Weka.IO – a stealthy startup swimming against the tide

Israeli firm wants to kick serious file system ass


Analysis Liran Zvibel sits across the table from me in a Tel Aviv restaurant eating pork chops. This is a man who doesn't mind going against the grain. He's co-founder and Chief Technical Officer of Weka.IO, a stealthy startup that is swimming against two perceived tides.

Storage folklore these days says that file systems generally can't scale and you need object storage to handle unstructured data at the billions-of-files level. This is a premise that Cleversafe, Scality, Amplidata, Caringo and Dell EMC's Atmos and EC2 have supported.

A second thing storage folks will say is that HPC-class file systems need special, parallel-access software such as Spectrum Scale, IBM's rebranded GPFS.

Another piece of storage wisdom is that, for good performance, a shared file system needs dedicated hardware; Isilon filers are the classic example.

Weka.IO says different. File systems can scale out to the billions-of-files level, and can do so using shared hardware, just needing a core on multi-core servers and one or more directly addressed SSDs. Its software aggregates many servers with their Weka-dedicated core and flash storage into a virtual filer – think VNAS conceptually similar to a VSAN, a virtual SAN.

That's swimming against the tide big time.

Technology glimpse

The software runs insde a virtual machine or container on its host server. Its file services communicate with applications and the file system manages clustering, data distribution, erasure coding, snapshots, rebuilds, and tiering. Object connector tiers cold data off to the cloud and this works with any S3 or Swift object store. It's qualified with HGST, Scality, IBM COS (Cleversafe), and Cloudian, and is policy-driven.

The founders say their software can handle billions of unstructured data files, hundreds of petabytes and millions of IOPS, with linear scalability through adding compute (cores) and capacity (SSDs). The system also has sub-millisecond latency (300 microseconds for reads and writes) and it's claimed to have cloud economics, with Posix semantics at cloud-scale.

A huge amount of development effort has been focussed on erasure coding making it more efficient in terms of minimising the space overhead. Weka.IO does distributed erasure coding to +4 (16 +4) = 20 stripes, meaning a 20 per cent overhead. Erasure coding is the foundational piece of the product code, and some of this technology is already patented.

In general the software has been designed for flash. Weka.IO says Isilon software has a disk-focus and even Qumulo, the scalable filer startup, has a disk focus for capacity with flash only being used for metadata storage.

Encryption is coming for object store tiering, and compression and dedupe are on the roadmap for SSDs.

Founders and funding

Weka.IO has three co-founders who set up the firm in 2013:

  • Liran Zvibel – Chief Technology Officer
  • Omri Palmon – Chief Product Officer
  • Maor Ben-Dayan – Chief Architect

They all came from IBM-acquired XIV, Moshe Yanai's second storage array company, and were somewhat disillusioned during IBM's ownership by what they perceived as missed opportunities for developing the product. Starting Weka.IO is their way of doing it right, focussed on two principles: it's a software-defined world and commodity hardware rules.

There have been two funding rounds: a $10 million A-round in January 2014 and a $22.25 million B-round in December 2015, which amounts to serious cash for a software company. They knew from the start that they had to crack the US market so opened a San Jose office and hired a professional, seasoned CEO, Michael Raam, in August last year. He was Sandforce's CEO when LSI bought that flash controller business.

Raam hired Barbara Murphy as VP Marketing. She comes from HGST (marketing Amplidata kit) with a three-year stint as Chief Marketing Officer at Panasas (HPC filer) before that. He has just hired Richard Dyke as VP of Sales, coming from Hedvig.

There are 38 employees, which should rise to 44 by the end of the year. We can expect a formal stealth exit in spring next year – an appropriate time of year for growth to emerge.

And what does Weka.IO mean? The .IO part is straightforward, but Weka? Wikipedia says it's a flightless bird from New Zealand, but that seems to be the wrong kind of idea. Could it be "We Kick Ass?" No. Zvibel says: "Weka is 1030 (peta, exa, zeta, yota, hela, weka) or a NZ bird – pick which one you want. It's going to take a while until we reach weka scale."

Weka.IO has convinced its VCs that it has seriously good technology under development, and execs are suppressing their excitement about POCs progressing towards orders.

This currently flightless bird could take off. We'll see how it looks as it exits stealth and moves to the runway in May next year. ®

Similar topics


Other stories you might like

  • Carnival Cruises torpedoed by US states, agrees to pay $6m after waves of cyberattacks
    Now those are some phishing boats

    Carnival Cruise Lines will cough up more than $6 million to end two separate lawsuits filed by 46 states in the US after sensitive, personal information on customers and employees was accessed in a string of cyberattacks.

    A couple of years ago, as the coronavirus pandemic was taking hold, the Miami-based biz revealed intruders had not only encrypted some of its data but also downloaded a collection of names and addresses; Social Security info, driver's license, and passport numbers; and health and payment information of thousands of people in almost every American state.

    It all started to go wrong more than a year prior, as the cruise line became aware of suspicious activity in May 2019. This apparently wasn't disclosed until 10 months later, in March 2020.

    Continue reading
  • India extends deadline for compliance with infosec logging rules by 90 days
    Helpfully announced extension on deadline day

    India's Ministry of Electronics and Information Technology (MeitY) and the local Computer Emergency Response Team (CERT-In) have extended the deadline for compliance with the Cyber Security Directions introduced on April 28, which were due to take effect yesterday.

    The Directions require verbose logging of users' activities on VPNs and clouds, reporting of infosec incidents within six hours of detection - even for trivial things like unusual port scanning - exclusive use of Indian network time protocol servers, and many other burdensome requirements. The Directions were purported to improve the security of local organisations, and to give CERT-In information it could use to assess threats to India. Yet the Directions allowed incident reports to be sent by fax – good ol' fax – to CERT-In, which offered no evidence it operates or would build infrastructure capable of ingesting or analyzing the millions of incident reports it would be sent by compliant organizations.

    The Directions were roundly criticized by tech lobby groups that pointed out requirements such as compelling clouds to store logs of customers' activities was futile, since clouds don't log what goes on inside resources rented by their customers. VPN providers quit India and moved their servers offshore, citing the impossibility of storing user logs when their entire business model rests on not logging user activities. VPN operators going offshore means India's government is therefore less able to influence such outfits.

    Continue reading
  • Hangouts hangs up: Google chat app shuts this year
    How many messaging services does this web giant need? It's gotta be over 9,000

    Google is winding down its messaging app Hangouts before it officially shuts in November, the web giant announced on Monday.

    Users of the mobile app will see a pop-up asking them to move their conversations onto Google Chat, which is yet another one of its online services. It can be accessed via Gmail as well as its own standalone application. Next month, conversations in the web version of Hangouts will be ported over to Chat in Gmail. 

    Continue reading

Biting the hand that feeds IT © 1998–2022