Tabrez Ahmed Khan

Software Engineer

Bengaluru, Karnataka, India19 yrs 9 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Expert in High Availability and Fault Tolerance systems.
Proven track record in enhancing network protocols.
Strong background in embedded systems and software development.

Stackforce AI infers this person is a seasoned professional in Telecommunications and Embedded Systems with a focus on High Availability and Network Protocols.

Contact

Skills

Core Skills

Embedded SystemsNetwork ProgrammingSoftware Performance EngineeringSystems ProgrammingNetwork ProtocolsSoftware DevelopmentSystems DesignNetwork Operating SystemsStorage SystemsData ManagementOperating Systems

Other Skills

AlgorithmsCC++CIFSCMPContent RepositoryData StructuresDebuggingDevice DriversDistributed SystemsEEMEVPNFreeBSD KernelFreeBSD kernelGNU Debugger

About

*. Linear Protection Switching, APS, PSC for MPLS-TP Profile (LSPs & PWEs) *. Acquainted with JunOS (Juniper's Network Operating System) & EVO (next-generation software platform) *. Acquainted with Data ONTAP 8.0 C-mode(cluster) operations *. Worked on high performance embedded system (Nexus 7000 datacenter Switch) *. Firm grasp of ipc, check-pointing & State-machines (software modelled as Finite State Machines) *. Hands-on knowledge of High Availability (HA), Process restart & ISSU (In service software upgrade) *. Worked on core infrastructure component critical to the nexus environment namely Gold-Generic Online Diagnostics & EEM-Embedded Event Manager(Fault Manager on IOS) *. Well acquainted with system programming & debugging in an embedded environment Specialties: ========= 1. Designing Distributed Software Modules 2. Error Detection, Fault Tolerance & Fault Analysis 3. Asynchronous (Event driven) system design 4. Process redundancy (High Availability) and Graceful Restart 5. Shell scripting

Experience

Juniper networks

3 roles

Staff Engineer

Promoted

Aug 2020 – Present · 5 yrs 7 mos · Bengaluru, Karnataka, India · On-site

I currently develop & maintain L3 forwarding features for a home-grown custom ASIC known as "Paradise", spawned from Juniper's pipeline based Express Architecture. This ASIC is shipped on PTX & QFX product lines. The software stack used to program the data-plane is essentially a micro-kernel with the adjoining platform SDK and a slim userspace. It's a co-operative multi-tasking OS with run-to completion semantics & shared address space for all threads. It also supports basic Inter-Task Communication & Synchronization primitives. I routinely deal with features such as V4/V6 IP transit traffic stats, Mcast, MoFRR, Filter based GRE Tunnels, Telemetry via Jvision, VLAN, EVPN and have developed naive understanding of the aforementioned technologies.

L3 forwarding featurescustom ASICmicro-kernelInter-Task CommunicationVLANEVPN+2

Software Engineer Staff

Jul 2018 – Jan 2019 · 6 mos

Software Engineer, Junos Core - Kernel Development Team

Sep 2012 – Jan 2019 · 6 yrs 4 mos

Currently employed with High Availability Team.As an Engineering organization we are responsible for common code development across Juniper's product line inclusive of Spanning,Routing,Switching & Security Gear.Hence, the development within Junos Core tends to have maximum impact, being leveraged in the largest core routers, services edge, branch offices & across networks worldwide. Junos is the network operating system that drives most of Juniper's platform & is primarily based on FreeBSD Kernel with access to user-space(Unix shell).The OS adheres to industry standards for routing, MPLS & HA mechanisms such as Graceful Switchover/Restart.
HA is System Design approach & associated implementation to ensure a certain degree of operational performance is unconditionally met.It's the basic tenet for achieving Non Stop Active routing.Juniper's routing platform tend to fall under "Redundant component architecture" i.e; additional hardware within the same chassis to achieve resiliency in event failure.There are 2 RE's(Routing Engine/Control plane) in a single chassis.Each one of them independently run's a Junos image.This allows for the PFE(Packet Forwarding Engine/Data plane) to continue forwarding, agnostic of the operational state of the RE.
Critical attributes are portrayed as kernel objects such as Interfaces(physical/logical), NextHop, Routes, Arp cache & so on.These were injected into the Master RE kernel via user configuration or Routing protocols.The kernel maintains this chain of objects in temporal order but with hierarchy/dependencies amongst them.The communication between the kernel & user-space peer's occurs over rtsock(routing socket).To achieve Graceful RE Switchover, specialized kernel services periodically checkpoint configuration database & Kernel states on standby RE.Specialized infra written in FreeBSD kernel assists in state propagation & maintenance on the Slave RE.Across switchover, the PFE's are only required to reconnect to the New Master RE

High AvailabilityFreeBSD KernelRouting ProtocolsGraceful SwitchoverSystems DesignNetwork Operating Systems

Tejas networks

Staff Software Engineer

Jan 2019 – Aug 2020 · 1 yr 7 mos · Bengaluru Area, India

Currently, as part of the platform Performance & Scaling pursuit, I'm working towards improving overall Scale, Performance, Throughput & User Experience. Different techniques have been deployed such as effective utilization of underlying SMP architecture, multi-threading code-paths that are discreet & independent, increasing degree of Parallelization, delayed device initialization for faster boot-up, migration to binary logging & upgrading tool-chain to newer C++ standard.
Source is constantly scouted to identify hot-spots & bottlenecks, lock ordering issues, lock contention, dismantling a coarse grained lock to multiple fine-grained locks, using atomic operations where possible, replacing blocking calls with async variants without jeopardizing correctness & ensuring async signal safety of Signal Handlers.
=====================================
As my first deliverable, I've done end-2-end design & implementation of enhancing Linear Protection Switching feature for MPLS-TP(IP/MPLS) environment to incorporate 1:N Protection Switching for MPLS Tunnels. Prior to my work, TejNeOS (Tejas Network OS) supported Linear Protection Switching (LPS - RFC: 6378) flavor wherein the number of Protect/Work Entities was restricted to 1 in each Protected Tunnel Group. This involved making numerous changes to practically every layer of the L2 Stack (L2 Services Manager, ProtectionMgr, BCM HAL), including the UI to facilitate user configuration. This provides for a rapid but simple protection switching scheme. Given that there is only 1 Protect entity against N failures of Work, protection is on Best-Effort basis.

Performance improvementSMP architecturemulti-threadingParallelizationbinary loggingSoftware Performance Engineering+1

Netapp

Member Technical Staff

Jan 2011 – Sep 2012 · 1 yr 8 mos · Bangalore

In recent years, we have been seeing an explosive growth of unstructured data which cannot be addressed by traditional container-based storage/file system.Content Repository tries to address this by providing the notion of "Infinitely scalable boundless NAS container" providing one flat namespace for all user-content
The ever-pervasive nature of internet has led user's storing & retrieving data over the web using HTTP protocol apart from conventional NAS access(NFS, CIFS). The scope of such data is far beyond simple text contents & forays in complex types such as mp3, medical scans, jpeg & so forth.Such complex attributes associated with data can be used to query & retrieve the same. Content Repository models the same by introducing the novel concept of "Objects" where plain data is associated with its attributes or formally, meta-data
This Object notion allows for Unified Object Storage accessed over CDMI protocol (modified HTTP for Cloud environment) while extending the ONTAP scale-out unified architecture. Its cluster mode personality abstracts artificial storage constructs such as raid aggregates, volumes & simplifies management through an sophisticated policy engine where the user only see's high level policies applied to arbitrary object sets & not eccentric storage constructs. It also facilitates dynamic & transparent object migration
I am currently involved in enabling Multi-Protocol NAS Access. In context of file services to either Windows or Unix clients, this allows for uniform file access permissions to Windows & Unix users. Thus providing multiprotocol data access that transcends the enclosed perspective of general-purpose OS.It automatically maps Windows users to Unix users & vice-versa on the fly, simplifying the unification of the two separate namespaces. ACL(Access Control list) management is now seamless on both Windows & Nfsv4 clients. Windows clients can now manage Nfsv4 Acls over CIFS & so can the Unix clients do the same over Nfsv4 to NT Acls

Content RepositoryMulti-Protocol NAS AccessHTTP protocolNFSCIFSStorage Systems+1

Cisco systems

2 roles

Software Engineer

Jul 2006 – Jan 2011 · 4 yrs 6 mos

Development on Cisco's next generation Data Center Operating System called NxOs (linux centric) used on the Nexus 7000 data center platform. Nexus 7000 Series Switches are high-performance switching system which offer comprehensive, one-platform solution for data center core network. It's also characterized by aggregation, high-density, scalability & high resilience.
Prime Responsibilities
> Design & Development of Feature - from inception to Customer shipment in a timely manner
> Providing support in case of Customer issue's
> Providing guidance to QA/Systest
> Providing long term maintenance of features
> Catering to Customer request for enhancements
> Mentoring New College Graduates
Key areas of Work
> EEM (Embedded Event Manager)
> Gold (Generic OnLine Diagnostic)
> Bug Fixing on NetStack
> Bug Fixing on CMP
Niche areas of expertise (In NxOS environment)
> PSS - Persistence Storage System (Key-Value database for achieving State Persistence for HA across dual Supervisor cards)
> MTS - Message Transaction Services (Primary IPC mechanism in NxOS)
> Capabilities - Cross verifying feature parity across releases during ISSU
> FIPS - Fault Injection for effective Unit Testing for excercising error code paths
> TLV - Type, Length & Value container for data during message passing

NxOSEEMGoldNetStackCMPOperating Systems+1

Student Intern

Feb 2006 – Jul 2006 · 5 mos

Worked on fault analysis of an upcoming Leadership based Election protocol in a San Cache/Cluster environment. The concept was eventually productized.