By Richard J. Cichelli, the newspaper industry’s forgotten computer scientist
Times-Shamrock took a big step when it replaced system X, the popular application it had used for years for ad tracking. Its replacement is an innovative, low-cost, small computer solution from Software Consulting Services, LLC (SCS).
Times-Shamrock Creative Services (TSCS) is a wholly owned subsidiary of the Times-Shamrock Communications (TSC) company. TSC is a diversified, family-owned media company located in Scranton PA. TSCS serves the ad building and pagination needs of the Times-Shamrock newspapers and 24 other independent publications across North America.
SCS helped TSCS reduce their ad tracking system and management costs by more than 50%. User reaction has been uniformly positive with everyone saying the new system, SCS/Track, was more capable and easier to use. The switch was accomplished without any loss of functionality.
How was this possible? SCS doesn't follow conventional computing wisdom. Its innovations in system engineering and support set it apart- so much so that customers of other systems can’t wait to make the switch. Abandoning legacy systems isn’t always easy, but nearly all those who do say that SCS's solutions are better than they had ever hoped for.
SCS accomplishes this by disappointing players who are behind the curve. It has a 35 year history of doing this. SCS disappointed server system vendors when it replaced Digital Equipment Corporation's (DEC) servers and operating systems (OS) with low-cost, Unix-based servers from Dell Computer starting in 1987. SCS then disappointed OS vendors SCO and Microsoft by adopting free and open source Linux as its preferred operating system in 1996.
So what does an innovative contrarian do when serving an industry facing tough economic challenges? SCS knows that cost savings are much more easily realized than revenue growth, especially for newspapers. It's a shame, but it's true.
This is the cost saving platform SCS installed at TSCS.
It is typical of current SCS solutions. There are five NUCs (what Intel calls its “Next Unit of Computing”). Primary, live stand-by secondary, live remote disaster recovery pair and a test server. All the software used is either free open source or SCS intellectual property. This includes both software tools and application code. There are no proprietary third-party operating systems, databases or virtualization systems.
SCS/Track was brought online at TSCS on Monday March 19, 2018. Training for all users began on the prior Thursday. There was no extensive parallel running. At mid-afternoon Tuesday there were over 180 active users distributed over 25 newspaper sites working on SCS/Track.
A single primary NUC server at TSCS supported all these users. An average 1.5 threads (out of four) were active. (Intel Core i5.) Memory use topped out at 3GB (out of 32GB available.) All the primary active storage is 4TB of SSDs in each NUC. Two 12TB Synology Network Attached Storage (NAS) devices are used for long-term archiving (with full text indexing and natural language querying). The multi-level storage is managed in a way invisible to users.
Using SSDs in very active servers is tricky. Heavy swapping or paging to SSD virtual memory can exceed the recommended maximum rewrites for SSDs. SCS disables swapping and paging to carefully manage dynamic memory usage. The good part is that SSDs are at least three times as fast as the fastest rotating hard disk drives, even those with local disk caches and RAID. This use of SSDs offers faster and more reliable storage.
At TSCS, new servers leased from SCS were delivered to replace the “X” system’s servers. Glen Knarr showed SCS VP Kurt Jackson to the computer room.
"See, we’ve cleared a space for your racks of servers. When will they be delivered?" asked Glen.
"Glen", Kurt said, "We don't need anything like that much space. I have all five servers in this box. Another box holds the five uninterruptible power devices. A small table or shelf is all they’ll need."
Glen realized immediately that SCS's approach to systems upends the supply chain, and that fact would save both SCS and TSCS money.
SCS leases its systems to customers.
The lease payments cover managed services. Customers take care of their desktops and mobile devices and SCS does all the rest, including all server hardware and system and application software.
Servers are actively monitored 24/7 with automated check-ins every 15 minutes. Mirroring on shared-nothing servers provides nearly bullet-proof backup. In the rare situation where a NUC fails (only one has of the nearly 100 SCS has installed in the last two years.), SCS has spare ones ready to ship at a moment's notice.
SCS seeks to keep what is paid to third parties as low as possible. The savings can be passed onto its customers. That's what happened at TSCS and for many other SCS customers. The savings can be quite large.
Intel NUCs are sold as bare-bones devices with a three year warranty. SCS adds whatever else is needed. With managed services, their maintenance, which is very little, is SCS's responsibility.
Would Intel prefer to sell big server systems? Of course. These are Intel's most profitable sale. NUCs are a volume market for them. Gamers, for example, love NUCs for their cheap and reliable computing power. Demand assures that NUCs are quickly updated with the newest Intel chips, fastest processors and built-in reliability necessary for a successful volume business. It is a perfect commodity market for Intel.
Special engineering is necessary to take advantage of the unique capabilities provided by NUCs. Fortunately, this involves no more than current best practices of software engineering. SCS builds systems both large and small out of components that are tightly integrated but loosely coupled. This recent software engineering paradigm is called a microservices architecture.
Systems can get bigger and more functional in a number of ways, not all of them good. Systems built on a single monolithic database (an advertising database, or ad base, if you will) stagnate due to the complexity of this approach.
A microservices architecture is composed of modules that operate as services acting on their own data and none other. Network compatible software is used to make an enterprise service bus (ESB or MESB). This is used to pass data securely among the independent service modules. This independence makes integrating services provided by third-parties much easier. Update one, either from SCS or others, and it still works properly with the others.
The ESB manages the locations of microservices. A service can run on any processor accessible anywhere over the network. To make a more powerful platform, simply add NUCs- as many as one might wish. (SCS calls its ESB, MOAI - Mother Of All Interfaces.)
A microservice architecture is inherently more agile. SCS customers get more frequent updates (dozens to hundreds per year), which are nearly all included as part of SCS's managed services.
Transitioning to Microservices.
A system based on the microservice architecture (MA) is inherently modular since MA is a type of modularization. There are just additional rules for it. One can write applications that are well built using a modular style. However, to get the benefits of MA, one needs to build components using rules that make tight integration and loose coupling work better. Modules must function as independent services. They must own their own data. Sharing data among modules is done through an intermediary, the MESB, which allows sharing without one service knowing how another service does its job. A microservice module should be able to be replaced without changing other modules.
A more concrete example.
Layout-8000 version 16 is being released. Going from V15.x to V16 means there is a significant change in Layout's system architecture. Layout-8000 users will all get new functionality with the release. However, a lot of the changes in version 16 are under the skin.
StandbyAdBoss (SAB), which, prior to V16, was an optional, extra-cost add on, is now bundled with the distribution at no additional cost. The new SAB is particularly important since it is now configured as a microservice. Previously, it was internal to the Layout-8000 main code and accessed Layout's ad database, and its availability to a site was controlled by a switch. Now, all that functionality and complexity has been factored out. Room has been made for easily attaching additional microservices. Updating SAB now doesn't require updating Layout-8000 or vice versa.
What if you are already committed to a cloud or virtualized architecture?
Many of our customers already are. SCS's applications run in these contexts very easily. The modular and platform independent architecture used for all SCS's systems is just better with microservices.
In these times, few would argue that a monolithic approach is better. In fact, there are still some architectures used by newspaper vendors that one should be avoiding wherever possible.
Connecting remote desktops to mainframe-based applications via Citrix seems to be one of these. A Citrix server farm may consume as much computing resources (expense) as simply running the application by itself.
Would it not be cool if you could access SCS's applications remotely from within a browser?
…Especially if the speed seemed like that of a local connection? This is coming very soon. No other method of remote access provides such a level of convenience. PuTTY and other connecting software will be replaced.
SCS deploys its systems on a number of the cloud-based platforms its customers have. However, SCS's preference is for NUCs and the flexibility and cost savings they provide as local appliances.
NUCs and microservices have the power, low cost, and flexibility needed to provide timely, innovative solutions. This was the case at TSCS, where corporate-wide consolidation of ad building yielded significant savings.
Whether you’re thinking of public cloud deployments (the cloud is just someone else's computer) or virtualized systems, check the expense and complexity of supporting them. SCS has found better solutions. Contact Phil (email@example.com or 610-746-7700) and let us prove it to you.
You won’t be disappointed.
Richard J. Cichelli
Send email to me to comment
What is implied by a "doing more with less" philosophy?
What constitutes less?
Picking appropriate technology enables business transformations
Paying lots to share computing power in a complex way?
A machine intelligence revolution that will rival the industrial revolution? (Warning. About 4 minutes.)
Blueprint for SCS enhancement of its search technology