AWS’ big bet on Arm just got real

The Big Story

AWS throws more chips on the table

The first real experiment with alternative processors in the cloud-computing era kicked off this week after AWS made its second-generation Graviton processors generally available to customers, six months after introducing the chips at re:Invent 2019. The modern cloud was built on Intel's x86 server processors, but the Graviton chips could change the way cloud customers think about building their applications.

• Chips for PCs and servers built by Intel (and AMD) use the x86 instruction set, which is more or less a blueprint for how those chips should do their work. This instruction set has been around since the late 1970s, and an enormous amount of the world's software was written with it in mind.
• Almost all chips built for smartphones, on the other hand, use the Arm instruction set for their marching orders. Arm-based chips have long been intriguing to data-center customers because of their power efficiency, but so far haven't even approached the performance levels required to be a true server alternative.

But the new AWS EC2 M6g instances (a name only an engineer could love) do use the Arm instruction set. And AWS says the instances could provide a "40% improvement on cost-performance ratio" compared to other general-purpose x86 instances available from the company. Application performance might be the ultimate expression of "your mileage may vary," but for some customers the new instances could provide big cost savings.

As is often the case, though, there is a catch.

• Some work will be required to make applications ready for Arm-based instances. Depending on the scope of that work, the cost savings might be moot.
• AWS said that applications based on interpreted languages — such as Java, NodeJS, Python, Go and others — should "run unmodified on M6g instances." But as we've learned over the last few months, there are tons of applications built around far older technologies in data centers, and they'll require extra work to get running.
• Applications that require advanced levels of compute performance or memory aren't suitable for the instances released this week. But AWS said it plans to release Graviton-based instances that are suitable for those categories in the future.

Still, this is potentially an interesting inflection point for cloud customers.

• Intel has dominated the data center market for generations of software developers and enterprise tech customers, and any competition is healthy.
• At the same time, cloud pricing looks set to get competitive again. A ceasefire in the pricing wars from years ago appears to be ending, with Oracle practically giving away cloud business in an attempt to gain ground and AWS changing the way it handles data-transfer pricing in some regions and cutting overall pricing for some customers.
• If price really does reassert itself as a competitive force heading into a very uncertain economic period and Graviton's potential cost-savings actually materialize, the new Arm instances could be a serious competitive advantage for AWS: No other cloud provider has yet released even a first-generation Arm server chip.

It's likely that the new instances will appeal mostly to startups or small companies without a lot of technical debt, or skunkworks projects inside big companies that have freedom to experiment.

• But remember: Those are the same folks who built AWS into the powerhouse it is today.

A MESSAGE FROM NASDAQ

Tailored to meet client demand, the Nasdaq Cloud Data Service (NCDS) provides real-time streaming of exchange, index, fund and analytic data. Data is made available through a suite of APIs, allowing for effortless integration and a dramatic reduction in time to market for customer-designed applications.

www.Nasdaq.com/Cloud-Data-Service

This Week in Protocol

Apple harvest: It was always weird that Apple's obsession with detail never extended to its dysfunctional web and tech infrastructure operations, but it sounds like that's changing. Apple has hired some serious heavy-hitters in the cloud and enterprise open-source space over the last few months, including ex-Docker engineer Michael Crosby, who one source told me "is who we can thank for containers as they exist today."

IP-Nope: Snowflake CEO Frank Slootman alluded to this in an interview earlier this year, but my colleague Biz Carson confirmed it in her new Protocol Pipeline newsletter: "The IPO market is dead for 2020." Cloud companies might be an exception toward the end of the year if demand for their services continues to increase along with the growth in remote work, but those companies should also find it easier to raise additional private funds than other startups.

High finance: This week's Protocol Braintrust built on one of my stories from earlier this year on how financial services companies are realizing that they need to get moving when it comes to modernizing their tech infrastructure. What's holding them back? Several experts weighed in.

Five Questions For...

Jennifer Tejada, CEO of PagerDuty

Thanks to everyone who responded with suggestions for great people to include in this new Q&A section. I received a ton of great ideas and will reach out individually over the coming weeks. New ideas are always welcome: Email me at cloud@protocol.com.

Anyway, this week it's five questions for Jennifer Tejada, CEO of PagerDuty.

What was your first tech job?

My first gig in tech was as the head of global marketing for i2, which at the time was a hyper-growth supply-chain automation software company that transformed manufacturing. I was fresh out of P&G with a consumer-brand marketing and sales background and thrown directly into the deep end, marketing the financial value proposition for complex technology to C-suite executives from Japan to Paris and everyplace in between.

What has changed the most at your company over the past two months?

We are accelerating components of our strategic plan in service of our customers. Our platform was built to automate and orchestrate distributed, mission-critical unplanned work so we are benefiting from some tailwinds in the current environment: cloud scaling, accelerated digitization of business models and customer service.

I know from previous recessions that companies like ours, cost-efficient market leaders with strong balance sheets, find opportunities to build market share. On the flip side, we are also ruthlessly prioritizing, saying 'no' to things that are not a high priority, or killing projects that just months ago seemed paramount.

Will the pandemic usher in a new era of remote working, or will we all come back together when it is safe to do so?

We are living in an incredible human experiment that will leave the workforce forever changed post COVID-19. Innovation we thought was five years away is five months away now. I cannot imagine going back to the old way of doing things and we are working hard to make sure we can embrace all the new muscle we have built as a 100% remote company.

What's the best piece of advice you could give to someone starting their first tech job?

Listen with curiosity and experiment both in your role and with your career — while many will encourage you to specialize, try roles across the business that build your understanding of how it all comes together.

Mac or PC?

Mac — I am all in at home and at work. Except for Excel — I love the marriage of a solid PC and a huge spreadsheet for analysis.

Around the Cloud

• Someone tripped over a patch cable at Slack, or something, which led to a prolonged outage late in the West Coast afternoon on Tuesday before it was resolved a few hours later.
• IBM's fleet of quantum computers has grown to 18, which is more than most of its rivals looking to perfect quantum computing and roughly 18 more than the market needs right now.
• Sales of conventional servers are skyrocketing, up 30% in the first quarter of the year, as providers scrambled to increase capacity, according to The Register.
• Amid that surge in demand, the Trump administration and several chip companies including Intel are reportedly talking about finding ways to build more factories in the U.S.
• And talking of production, some Intel factory employees are worried about coronavirus. At one of its chip fabs in Arizona, workers complained that the company has not provided enough protective equipment and has not isolated infected employees.
• What does "hyperscale" really mean? Rick Altherr of Oxide Computing took a stab at defining the term, which gets tossed around a lot when referring to the big cloud players but has … murky origins.
• AWS and Microsoft continue to snipe like politicians over the JEDI contract, but the CIO of the Department of Defense laid out a case for why the military wants to use a single cloud vendor for its modernization project.
• Funding deals for cloud companies continue apace despite an uncertain IPO market, and MemSQL just landed $50 million for its AI-oriented database technology.
• If, like me, you've read countless engineer blog posts from teams about their new infrastructure strategies, settle in with this hilarious sendup of the formulaic "how we did it" post from Saagar Jha.
• HPE announced another strategic realignment, blaming the pandemic for accelerating the need to become an "agile organization," which will also see CTO Mark Potter leave the company.
• Microsoft continued its work with the Open Compute Project this week, introducing some new open-source components of its server design technology and updating some existing projects.
• Atlassian snapped up a Slack-oriented help desk company cheekily titled Halp, two years after it threw in the towel on its attempt to compete directly with Slack in workplace-collaboration software.
• Chef was once an influential infrastructure-automation company but it's struggled to find its footing in the cloud era. It recently hired Guggenheim Partners to help it find new ways to raise money.