propagation in the foreseeable future. The focus of optimization hence will completely shift to efficiently handling RPCs at the endpoints as soon as they are delivered from the network.
Furthermore, the evaluation shows that “RPCValet leaves no significant room for improvement” when compared against the theoretical ideal (it comes within 3-15%). So what we have here is a glimpse of the limits for low-latency RPCs under load. When it’s no longer physically possible to go meaningfully faster, further application-level performance Continue reading
Here is a transcript of a slightly adjusted version of that talk.
At Cloudflare we run Linux on our servers. We operate two categories of data centers: large "Core" data centers, processing logs, analyzing attacks, computing analytics, and the "Edge" server fleet, delivering customer content from 180 locations across the world.
In this talk, we will focus on the "Edge" servers. It's here where we use the newest Linux features, optimize for performance and care deeply about DoS resilience.
Our edge service is special due to our network configuration - we are extensively using anycast routing. Anycast means that the same set of IP addresses are announced by all our data centers.
This design has great advantages. First, it guarantees the optimal speed for end users. No matter where you are located, you will always reach the closest data center. Then, anycast helps us to spread out DoS traffic. During attacks each of the locations receives a small fraction of Continue reading
Are you based in Moscow? Cloudflare is partnering with Yandex to produce a meetup this month in Yandex's Moscow headquarters. We would love to invite you to join us to learn about the newest in the Internet industry. You'll join Cloudflare's users, stakeholders from the tech community, and Engineers and Product Managers from both Cloudflare and Yandex.
Talks will include "Performance and scalability at Cloudflare”, "Security at Yandex Cloud", and "Edge computing".
Speakers will include Evgeny Sidorov, Information Security Engineer at Yandex, Ivan Babrou, Performance Engineer at Cloudflare, Alex Cruz Farmer, Product Manager for Firewall at Cloudflare, and Olga Skobeleva, Solutions Engineer at Cloudflare.
Agenda:
18:00 - 19:00 - Registration and welcome cocktail
19:00 - 19:10 - Cloudflare overview
19:10 - 19:40 - Performance and scalability at Cloudflare
19:40 - 20:10 - Security at Yandex Cloud
20:10 - 20:40 - Cloudflare security solutions and industry security trends
20:40 - 21:10 - Edge computing
Q&A
The talks will be followed by food, drinks, and networking.
I am currently using a Windows 10 laptop as my main OS. IMHO Microsoft has
really turned it around as a company in recent years and I no longer think of
them as M$ Virus Corp!! I still live in a terminal session for a lot of my day
and I am not ready give up on Bash just yet. The ability to...
Following on the heels of DockerCon SF, the team is packing their bags and heading to Barcelona for KubeCon + CloudNativeCon EU from May 20- 23. Docker employees, community members and Docker captains will be there speaking about and demonstrating Docker and Kubernetes.
Stop by Booth G14 to learn more about our Docker Kubernetes Services (DKS), which is part of the recently announced Docker Enterprise 3.0. Docker Enterprise 3.0 is the only container platform that provides a simple and integrated desktop-to-cloud experience for both Docker and Kubernetes.
Get Involved with Open Source
Get involved in and learn more about some of the projects Docker has been working on with the Kubernetes community:
containerd – the core container runtime that was recently graduated from the CNCF and is in use by millions of users
Notary/TUF – a project designed to address the key security challenge for enterprises working with containers
Docker Compose on Kubernetes – a recently open-sourced project that enables users to take a Docker Compose file and translates it into Kubernetes resources.
Microsoft took the rare step of issuing security fixes for both the server and desktop versions of Windows that are long out of support, so you know this is serious.The vulnerability (CVE-2019-0708) is in the Remote Desktop Services component built into all versions of Windows. RDP, formerly known as Terminal Services, itself is not vulnerable. CVE-2019-0708 is pre-authentication and requires no user interaction, meaning any future malware could self-propagate from one vulnerable machine to another.CVE-2019-0708 affects Windows XP, Windows 7, Windows Server 2003, Windows Server 2008 R2, and Windows Server 2008. It does not impact Microsoft’s newest operating systems; Windows 8 through 10 and Windows Server 2012 through 2019 are not affected.To read this article in full, please click here
Microsoft took the rare step of issuing security fixes for both the server and desktop versions of Windows that are long out of support, so you know this is serious.The vulnerability (CVE-2019-0708) is in the Remote Desktop Services component built into all versions of Windows. RDP, formerly known as Terminal Services, itself is not vulnerable. CVE-2019-0708 is pre-authentication and requires no user interaction, meaning any future malware could self-propagate from one vulnerable machine to another.CVE-2019-0708 affects Windows XP, Windows 7, Windows Server 2003, Windows Server 2008 R2, and Windows Server 2008. It does not impact Microsoft’s newest operating systems; Windows 8 through 10 and Windows Server 2012 through 2019 are not affected.To read this article in full, please click here
HPE has agreed to buy supercomputer-maker Cray for $1.3 billion, a deal that the companies say will bring their corporate customers high-performance computing as a service to help with analytics needed for artificial intelligence and machine learning, but also products supporting high-performance storage, compute and software.In addition to bringing HPC capabilities that can blend with and expand HPE’s current products, Cray brings with it customers in government and academia that might be interested in HPE’s existing portfolio as well.[ Now read: Who's developing quantum computers ]
The companies say they expect to close the cash deal by the end of next April.To read this article in full, please click here
HPE has agreed to buy supercomputer-maker Cray for $1.3 billion, a deal that the companies say will bring their corporate customers high-performance computing as a service to help with analytics needed for artificial intelligence and machine learning, but also products supporting high-performance storage, compute and software.In addition to bringing HPC capabilities that can blend with and expand HPE’s current products, Cray brings with it customers in government and academia that might be interested in HPE’s existing portfolio as well.[ Now read: Who's developing quantum computers ]
The companies say they expect to close the cash deal by the end of next April.To read this article in full, please click here
Hewlett Packard Enterprise has reached the top of the HPC sector in an unconventional way, but one that could prove truly transformative for its new acquisition, legendary supercomputer maker Cray. …
If you’ve landed on this page, you likely already have a good idea of what ngrok is and what it does. For those that don’t, the reader’s digest version is that it’s a simple way to securely tunnel to a device that sits behind a firewall/NAT device. It’s a slick implementation that is easy to install and allows a few different tunneling options. For the purpose of this blog, we’re using ssh and eliminating the need for port forwarding on the firewall.
Here are step-by-step instructions for turning up ngrok ssh services on Cumulus Linux. Note that these instructions work on the default VRF. You’ll need to take additional configuration steps to get this to work on Cumulus Linux with mgmt VRF enabled.
First, install the unzip package from the repo
Then wget the ngrok application, or optionally add the appropriate repo to your /etc/apt/sources.list and use apt to pull the package. You’ll obviously want to find the appropriate package for your switch (x86 or ARM).