Tales from a Trading Desk: Resiliency Made Easy

By By Mike Stolz, Vice President of Architecture, GemStone Systems

November 5, 2007

To Keep It Running, Keep It Simple

Today’s electronic world has resulted in a major shift in terms of how organizations think about resiliency. Firms of all sizes are not only faced with the challenge of determining how resilient their mission-critical systems need to be, but also how they can efficiently and cost-effectively architect a “resilient” system. With millions of dollars per minute running through electronic channels 24×7, traditional high availability and disaster recovery notions are no longer good enough.

For the past five or 10 years, high availability meant the ability to recover from a server outage within about 15 minutes. Solutions like N+1 clustering and storage area network replication were perfectly acceptable. Today, however, the recovery time associated with these high-availability schemes can cause millions of dollars in lost revenue.

To avoid a potentially massive loss in revenue and efficiency in today’s fast-moving markets, firms must significantly improve their enterprise resiliency. Continuous availability is now the acceptable level of resiliency, and it is quite common in Web-based or other electronic channels to use load-balanced, hot/hot clusters of servers to serve up the business logic. These servers typically are stateless in design, so it is easy to add or remove servers and re-balance the work load. The difficult part is designing a resiliency architecture that makes the data behind those business services hot/hot.

Meeting the Resiliency Challenge: An EDF Approach

The best way to provide nearly 100 percent uptime for data and deliver maximum resiliency is by using data management middleware to ensure there are multiple consistent copies of the active business objects in-memory at all times. As firms strive to get ever closer to 100 percent uptime and ensure resiliency, distributed data caching is gaining in popularity.

Solutions such as an enterprise data fabric (EDF) are ideal for meeting those demands. Presented as a simple HashMap API, the EDF programming paradigm is extremely simple and familiar yet delivers maximum value behind the scenes: You simply “put” your state into the HashMap and, under the covers, the middleware takes care of replicating this business object to multiple additional servers.

Sounds easy, right? It is — until you start to think about the various failure modes, guarantees around zero data loss, low latency and scalability. That’s what makes a product like an EDF worth its weight in gold. The most difficult parts of data management are resiliency, scalability, throughput, latency and dataset size — and you have to get it right. Every time.

By deploying an EDF, firms will benefit from a very fast, highly scalable distributed caching system. An EDF is designed for use in many diverse data management situations, but is especially useful for high-volume, latency-sensitive, mission-critical, transactional systems. There are several critical features to consider when evaluating an EDF, including:

  • Language neutrality. This is the ability to access the data natively from common programming languages like Java, C++ and C#.
  • Cache coherency, which is especially important in globally distributed systems.
  • Persistence/overflow so no data is ever lost regardless of circumstances.
  • Highly reliable business object replication, both synchronous and asynchronous, to multiple locations for safe, high-volume transactional environments.
  • Horizontal scalability to thousands of cache nodes.
  • A loosely coupled WAN gateway for long-haul distribution of data.
  • And all of this with continuous availability — never any unexpected down time.

So how does it work? As soon as an application puts data into the cache it is replicated synchronously to at least one additional member of the cache. It also can be replicated to additional members or written to a persistent store, but this can be done on a low priority, asynchronous thread so it doesn’t hold up mainstream processing.

Leveraging Multiple Topologies to Deliver Maximum Value

A true EDF should use three topologies in order to achieve the highest levels of reliability, scalability and speed. The first — and the backbone of the system — is the peer-to-peer topology. In this configuration, everybody knows about everybody else. If a new node joins the distributed system, everybody gets notified, and if a node leaves, everybody gets notified. This enables users to dynamically adjust distributed systems. There is no notion of a “broker” and no single point of failure; fault tolerance is designed right in.

The trouble with peer-to-peer architectures is that they have so much metadata flying around that they can only have limited scale. In most cases, this topology should only be scaled up to about 100 or 200 nodes.

Scalability can be improved by using a second type of topology — client-server — where we elect some of the peers from the peer-to-peer backbone to be servers for client applications (your business logic servers). Each server should be able to manage as many as 100 clients. As there is much less metadata overhead in this topology, it can scale to thousands of nodes.

The third topology is a WAN gateway topology, which can glue together multiple client-server distributed systems. This is an ideal way of creating an enterprise data grid that is globally distributed and appears as one large distributed system, even though it is really many distributed systems glued together.

Appropriate use of these three topologies will enable you to achieve your business requirements around recovery point objective and recovery time objective. Data is replicated across the entire distributed cache, and replication is transactional and performed at the in-memory object level. As soon as an object is put into the cache, it is replicated in-memory to at least one additional node. The data can be replicated to additional nodes either synchronously or asynchronously depending on sensitivity to latency and tolerance for data loss in the event of a catastrophic failure. Write-through to a database or other persistent store is done asynchronously as time permits. In essence, the distributed cache behaves much like RAID for the enterprise.

Additionally, the data can be actively used in both the primary and secondary sites. In fact, the only thing that typically drives the notion of one site even being primary is the external connectivity to the exchanges or ECNs.

Another factor to consider when evaluating an EDF is what we’ll term a “shared nothing” architecture. Because the data in an EDF can be mirrored across multiple nodes in a distributed cache, it eliminates the need for any type of fancy shared storage. In fact, the local disks that are on the blades themselves are often sufficient. In the event that a disk fails, only one node is taken down in the distributed system and there are other nodes alive and ready to take over that workload. Finally, the workload itself is distributed across all the nodes in the distributed system. Exchanges may be split between the two sites and clients will likely be distributed across the two sites, as everything except external connectivity is in a hot/hot configuration.

Let’s walk through the simple H/A recovery process for a single node failure: detect the failure; reconnect the clients; recovery is complete. In total, there is less than 1 second from detection of the failure to complete recovery. A little better than 15 minutes! Because the data is all in-memory in the form of business objects all the time, there is no re-booting, no re-fetching of data and no re-creation of objects.

But what about a catastrophic failure? EDF clusters are virtual, so the nodes needn’t be located close together within the datacenter — they can be on separate subnets, using separate routers, power sources, etc. In fact, some of the nodes actually can be physically located in a different site. Therefore, the notion of losing a “cluster” is non-existent; we’re actually talking about loss of an entire datacenter.

If a disaster occurs and the entire primary data-center fails, the recovery process goes like this: detect the failure; reconnect the exchange at the alternate site; reconnect the clients; recovery is complete. The typical time to recover from point of detection is around 1 second. That’s a huge difference to the typical 1-4 hour disaster recovery time common in business today!

Summary

As distributed computing deployments become the norm rather than the exception, resiliency will become one of the most critical issues facing global corporations. By using an EDF, firms can achieve nearly instantaneous recovery from outages — real business continuity — while simultaneously simplifying their architectures. This one product takes the place of an H/A solution, a shared-storage environment, storage-level replication and wide-area data distribution, removing the need to design a data resiliency architecture for mission-critical systems.

About Mike Stolz

Mike Stolz is vice president of architecture and strategy for financial services at GemStone Systems. In his role, Stolz leverages his expertise in targeting, developing and delivering innovative technology solutions to expand GemStone’s global financial services offering and cultivate its growing capital markets division. Stolz served during the last nine years as director and chief architect of Merrill Lynch’s global markets and investment banking debt division. In this role, Stolz was responsible for the design and development of trading systems and trading support systems for interest rate, credit and asset backed derivatives, as well as FX and repos and fixed income products.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

Mira Supercomputer Enables Cancer Research Breakthrough

November 11, 2019

Dynamic partial-wave spectroscopic (PWS) microscopy allows researchers to observe intracellular structures as small as 20 nanometers – smaller than those visible by optical microscopes – in three dimensions at a mill Read more…

By Staff report

IBM Adds Support for Ion Trap Quantum Technology to Qiskit

November 11, 2019

After years of percolating in the shadow of quantum computing research based on superconducting semiconductors – think IBM, Rigetti, Google, and D-Wave (quantum annealing) – ion trap technology is edging into the QC Read more…

By John Russell

Tackling HPC’s Memory and I/O Bottlenecks with On-Node, Non-Volatile RAM

November 8, 2019

On-node, non-volatile memory (NVRAM) is a game-changing technology that can remove many I/O and memory bottlenecks and provide a key enabler for exascale. That’s the conclusion drawn by the scientists and researcher Read more…

By Jan Rowell

What’s New in HPC Research: Cosmic Magnetism, Cryptanalysis, Car Navigation & More

November 8, 2019

In this bimonthly feature, HPCwire highlights newly published research in the high-performance computing community and related domains. From parallel programming to exascale to quantum computing, the details are here. Read more…

By Oliver Peckham

Machine Learning Fuels a Booming HPC Market

November 7, 2019

Enterprise infrastructure investments for training machine learning models have grown more than 50 percent annually over the past two years, and are expected to shortly surpass $10 billion, according to a new market fore Read more…

By George Leopold

AWS Solution Channel

Making High Performance Computing Affordable and Accessible for Small and Medium Businesses with HPC on AWS

High performance computing (HPC) brings a powerful set of tools to a broad range of industries, helping to drive innovation and boost revenue in finance, genomics, oil and gas extraction, and other fields. Read more…

IBM Accelerated Insights

Atom by Atom, Supercomputers Shed Light on Alloys

November 7, 2019

Alloys are at the heart of human civilization, but developing alloys in the Information Age is much different than it was in the Bronze Age. Trial-by-error smelting has given way to the use of high-performance computing Read more…

By Oliver Peckham

IBM Adds Support for Ion Trap Quantum Technology to Qiskit

November 11, 2019

After years of percolating in the shadow of quantum computing research based on superconducting semiconductors – think IBM, Rigetti, Google, and D-Wave (quant Read more…

By John Russell

Tackling HPC’s Memory and I/O Bottlenecks with On-Node, Non-Volatile RAM

November 8, 2019

On-node, non-volatile memory (NVRAM) is a game-changing technology that can remove many I/O and memory bottlenecks and provide a key enabler for exascale. Th Read more…

By Jan Rowell

MLPerf Releases First Inference Benchmark Results; Nvidia Touts its Showing

November 6, 2019

MLPerf.org, the young AI-benchmarking consortium, today issued the first round of results for its inference test suite. Among organizations with submissions wer Read more…

By John Russell

Azure Cloud First with AMD Epyc Rome Processors

November 6, 2019

At Ignite 2019 this week, Microsoft's Azure cloud team and AMD announced an expansion of their partnership that began in 2017 when Azure debuted Epyc-backed ins Read more…

By Tiffany Trader

Nvidia Launches Credit Card-Sized 21 TOPS Jetson System for Edge Devices

November 6, 2019

Nvidia has launched a new addition to its Jetson product line: a credit card-sized (70x45mm) form factor delivering up to 21 trillion operations/second (TOPS) o Read more…

By Doug Black

In Memoriam: Steve Tuecke, Globus Co-founder

November 4, 2019

HPCwire is deeply saddened to report that Steve Tuecke, longtime scientist at Argonne National Lab and University of Chicago, has passed away at age 52. Tuecke Read more…

By Tiffany Trader

Spending Spree: Hyperscalers Bought $57B of IT in 2018, $10B+ by Google – But Is Cloud on Horizon?

October 31, 2019

Hyperscalers are the masters of the IT universe, gravitational centers of increasing pull in the emerging age of data-driven compute and AI.  In the high-stake Read more…

By Doug Black

Cray Debuts ClusterStor E1000 Finishing Remake of Portfolio for ‘Exascale Era’

October 30, 2019

Cray, now owned by HPE, today introduced the ClusterStor E1000 storage platform, which leverages Cray software and mixes hard disk drives (HDD) and flash memory Read more…

By John Russell

Supercomputer-Powered AI Tackles a Key Fusion Energy Challenge

August 7, 2019

Fusion energy is the Holy Grail of the energy world: low-radioactivity, low-waste, zero-carbon, high-output nuclear power that can run on hydrogen or lithium. T Read more…

By Oliver Peckham

Using AI to Solve One of the Most Prevailing Problems in CFD

October 17, 2019

How can artificial intelligence (AI) and high-performance computing (HPC) solve mesh generation, one of the most commonly referenced problems in computational engineering? A new study has set out to answer this question and create an industry-first AI-mesh application... Read more…

By James Sharpe

Cray Wins NNSA-Livermore ‘El Capitan’ Exascale Contract

August 13, 2019

Cray has won the bid to build the first exascale supercomputer for the National Nuclear Security Administration (NNSA) and Lawrence Livermore National Laborator Read more…

By Tiffany Trader

DARPA Looks to Propel Parallelism

September 4, 2019

As Moore’s law runs out of steam, new programming approaches are being pursued with the goal of greater hardware performance with less coding. The Defense Advanced Projects Research Agency is launching a new programming effort aimed at leveraging the benefits of massive distributed parallelism with less sweat. Read more…

By George Leopold

AMD Launches Epyc Rome, First 7nm CPU

August 8, 2019

From a gala event at the Palace of Fine Arts in San Francisco yesterday (Aug. 7), AMD launched its second-generation Epyc Rome x86 chips, based on its 7nm proce Read more…

By Tiffany Trader

D-Wave’s Path to 5000 Qubits; Google’s Quantum Supremacy Claim

September 24, 2019

On the heels of IBM’s quantum news last week come two more quantum items. D-Wave Systems today announced the name of its forthcoming 5000-qubit system, Advantage (yes the name choice isn’t serendipity), at its user conference being held this week in Newport, RI. Read more…

By John Russell

Ayar Labs to Demo Photonics Chiplet in FPGA Package at Hot Chips

August 19, 2019

Silicon startup Ayar Labs continues to gain momentum with its DARPA-backed optical chiplet technology that puts advanced electronics and optics on the same chip Read more…

By Tiffany Trader

Crystal Ball Gazing: IBM’s Vision for the Future of Computing

October 14, 2019

Dario Gil, IBM’s relatively new director of research, painted a intriguing portrait of the future of computing along with a rough idea of how IBM thinks we’ Read more…

By John Russell

Leading Solution Providers

ISC 2019 Virtual Booth Video Tour

CRAY
CRAY
DDN
DDN
DELL EMC
DELL EMC
GOOGLE
GOOGLE
ONE STOP SYSTEMS
ONE STOP SYSTEMS
PANASAS
PANASAS
VERNE GLOBAL
VERNE GLOBAL

Intel Confirms Retreat on Omni-Path

August 1, 2019

Intel Corp.’s plans to make a big splash in the network fabric market for linking HPC and other workloads has apparently belly-flopped. The chipmaker confirmed to us the outlines of an earlier report by the website CRN that it has jettisoned plans for a second-generation version of its Omni-Path interconnect... Read more…

By Staff report

Kubernetes, Containers and HPC

September 19, 2019

Software containers and Kubernetes are important tools for building, deploying, running and managing modern enterprise applications at scale and delivering enterprise software faster and more reliably to the end user — while using resources more efficiently and reducing costs. Read more…

By Daniel Gruber, Burak Yenier and Wolfgang Gentzsch, UberCloud

Dell Ramps Up HPC Testing of AMD Rome Processors

October 21, 2019

Dell Technologies is wading deeper into the AMD-based systems market with a growing evaluation program for the latest Epyc (Rome) microprocessors from AMD. In a Read more…

By John Russell

Intel Debuts Pohoiki Beach, Its 8M Neuron Neuromorphic Development System

July 17, 2019

Neuromorphic computing has received less fanfare of late than quantum computing whose mystery has captured public attention and which seems to have generated mo Read more…

By John Russell

Rise of NIH’s Biowulf Mirrors the Rise of Computational Biology

July 29, 2019

The story of NIH’s supercomputer Biowulf is fascinating, important, and in many ways representative of the transformation of life sciences and biomedical res Read more…

By John Russell

Xilinx vs. Intel: FPGA Market Leaders Launch Server Accelerator Cards

August 6, 2019

The two FPGA market leaders, Intel and Xilinx, both announced new accelerator cards this week designed to handle specialized, compute-intensive workloads and un Read more…

By Doug Black

With the Help of HPC, Astronomers Prepare to Deflect a Real Asteroid

September 26, 2019

For years, NASA has been running simulations of asteroid impacts to understand the risks (and likelihoods) of asteroids colliding with Earth. Now, NASA and the European Space Agency (ESA) are preparing for the next, crucial step in planetary defense against asteroid impacts: physically deflecting a real asteroid. Read more…

By Oliver Peckham

When Dense Matrix Representations Beat Sparse

September 9, 2019

In our world filled with unintended consequences, it turns out that saving memory space to help deal with GPU limitations, knowing it introduces performance pen Read more…

By James Reinders

  • arrow
  • Click Here for More Headlines
  • arrow
Do NOT follow this link or you will be banned from the site!
Share This